Annotation of researchv10no/sys/io/bio.c, revision 1.1.1.1

1.1       root        1: #include "sys/param.h"
                      2: #include "sys/user.h"
                      3: #include "sys/buf.h"
                      4: #include "sys/conf.h"
                      5: #include "sys/proc.h"
                      6: #include "sys/pte.h"
                      7: #include "sys/vm.h"
                      8: 
                      9: /*
                     10:  * The following several routines allocate and free
                     11:  * buffers with various side effects.  In general the
                     12:  * arguments to an allocate routine are a device and
                     13:  * a block number, and the value is a pointer to
                     14:  * to the buffer header; the buffer is marked "busy"
                     15:  * so that no one else can touch it.  If the block was
                     16:  * already in core, no I/O need be done; if it is
                     17:  * already busy, the process waits until it becomes free.
                     18:  * The following routines allocate a buffer:
                     19:  *     getblk
                     20:  *     bread
                     21:  *     breada
                     22:  *     baddr   (if it is incore)
                     23:  * Eventually the buffer must be released, possibly with the
                     24:  * side effect of writing it out, by using one of
                     25:  *     bwrite
                     26:  *     bdwrite
                     27:  *     bawrite
                     28:  *     brelse
                     29:  */
                     30: 
                     31: struct buf bfreelist[BQUEUES];
                     32: struct buf bswlist, *bclnlist;
                     33: 
                     34: extern int bufhcnt;
                     35: extern struct bufhd bufhash[];
                     36: #define        BUFHASH(dev, dblkno)    \
                     37:                ((struct buf *)&bufhash[((int)(dev)+(int)(dblkno)) % bufhcnt])
                     38: 
                     39: /*
                     40:  * Initialize hash links for buffers.
                     41:  */
                     42: bhinit()
                     43: {
                     44:        register int i;
                     45:        register struct bufhd *bp;
                     46: 
                     47:        for (bp = bufhash, i = 0; i < bufhcnt; i++, bp++)
                     48:                bp->b_forw = bp->b_back = (struct buf *)bp;
                     49: }
                     50: 
                     51: #define        DISKMON 1
                     52: 
                     53: #ifdef DISKMON
                     54: struct {
                     55:        long    ndwrite;        /* bdwrite, not real io */
                     56:        long    nread;          /* someone's waiting for this one */
                     57:        long    nreada;         /* readahead from disk */
                     58:        long    ncache;         /* blocks found in cache, so not read */
                     59:        long    nwrite;         /* acutally written */
                     60:        long    bufcount[64];
                     61: } io_info;
                     62: #endif
                     63: 
                     64: #define CECMON 0
                     65: 
                     66: #define CFS 64
                     67: 
                     68: #ifdef CECMON
                     69: struct {
                     70:        long    cread;          /* someone's waiting for this one */
                     71:        long    creada;         /* readahead from disk */
                     72:        long    ccache;         /* blocks found in cache, so not read */
                     73:        long    cwrite;         /* acutally written */
                     74: } cec_info[CFS];
                     75: #endif
                     76: 
                     77: 
                     78: /*
                     79:  * Swap IO headers -
                     80:  * They contain the necessary information for the swap I/O.
                     81:  * At any given time, a swap header can be in three
                     82:  * different lists. When free it is in the free list, 
                     83:  * when allocated and the I/O queued, it is on the swap 
                     84:  * device list, and finally, if the operation was a dirty 
                     85:  * page push, when the I/O completes, it is inserted 
                     86:  * in a list of cleaned pages to be processed by the pageout daemon.
                     87:  */
                     88: extern struct buf swapbuf[];
                     89: extern struct swapinfo swapinfo[];
                     90: 
                     91: #ifndef        UNFAST
                     92: #define        notavail(bp) \
                     93: { \
                     94:        int s = spl6(); \
                     95:        (bp)->av_back->av_forw = (bp)->av_forw; \
                     96:        (bp)->av_forw->av_back = (bp)->av_back; \
                     97:        (bp)->b_flags |= B_BUSY; \
                     98:        splx(s); \
                     99: }
                    100: #endif
                    101: 
                    102: /*
                    103:  * Read in (if necessary) the block and return a buffer pointer.
