Annotation of researchv8dc/sys/dev/bio.c, revision 1.1.1.1

1.1       root        1: /*     bio.c   4.21    81/05/08        */
                      2: 
                      3: #include "../h/param.h"
                      4: #include "../h/systm.h"
                      5: #include "../h/dir.h"
                      6: #include "../h/user.h"
                      7: #include "../h/buf.h"
                      8: #include "../h/conf.h"
                      9: #include "../h/proc.h"
                     10: #include "../h/seg.h"
                     11: #include "../h/pte.h"
                     12: #include "../h/vm.h"
                     13: #include "../h/trace.h"
                     14: 
                     15: /*
                     16:  * The following several routines allocate and free
                     17:  * buffers with various side effects.  In general the
                     18:  * arguments to an allocate routine are a device and
                     19:  * a block number, and the value is a pointer to
                     20:  * to the buffer header; the buffer is marked "busy"
                     21:  * so that no one else can touch it.  If the block was
                     22:  * already in core, no I/O need be done; if it is
                     23:  * already busy, the process waits until it becomes free.
                     24:  * The following routines allocate a buffer:
                     25:  *     getblk
                     26:  *     bread
                     27:  *     breada
                     28:  *     baddr   (if it is incore)
                     29:  * Eventually the buffer must be released, possibly with the
                     30:  * side effect of writing it out, by using one of
                     31:  *     bwrite
                     32:  *     bdwrite
                     33:  *     bawrite
                     34:  *     brelse
                     35:  */
                     36: 
                     37: struct buf bfreelist[BQUEUES];
                     38: struct buf bswlist, *bclnlist;
                     39: 
                     40: #define        BUFHSZ  63
                     41: struct bufhd bufhash[BUFHSZ];
                     42: #define        BUFHASH(dev, dblkno)    \
                     43:                ((struct buf *)&bufhash[((int)(dev)+(int)(dblkno)) % BUFHSZ])
                     44: 
                     45: /*
                     46:  * Initialize hash links for buffers.
                     47:  */
                     48: bhinit()
                     49: {
                     50:        register int i;
                     51:        register struct bufhd *bp;
                     52: 
                     53:        for (bp = bufhash, i = 0; i < BUFHSZ; i++, bp++)
                     54:                bp->b_forw = bp->b_back = (struct buf *)bp;
                     55: }
                     56: 
                     57: /* #define     DISKMON 1 */
                     58: 
                     59: #ifdef DISKMON
                     60: struct {
                     61:        int     nbuf;
                     62:        long    nread;
                     63:        long    nreada;
                     64:        long    ncache;
                     65:        long    nwrite;
                     66:        long    bufcount[64];
                     67: } io_info;
                     68: #endif
                     69: 
                     70: /*
                     71:  * Swap IO headers -
                     72:  * They contain the necessary information for the swap I/O.
                     73:  * At any given time, a swap header can be in three
                     74:  * different lists. When free it is in the free list, 
                     75:  * when allocated and the I/O queued, it is on the swap 
                     76:  * device list, and finally, if the operation was a dirty 
                     77:  * page push, when the I/O completes, it is inserted 
                     78:  * in a list of cleaned pages to be processed by the pageout daemon.
                     79:  */
                     80: struct buf *swbuf;
                     81: short  *swsize;                /* CAN WE JUST USE B_BCOUNT? */
                     82: int    *swpf;
                     83: 
                     84: 
                     85: #ifndef        UNFAST
                     86: #define        notavail(bp) \
                     87: { \
                     88:        int s = spl6(); \
                     89:        (bp)->av_back->av_forw = (bp)->av_forw; \
                     90:        (bp)->av_forw->av_back = (bp)->av_back; \
                     91:        (bp)->b_flags |= B_BUSY; \
                     92:        splx(s); \
                     93: }
                     94: #endif
                     95: 
                     96: /*
                     97:  * Read in (if necessary) the block and return a buffer pointer.