                    104:  */
                    105: struct buf *
                    106: bread(dev, blkno)
                    107: dev_t dev;
                    108: daddr_t blkno;
                    109: {
                    110:        register struct buf *bp;
                    111: #ifdef  CECMON
                    112:        register int sub;
                    113: #endif
                    114: 
                    115:        bp = getblk(dev, blkno);
                    116:        if (bp->b_flags&B_DONE) {
                    117: #ifdef DISKMON
                    118:                io_info.ncache++;
                    119: #endif
                    120: #ifdef CECMON
                    121:                sub = minor(dev);
                    122:                cec_info[(sub >= CFS)?(sub - CFS) : sub].ccache++;
                    123: #endif
                    124:                return(bp);
                    125:        }
                    126:        bp->b_flags |= B_READ;
                    127:        bp->b_bcount = BSIZE(dev);
                    128:        (*bdevsw[major(dev)]->d_strategy)(bp);
                    129: #ifdef DISKMON
                    130:        io_info.nread++;
                    131: #endif
                    132: #ifdef CECMON
                    133:        sub = minor(dev);
                    134:        cec_info[(sub >= CFS)?(sub - CFS) : sub].cread++;
                    135: #endif
                    136:        u.u_vm.vm_inblk++;              /* pay for read */
                    137:        iowait(bp);
                    138:        return(bp);
                    139: }
                    140: 
                    141: /*
                    142:  * Read in the block, like bread, but also start I/O on the
                    143:  * read-ahead block (which is not allocated to the caller)
                    144:  */
                    145: struct buf *
                    146: breada(dev, blkno, rablkno)
                    147: dev_t dev;
                    148: daddr_t blkno, rablkno;
                    149: {
                    150:        register struct buf *bp, *rabp;
                    151: #ifdef  CECMON
                    152:        register int sub;
                    153: #endif
                    154: 
                    155:        bp = NULL;
                    156:        if (!incore(dev, blkno)) {
                    157:                bp = getblk(dev, blkno);
                    158:                if ((bp->b_flags&B_DONE) == 0) {
                    159:                        bp->b_flags |= B_READ;
                    160:                        bp->b_bcount = BSIZE(dev);
                    161:                        (*bdevsw[major(dev)]->d_strategy)(bp);
                    162: #ifdef DISKMON
                    163:                        io_info.nread++;
                    164: #endif
                    165: #ifdef CECMON
                    166:                        sub = minor(dev);
                    167:                        cec_info[(sub >= CFS) ? (sub - CFS) : sub].cread++;
                    168: #endif
                    169:                        u.u_vm.vm_inblk++;              /* pay for read */
                    170:                }
                    171:        }
                    172:        if (rablkno && !incore(dev, rablkno)) {
                    173:                rabp = getblk(dev, rablkno);
                    174:                if (rabp->b_flags & B_DONE) {
                    175:                        brelse(rabp);
                    176:                } else {
                    177:                        rabp->b_flags |= B_READ|B_ASYNC;
                    178:                        rabp->b_bcount = BSIZE(dev);
                    179:                        (*bdevsw[major(dev)]->d_strategy)(rabp);
                    180: #ifdef DISKMON
                    181:                        io_info.nreada++;
                    182: #endif
                    183: #ifdef CECMON
                    184:                        sub = minor(dev);
                    185:                        cec_info[(sub >= CFS) ? (sub - CFS) : sub].creada++;
                    186: #endif
                    187:                        u.u_vm.vm_inblk++;              /* pay in advance */
                    188:                }
                    189:        }
                    190:        if(bp == NULL)
                    191:                return(bread(dev, blkno));
                    192:        iowait(bp);
                    193:        return(bp);
                    194: }
                    195: 
                    196: /*
                    197:  * Write the buffer, waiting for completion.
                    198:  * Then release the buffer.
                    199:  */
                    200: bwrite(bp)
                    201: register struct buf *bp;
                    202: {
                    203:        register flag;
                    204: #ifdef  CECMON
                    205:        register int sub;
                    206: #endif
                    207: 
                    208:        flag = bp->b_flags;
                    209:        bp->b_flags &= ~(B_READ | B_DONE | B_ERROR | B_DELWRI | B_AGE);
                    210:        bp->b_bcount = BSIZE(bp->b_dev);
                    211: #ifdef DISKMON
                    212:        io_info.nwrite++;
                    213: #endif
                    214: #ifdef CECMON
                    215:        sub = minor(bp->b_dev);
                    216:        cec_info[(sub >= CFS) ? (sub - CFS) : sub].cwrite++;
                    217: #endif
                    218:        if ((flag&B_DELWRI) == 0)
                    219:                u.u_vm.vm_oublk++;              /* noone paid yet */
                    220:        (*bdevsw[major(bp->b_dev)]->d_strategy)(bp);
                    221:        if ((flag&B_ASYNC) == 0) {
                    222:                iowait(bp);
                    223:                brelse(bp);
                    224:        } else if (flag & B_DELWRI)
                    225:                bp->b_flags |= B_AGE;
                    226:        else
                    227:                geterror(bp);
                    228: }
                    229: 
                    230: /*
                    231:  * Release the buffer, marking it so that if it is grabbed
                    232:  * for another purpose it will be written out before being
                    233:  * given up (e.g. when writing a partial block where it is
                    234:  * assumed that another write for the same block will soon follow).
                    235:  * This can't be done for magtape, since writes must be done
                    236:  * in the same order as requested.
                    237:  */
                    238: bdwrite(bp)
                    239: register struct buf *bp;
                    240: {
                    241: 
                    242:        if ((bp->b_flags&B_DELWRI) == 0)
                    243:                u.u_vm.vm_oublk++;              /* noone paid yet */
                    244:        if(bdevsw[major(bp->b_dev)]->d_flags & B_TAPE)
                    245:                bawrite(bp);
                    246:        else {
                    247: #ifdef DISKMON
                    248:                io_info.ndwrite++;
                    249: #endif
                    250:                bp->b_flags |= B_DELWRI | B_DONE;
                    251:                brelse(bp);
                    252:        }
                    253: }
                    254: 
                    255: /*
                    256:  * Release the buffer, start I/O on it, but don't wait for completion.
                    257:  */
                    258: bawrite(bp)
                    259: register struct buf *bp;
                    260: {
                    261: 
                    262:        bp->b_flags |= B_ASYNC;
                    263:        bwrite(bp);
                    264: }
                    265: 
                    266: /*
                    267:  * release the buffer, with no I/O implied.
                    268:  */
                    269: brelse(bp)
                    270: register struct buf *bp;
                    271: {
                    272:        register struct buf *flist;
                    273:        register s;
                    274: 
                    275:        if (bp->b_flags&B_WANTED)
                    276:                wakeup((caddr_t)bp);
                    277:        if (bfreelist[0].b_flags&B_WANTED) {
                    278:                bfreelist[0].b_flags &= ~B_WANTED;
                    279:                wakeup((caddr_t)bfreelist);
                    280:        }
                    281:        if (bp->b_flags&B_ERROR)
                    282:                if (bp->b_flags & B_LOCKED)
                    283:                        bp->b_flags &= ~B_ERROR;        /* try again later */
                    284:                else
                    285:                        bp->b_dev = NODEV;              /* no assoc */
                    286:        s = spl6();
                    287:        if (bp->b_flags & (B_ERROR|B_INVAL)) {
                    288:                /* block has no info ... put at front of most free list */
                    289:                flist = &bfreelist[BQUEUES-1];
                    290:                flist->av_forw->av_back = bp;
                    291:                bp->av_forw = flist->av_forw;
                    292:                flist->av_forw = bp;
                    293:                bp->av_back = flist;
                    294:        } else {
                    295:                if (bp->b_flags & B_LOCKED)
                    296:                        flist = &bfreelist[BQ_LOCKED];
                    297:                else if (bp->b_flags & B_AGE)
                    298:                        flist = &bfreelist[BQ_AGE];
                    299:                else
                    300:                        flist = &bfreelist[BQ_LRU];
                    301:                flist->av_back->av_forw = bp;
                    302:                bp->av_back = flist->av_back;
                    303:                flist->av_back = bp;
                    304:                bp->av_forw = flist;
                    305:        }
                    306:        bp->b_flags &= ~(B_WANTED|B_BUSY|B_ASYNC|B_AGE);
                    307:        splx(s);
                    308: }
                    309: 
                    310: /*
                    311:  * See if the block is associated with some buffer
                    312:  * (mainly to avoid getting hung up on a wait in breada)
                    313:  */
                    314: incore(dev, blkno)
                    315: dev_t dev;
                    316: daddr_t blkno;
                    317: {
                    318:        register struct buf *bp;
                    319:        register struct buf *dp;
                    320:        register int dblkno = fsbtodb(dev, blkno);
                    321: 
                    322:        dp = BUFHASH(dev, dblkno);
                    323:        for (bp = dp->b_forw; bp != dp; bp = bp->b_forw)
                    324:                if (bp->b_blkno == dblkno && bp->b_dev == dev &&
                    325:                    !(bp->b_flags & B_INVAL))
                    326:                        return (1);
                    327:        return (0);
                    328: }
                    329: 
                    330: struct buf *
                    331: baddr(dev, blkno)
                    332: dev_t dev;
                    333: daddr_t blkno;
                    334: {
                    335: 
                    336:        if (incore(dev, blkno))
                    337:                return (bread(dev, blkno));
                    338:        return (0);
                    339: }
                    340: 
                    341: /*
                    342:  * Assign a buffer for the given block.  If the appropriate
                    343:  * block is already associated, return it; otherwise search
                    344:  * for the oldest non-busy buffer and reassign it.