                     98:  */
                     99: struct buf *
                    100: bread(dev, blkno)
                    101: dev_t dev;
                    102: daddr_t blkno;
                    103: {
                    104:        register struct buf *bp;
                    105: 
                    106:        bp = getblk(dev, blkno);
                    107:        if (bp->b_flags&B_DONE) {
                    108: #ifdef DISKMON
                    109:                io_info.ncache++;
                    110: #endif
                    111:                return(bp);
                    112:        }
                    113:        bp->b_flags |= B_READ;
                    114:        bp->b_bcount = BSIZE(dev);
                    115:        (*bdevsw[major(dev)].d_strategy)(bp);
                    116: #ifdef DISKMON
                    117:        io_info.nread++;
                    118: #endif
                    119:        u.u_vm.vm_inblk++;              /* pay for read */
                    120:        iowait(bp);
                    121:        return(bp);
                    122: }
                    123: 
                    124: /*
                    125:  * Read in the block, like bread, but also start I/O on the
                    126:  * read-ahead block (which is not allocated to the caller)
                    127:  */
                    128: struct buf *
                    129: breada(dev, blkno, rablkno)
                    130: dev_t dev;
                    131: daddr_t blkno, rablkno;
                    132: {
                    133:        register struct buf *bp, *rabp;
                    134: 
                    135:        bp = NULL;
                    136:        if (!incore(dev, blkno)) {
                    137:                bp = getblk(dev, blkno);
                    138:                if ((bp->b_flags&B_DONE) == 0) {
                    139:                        bp->b_flags |= B_READ;
                    140:                        bp->b_bcount = BSIZE(dev);
                    141:                        (*bdevsw[major(dev)].d_strategy)(bp);
                    142: #ifdef DISKMON
                    143:                        io_info.nread++;
                    144: #endif
                    145:                        u.u_vm.vm_inblk++;              /* pay for read */
                    146:                }
                    147:        }
                    148:        if (rablkno && !incore(dev, rablkno)) {
                    149:                rabp = getblk(dev, rablkno);
                    150:                if (rabp->b_flags & B_DONE) {
                    151:                        brelse(rabp);
                    152:                } else {
                    153:                        rabp->b_flags |= B_READ|B_ASYNC;
                    154:                        rabp->b_bcount = BSIZE(dev);
                    155:                        (*bdevsw[major(dev)].d_strategy)(rabp);
                    156: #ifdef DISKMON
                    157:                        io_info.nreada++;
                    158: #endif
                    159:                        u.u_vm.vm_inblk++;              /* pay in advance */
                    160:                }
                    161:        }
                    162:        if(bp == NULL)
                    163:                return(bread(dev, blkno));
                    164:        iowait(bp);
                    165:        return(bp);
                    166: }
                    167: 
                    168: /*
                    169:  * Write the buffer, waiting for completion.
                    170:  * Then release the buffer.
                    171:  */
                    172: bwrite(bp)
                    173: register struct buf *bp;
                    174: {
                    175:        register flag;
                    176: 
                    177:        flag = bp->b_flags;
                    178:        bp->b_flags &= ~(B_READ | B_DONE | B_ERROR | B_DELWRI | B_AGE);
                    179:        bp->b_bcount = BSIZE(bp->b_dev);
                    180: #ifdef DISKMON
                    181:        io_info.nwrite++;
                    182: #endif
                    183:        if ((flag&B_DELWRI) == 0)
                    184:                u.u_vm.vm_oublk++;              /* noone paid yet */
                    185:        (*bdevsw[major(bp->b_dev)].d_strategy)(bp);
                    186:        if ((flag&B_ASYNC) == 0) {
                    187:                iowait(bp);
                    188:                brelse(bp);
                    189:        } else if (flag & B_DELWRI)
                    190:                bp->b_flags |= B_AGE;
                    191:        else
                    192:                geterror(bp);
                    193: }
                    194: 
                    195: /*
                    196:  * Release the buffer, marking it so that if it is grabbed
                    197:  * for another purpose it will be written out before being
                    198:  * given up (e.g. when writing a partial block where it is
                    199:  * assumed that another write for the same block will soon follow).
                    200:  * This can't be done for magtape, since writes must be done
                    201:  * in the same order as requested.
                    202:  */
                    203: bdwrite(bp)
                    204: register struct buf *bp;
                    205: {
                    206:        register int flags;
                    207: 
                    208:        if ((bp->b_flags&B_DELWRI) == 0)
                    209:                u.u_vm.vm_oublk++;              /* noone paid yet */
                    210:        flags = bdevsw[major(bp->b_dev)].d_flags;
                    211:        if(flags & B_TAPE)
                    212:                bawrite(bp);
                    213:        else {
                    214:                bp->b_flags |= B_DELWRI | B_DONE;
                    215:                brelse(bp);
                    216:        }
                    217: }
                    218: 
                    219: /*
                    220:  * Release the buffer, start I/O on it, but don't wait for completion.
                    221:  */
                    222: bawrite(bp)
                    223: register struct buf *bp;
                    224: {
                    225: 
                    226:        bp->b_flags |= B_ASYNC;
                    227:        bwrite(bp);
                    228: }
                    229: 
                    230: /*
                    231:  * release the buffer, with no I/O implied.