                    345:  */
                    346: struct buf *
                    347: getblk(dev, blkno)
                    348: dev_t dev;
                    349: daddr_t blkno;
                    350: {
                    351:        register struct buf *bp, *dp, *ep;
                    352:        register daddr_t dblkno;
                    353: #ifdef DISKMON
                    354:        register int i;
                    355: #endif
                    356: 
                    357:        if ((unsigned)blkno >= 1 << (sizeof(int)*NBBY-PGSHIFT))
                    358:                blkno = 1 << ((sizeof(int)*NBBY-PGSHIFT) + 1);
                    359:        dblkno = fsbtodb(dev, blkno);
                    360:        dp = BUFHASH(dev, dblkno);
                    361:     loop:
                    362:        (void) spl0();
                    363:        for (bp = dp->b_forw; bp != dp; bp = bp->b_forw) {
                    364:                if (bp->b_blkno != dblkno || bp->b_dev != dev ||
                    365:                    bp->b_flags&B_INVAL)
                    366:                        continue;
                    367:                (void) spl6();
                    368:                if (bp->b_flags&B_BUSY) {
                    369:                        bp->b_flags |= B_WANTED;
                    370:                        sleep((caddr_t)bp, PRIBIO+1);
                    371:                        goto loop;
                    372:                }
                    373:                (void) spl0();
                    374: #ifdef DISKMON
                    375:                i = 0;
                    376:                dp = bp->av_forw;
                    377:                while ((dp->b_flags & B_HEAD) == 0) {
                    378:                        i++;
                    379:                        dp = dp->av_forw;
                    380:                }
                    381:                if (i<64)
                    382:                        io_info.bufcount[i]++;
                    383: #endif
                    384:                notavail(bp);
                    385:                return(bp);
                    386:        }
                    387:        if (major(dev) >= nblkdev)
                    388:                panic("blkdev");
                    389:        (void) spl6();
                    390:        for (ep = &bfreelist[BQUEUES-1]; ep > bfreelist; ep--)
                    391:                if (ep->av_forw != ep)
                    392:                        break;
                    393:        if (ep == bfreelist) {          /* no free blocks at all */
                    394:                ep->b_flags |= B_WANTED;
                    395:                sleep((caddr_t)ep, PRIBIO+1);
                    396:                goto loop;
                    397:        }
                    398:        (void) spl0();
                    399:        bp = ep->av_forw;
                    400:        notavail(bp);
                    401:        if (bp->b_flags & B_DELWRI) {
                    402:                bp->b_flags |= B_ASYNC;
                    403:                bwrite(bp);
                    404:                goto loop;
                    405:        }
                    406:        bp->b_flags = B_BUSY;
                    407:        bp->b_back->b_forw = bp->b_forw;
                    408:        bp->b_forw->b_back = bp->b_back;
                    409:        bp->b_forw = dp->b_forw;
                    410:        bp->b_back = dp;
                    411:        dp->b_forw->b_back = bp;
                    412:        dp->b_forw = bp;
                    413:        bp->b_dev = dev;
                    414:        bp->b_blkno = dblkno;
                    415:        return(bp);
                    416: }
                    417: 
                    418: /*
                    419:  * get an empty block,
                    420:  * not assigned to any particular device
                    421:  */
                    422: struct buf *
                    423: geteblk()
                    424: {
                    425:        register struct buf *bp, *dp;
                    426:        register int s;
                    427: 
                    428: loop:
                    429:        s = spl6();
                    430:        for (dp = &bfreelist[BQUEUES-1]; dp > bfreelist; dp--)
                    431:                if (dp->av_forw != dp)
                    432:                        break;
                    433:        if (dp == bfreelist) {          /* no free blocks */
                    434:                dp->b_flags |= B_WANTED;
                    435:                sleep((caddr_t)dp, PRIBIO+1);
                    436:                goto loop;
                    437:        }
                    438:        (void) splx(s);
                    439:        bp = dp->av_forw;
                    440:        notavail(bp);
                    441:        if (bp->b_flags & B_DELWRI) {
                    442:                bp->b_flags |= B_ASYNC;
                    443:                bwrite(bp);
                    444:                goto loop;
                    445:        }
                    446:        bp->b_flags = B_BUSY|B_INVAL;
                    447:        bp->b_back->b_forw = bp->b_forw;
                    448:        bp->b_forw->b_back = bp->b_back;
                    449:        bp->b_forw = dp->b_forw;
                    450:        bp->b_back = dp;
                    451:        dp->b_forw->b_back = bp;
                    452:        dp->b_forw = bp;
                    453:        bp->b_dev = (dev_t)NODEV;
                    454:        bp->b_bcount = BUFSIZE;
                    455:        return(bp);
                    456: }
                    457: 
                    458: /*
                    459:  * Wait for I/O completion on the buffer; return errors
                    460:  * to the user.
                    461:  */
                    462: iowait(bp)
                    463: register struct buf *bp;
                    464: {
                    465: 
                    466:        (void) spl6();
                    467:        while ((bp->b_flags&B_DONE)==0)
                    468:                sleep((caddr_t)bp, PRIBIO);
                    469:        (void) spl0();
                    470:        geterror(bp);
                    471: }
                    472: 
                    473: #ifdef UNFAST
                    474: /*
                    475:  * Unlink a buffer from the available list and mark it busy.
                    476:  * (internal interface)
                    477:  */
                    478: notavail(bp)
                    479: register struct buf *bp;
                    480: {
                    481:        register s;
                    482: 
                    483:        s = spl6();
                    484:        bp->av_back->av_forw = bp->av_forw;
                    485:        bp->av_forw->av_back = bp->av_back;
                    486:        bp->b_flags |= B_BUSY;
                    487:        splx(s);
                    488: }
                    489: #endif
                    490: 
                    491: /*
                    492:  * Mark I/O complete on a buffer. If the header
                    493:  * indicates a dirty page push completion, the
                    494:  * header is inserted into the ``cleaned'' list
                    495:  * to be processed by the pageout daemon. Otherwise
                    496:  * release it if I/O is asynchronous, and wake 
                    497:  * up anyone waiting for it.
                    498:  */
                    499: iodone(bp)
                    500: register struct buf *bp;
                    501: {
                    502:        register int s;
                    503: 
                    504:        if (bp->b_flags & B_DONE)
                    505:                panic("dup iodone");
                    506:        bp->b_flags |= B_DONE;
                    507:        if (bp->b_flags & B_DIRTY) {
                    508:                if (bp->b_flags & B_ERROR)
                    509:                        panic("IO err in push");
                    510:                s = spl6();
                    511:                bp->av_forw = bclnlist;
                    512:                bp->b_bcount = swapinfo[bp - swapbuf].swsize;
                    513:                bp->b_pfcent = swapinfo[bp - swapbuf].swpf;
                    514:                cnt.v_pgout++;
                    515:                cnt.v_pgpgout += bp->b_bcount / NBPG;
                    516:                bclnlist = bp;
                    517:                if (bswlist.b_flags & B_WANTED)
                    518:                        wakeup((caddr_t)&proc[PAGEPID]);
                    519:                splx(s);
                    520:                return;
                    521:        }
                    522:        if (bp->b_flags&B_ASYNC)
                    523:                brelse(bp);
                    524:        else {
                    525:                bp->b_flags &= ~B_WANTED;
                    526:                wakeup((caddr_t)bp);
                    527:        }
                    528: }
                    529: 
                    530: /*
                    531:  * Zero the core associated with a buffer.
                    532:  */
                    533: clrbuf(bp)
                    534: struct buf *bp;
                    535: {
                    536:        register *p;
                    537:        register c;
                    538: 
                    539:        p = bp->b_un.b_words;
                    540:        c = BUFSIZE/sizeof(int);
                    541:        do
                    542:                *p++ = 0;
                    543:        while (--c);
                    544:        bp->b_resid = 0;
                    545: }
                    546: 
                    547: /*
                    548:  * swap I/O -
                    549:  *
                    550:  * If the flag indicates a dirty page push initiated
                    551:  * by the pageout daemon, we map the page into the i th
                    552:  * virtual page of process 2 (the daemon itself) where i is
                    553:  * the index of the swap header that has been allocated.
                    554:  * We simply initialize the header and queue the I/O but
                    555:  * do not wait for completion. When the I/O completes,
                    556:  * iodone() will link the header to a list of cleaned
                    557:  * pages to be processed by the pageout daemon.
                    558:  */
                    559: swap(p, dblkno, addr, nbytes, rdflg, flag, dev, pfcent)
                    560:        struct proc *p;
                    561:        swblk_t dblkno;
                    562:        caddr_t addr;
                    563:        int flag, nbytes;
                    564:        dev_t dev;
                    565:        unsigned pfcent;
                    566: {
                    567:        register struct buf *bp;
                    568:        register int c;
                    569:        int p2dp;
                    570:        register struct pte *dpte, *vpte;
                    571: 
                    572:        (void) spl6();
                    573:        while (bswlist.av_forw == NULL) {
                    574:                bswlist.b_flags |= B_WANTED;
                    575:                sleep((caddr_t)&bswlist, PSWP+1);
                    576:        }
                    577:        bp = bswlist.av_forw;
                    578:        bswlist.av_forw = bp->av_forw;
                    579:        (void) spl0();
                    580: 
                    581:        bp->b_flags = B_BUSY | B_PHYS | rdflg | flag;
                    582:        if ((bp->b_flags & (B_DIRTY|B_PGIN)) == 0)
                    583:                if (rdflg == B_READ)
                    584:                        sum.v_pswpin += btoc(nbytes);
                    585:                else
                    586:                        sum.v_pswpout += btoc(nbytes);
                    587:        bp->b_proc = p;
                    588:        if (flag & B_DIRTY) {
                    589:                p2dp = ((bp - swapbuf) * CLSIZE) * KLMAX;
                    590:                dpte = dptopte(&proc[PAGEPID], p2dp);
                    591:                vpte = vtopte(p, btop(addr));
                    592:                for (c = 0; c < nbytes; c += NBPG) {
                    593:                        if (vpte->pg_pfnum == 0 || vpte->pg_fod)
                    594:                                panic("swap bad pte");
                    595:                        *dpte++ = *vpte++;
                    596:                }
                    597:                bp->b_un.b_addr = (caddr_t)ctob(p2dp);
                    598:        } else
                    599:                bp->b_un.b_addr = addr;
                    600:        while (nbytes > 0) {
                    601:                c = imin(ctob(120), nbytes);
                    602:                bp->b_bcount = c;
                    603:                bp->b_blkno = dblkno;
                    604:                bp->b_dev = dev;
                    605:                if (flag & B_DIRTY) {
                    606:                        swapinfo[bp - swapbuf].swpf = pfcent;
                    607:                        swapinfo[bp - swapbuf].swsize = nbytes;
                    608:                }
                    609:                if (bdevsw[major(dev)] == NULL)
                    610:                        panic("swap");
                    611:                (*bdevsw[major(dev)]->d_strategy)(bp);
                    612:                if (flag & B_DIRTY) {
                    613:                        if (c < nbytes)
                    614:                                panic("big push");
                    615:                        return;
                    616:                }
                    617:                (void) spl6();
                    618:                while((bp->b_flags&B_DONE)==0)
                    619:                        sleep((caddr_t)bp, PSWP);
                    620:                (void) spl0();
                    621:                bp->b_un.b_addr += c;
                    622:                bp->b_flags &= ~B_DONE;
                    623:                if (bp->b_flags & B_ERROR) {
                    624:                        if ((flag & (B_UAREA|B_PAGET)) || rdflg == B_WRITE)
                    625:                                panic("hard IO err in swap");
                    626:                        swkill(p, (char *)0);
                    627:                }
                    628:                nbytes -= c;
                    629:                dblkno += btoc(c);
                    630:        }
                    631:        (void) spl6();
                    632:        bp->b_flags &= ~(B_BUSY|B_WANTED|B_PHYS|B_PAGET|B_UAREA|B_DIRTY);
                    633:        bp->av_forw = bswlist.av_forw;
                    634:        bswlist.av_forw = bp;
                    635:        if (bswlist.b_flags & B_WANTED) {
                    636:                bswlist.b_flags &= ~B_WANTED;
                    637:                wakeup((caddr_t)&bswlist);
                    638:                wakeup((caddr_t)&proc[PAGEPID]);
                    639:        }
                    640:        (void) spl0();
                    641: }
                    642: 
                    643: /*
                    644:  * Initialize linked list of free swap
                    645:  * headers. These do not actually point
                    646:  * to buffers, but rather to pages that
                    647:  * are being swapped in and out.
                    648:  */
                    649: bswinit()
                    650: {
                    651:        extern int swbufcnt;
                    652:        register int i;
                    653:        register struct buf *sp = swapbuf;
                    654: 
                    655:        bswlist.av_forw = sp;
                    656:        for (i=0; i<swbufcnt-1; i++, sp++)
                    657:                sp->av_forw = sp+1;
                    658:        sp->av_forw = NULL;
                    659: }
                    660: 
                    661: /*
                    662:  * If rout == 0 then killed on swap error, else
                    663:  * rout is the name of the routine where we ran out of
                    664:  * swap space.
                    665:  */
                    666: swkill(p, rout)
                    667:        struct proc *p;
                    668:        char *rout;
                    669: {
                    670: 
                    671:        printf("pid %d: ", p->p_pid);
                    672:        if (rout)
                    673:                printf("killed due to no swap space\n");
                    674:        else
                    675:                printf("killed on swap error\n");
                    676:        /*
                    677:         * To be sure no looping (e.g. in vmsched trying to
                    678:         * swap out) mark process locked in core (as though
                    679:         * done by user) after killing it so noone will try
                    680:         * to swap it out.
                    681:         */
                    682:        psignal(p, SIGKILL);
                    683:        p->p_flag |= SULOCK;
                    684: }
                    685: 
                    686: /*
                    687:  * make sure all write-behind blocks
                    688:  * on dev (or NODEV for all)
                    689:  * are flushed out.
                    690:  * (from umount and update)
                    691:  */
                    692: bflush(dev)
                    693: dev_t dev;
                    694: {
                    695:        register struct buf *bp;
                    696:        register struct buf *flist;
                    697:        register s;
                    698: 
                    699: loop:
                    700:        s = spl6();
                    701:        for (flist = bfreelist; flist < &bfreelist[BQUEUES]; flist++)
                    702:        for (bp = flist->av_forw; bp != flist; bp = bp->av_forw) {
                    703:                if (bp->b_flags&B_DELWRI && (dev == NODEV||dev==bp->b_dev)) {
                    704:                        bp->b_flags |= B_ASYNC;
                    705:                        notavail(bp);
                    706:                        splx(s);
                    707:                        bwrite(bp);
                    708:                        goto loop;
                    709:                }
                    710:        }
                    711:        splx(s);
                    712: }
                    713: 
                    714: /*
                    715:  * Raw I/O. The arguments are
                    716:  *     The strategy routine for the device
                    717:  *     A buffer, which will always be a special buffer
                    718:  *       header owned exclusively by the device for this purpose
                    719:  *     The device number
                    720:  *     Read/write flag
                    721:  * Essentially all the work is computing physical addresses and
                    722:  * validating them.
                    723:  * If the user has the proper access privilidges, the process is
                    724:  * marked 'delayed unlock' and the pages involved in the I/O are
                    725:  * faulted and locked. After the completion of the I/O, the above pages
                    726:  * are unlocked.
                    727:  */
                    728: 
                    729: physio(strat, bp, dev, rw, mincnt)
                    730: int (*strat)(); 
                    731: register struct buf *bp;
                    732: unsigned (*mincnt)();
                    733: {
                    734:        register int c;
                    735:        char *a;
                    736:        register int s;
                    737: 
                    738:        if (useracc(u.u_base,u.u_count,rw==B_READ?B_WRITE:B_READ) == NULL) {
                    739:                u.u_error = EFAULT;
                    740:                return;
                    741:        }
                    742:        s = spl6();
                    743:        while (bp->b_flags&B_BUSY) {
                    744:                bp->b_flags |= B_WANTED;
                    745:                /*sleep((caddr_t)bp, PRIBIO+1);*/
                    746:                switch(tsleep((caddr_t)bp, PRIBIO+1, 20)) {
                    747:                case TS_OK:
                    748:                        continue;
                    749:                case TS_SIG:    /* can't happen at PRIBIO+1*/
                    750:                        continue;
                    751:                case TS_TIME:
                    752:                        u.u_error = EIO;
                    753:                        (void) splx(s);
                    754:                        return;
                    755:                }
                    756:        }
                    757:        (void) splx(s);
                    758:        bp->b_error = 0;
                    759:        bp->b_proc = u.u_procp;
                    760:        bp->b_un.b_addr = u.u_base;
                    761:        while (u.u_count != 0) {
                    762:                bp->b_flags = B_BUSY | B_PHYS | rw;
                    763:                bp->b_dev = dev;
                    764:                bp->b_blkno = Lshift(u.u_offset, PGSHIFT);
                    765:                bp->b_bcount = u.u_count;
                    766:                (*mincnt)(bp);
                    767:                c = bp->b_bcount;
                    768:                u.u_procp->p_flag |= SPHYSIO;
                    769:                vslock(a = bp->b_un.b_addr, c);
                    770:                (*strat)(bp);
                    771:                s = spl6();
                    772:                while ((bp->b_flags&B_DONE) == 0)
                    773:                        sleep((caddr_t)bp, PRIBIO);
                    774:                vsunlock(a, c, rw);
                    775:                u.u_procp->p_flag &= ~SPHYSIO;
                    776:                if (bp->b_flags&B_WANTED)
                    777:                        wakeup((caddr_t)bp);
                    778:                (void) splx(s);
                    779:                bp->b_un.b_addr += c;
                    780:                u.u_count -= c;
                    781:                u.u_offset = Lladd(u.u_offset, c);
                    782:                if (bp->b_flags&B_ERROR)
                    783:                        break;
                    784:        }
                    785:        bp->b_flags &= ~(B_BUSY|B_WANTED|B_PHYS);
                    786:        u.u_count = bp->b_resid;
                    787:        geterror(bp);
                    788: }
                    789: 
                    790: /*ARGSUSED*/
                    791: unsigned
                    792: minphys(bp)
                    793: struct buf *bp;
                    794: {
                    795: 
                    796:        if (bp->b_bcount > (64*1024)-512)
                    797:                bp->b_bcount = (64*1024)-512;
                    798: }
                    799: 
                    800: /*
                    801:  * Pick up the device's error number and pass it to the user;
                    802:  * if there is an error but the number is 0 set a generalized
                    803:  * code.  Actually the latter is always true because devices
                    804:  * don't yet return specific errors.
                    805:  */
                    806: geterror(bp)
                    807: register struct buf *bp;
                    808: {
                    809: 
                    810:        if (bp->b_flags&B_ERROR)
                    811:                if ((u.u_error = bp->b_error)==0)
                    812:                        u.u_error = EIO;
                    813: }
                    814: 
                    815: /*
                    816:  * Invalidate in core blocks belonging to closed or umounted filesystem
                    817:  *
                    818:  * This is not nicely done at all - the buffer ought to be removed from the
                    819:  * hash chains & have its dev/blkno fields clobbered, but unfortunately we
                    820:  * can't do that here, as it is quite possible that the block is still
                    821:  * being used for i/o. Eventually, all disc drivers should be forced to
                    822:  * have a close routine, which ought ensure that the queue is empty, then
                    823:  * properly flush the queues. Until that happy day, this suffices for
                    824:  * correctness.                                                ... kre
                    825:  */
                    826: binval(dev)
                    827: dev_t dev;
                    828: {
                    829:        register struct buf *bp;
                    830:        register struct bufhd *hp;
                    831: #define dp ((struct buf *)hp)
                    832: 
                    833:        for (hp = &bufhash[bufhcnt-1]; hp >= bufhash; hp--)
                    834:                for (bp = dp->b_forw; bp != dp; bp = bp->b_forw)
                    835:                        if (bp->b_dev == dev)
                    836:                                bp->b_flags |= B_INVAL;
                    837: }

unix.superglobalmegacorp.com

This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.