                    232:  */
                    233: brelse(bp)
                    234: register struct buf *bp;
                    235: {
                    236:        register struct buf *flist;
                    237:        register s;
                    238: 
                    239:        if (bp->b_flags&B_WANTED)
                    240:                wakeup((caddr_t)bp);
                    241:        if (bfreelist[0].b_flags&B_WANTED) {
                    242:                bfreelist[0].b_flags &= ~B_WANTED;
                    243:                wakeup((caddr_t)bfreelist);
                    244:        }
                    245:        if (bp->b_flags&B_ERROR)
                    246:                if (bp->b_flags & B_LOCKED)
                    247:                        bp->b_flags &= ~B_ERROR;        /* try again later */
                    248:                else
                    249:                        bp->b_dev = NODEV;              /* no assoc */
                    250:        s = spl6();
                    251:        if (bp->b_flags & (B_ERROR|B_INVAL)) {
                    252:                /* block has no info ... put at front of most free list */
                    253:                flist = &bfreelist[BQUEUES-1];
                    254:                flist->av_forw->av_back = bp;
                    255:                bp->av_forw = flist->av_forw;
                    256:                flist->av_forw = bp;
                    257:                bp->av_back = flist;
                    258:        } else {
                    259:                if (bp->b_flags & B_LOCKED)
                    260:                        flist = &bfreelist[BQ_LOCKED];
                    261:                else if (bp->b_flags & B_AGE)
                    262:                        flist = &bfreelist[BQ_AGE];
                    263:                else
                    264:                        flist = &bfreelist[BQ_LRU];
                    265:                flist->av_back->av_forw = bp;
                    266:                bp->av_back = flist->av_back;
                    267:                flist->av_back = bp;
                    268:                bp->av_forw = flist;
                    269:        }
                    270:        bp->b_flags &= ~(B_WANTED|B_BUSY|B_ASYNC|B_AGE);
                    271:        splx(s);
                    272: }
                    273: 
                    274: /*
                    275:  * See if the block is associated with some buffer
                    276:  * (mainly to avoid getting hung up on a wait in breada)
                    277:  */
                    278: incore(dev, blkno)
                    279: dev_t dev;
                    280: daddr_t blkno;
                    281: {
                    282:        register struct buf *bp;
                    283:        register struct buf *dp;
                    284:        register int dblkno = fsbtodb(dev, blkno);
                    285: 
                    286:        dp = BUFHASH(dev, dblkno);
                    287:        for (bp = dp->b_forw; bp != dp; bp = bp->b_forw)
                    288:                if (bp->b_blkno == dblkno && bp->b_dev == dev &&
                    289:                    !(bp->b_flags & B_INVAL))
                    290:                        return (1);
                    291:        return (0);
                    292: }
                    293: 
                    294: struct buf *
                    295: baddr(dev, blkno)
                    296: dev_t dev;
                    297: daddr_t blkno;
                    298: {
                    299: 
                    300:        if (incore(dev, blkno))
                    301:                return (bread(dev, blkno));
                    302:        return (0);
                    303: }
                    304: 
                    305: /*
                    306:  * Assign a buffer for the given block.  If the appropriate
                    307:  * block is already associated, return it; otherwise search
                    308:  * for the oldest non-busy buffer and reassign it.
                    309:  */
                    310: struct buf *
                    311: getblk(dev, blkno)
                    312: dev_t dev;
                    313: daddr_t blkno;
                    314: {
                    315:        register struct buf *bp, *dp, *ep;
                    316:        register int dblkno = fsbtodb(dev, blkno);
                    317: #ifdef DISKMON
                    318:        register int i;
                    319: #endif
                    320: 
                    321:        if ((unsigned)blkno >= 1 << (sizeof(int)*NBBY-PGSHIFT))
                    322:                blkno = 1 << ((sizeof(int)*NBBY-PGSHIFT) + 1);
                    323:        dblkno = fsbtodb(dev, blkno);
                    324:        dp = BUFHASH(dev, dblkno);
                    325:     loop:
                    326:        (void) spl0();
                    327:        for (bp = dp->b_forw; bp != dp; bp = bp->b_forw) {
                    328:                if (bp->b_blkno != dblkno || bp->b_dev != dev ||
                    329:                    bp->b_flags&B_INVAL)
                    330:                        continue;
                    331:                (void) spl6();
                    332:                if (bp->b_flags&B_BUSY) {
                    333:                        bp->b_flags |= B_WANTED;
                    334:                        sleep((caddr_t)bp, PRIBIO+1);
                    335:                        goto loop;
                    336:                }
                    337:                (void) spl0();
                    338: #ifdef DISKMON
                    339:                i = 0;
                    340:                dp = bp->av_forw;
                    341:                while ((dp->b_flags & B_HEAD) == 0) {
                    342:                        i++;
                    343:                        dp = dp->av_forw;
                    344:                }
                    345:                if (i<64)
                    346:                        io_info.bufcount[i]++;
                    347: #endif
                    348:                notavail(bp);
                    349:                bp->b_flags |= B_CACHE;
                    350:                return(bp);
                    351:        }
                    352:        if (major(dev) >= nblkdev)
                    353:                panic("blkdev");
                    354:        (void) spl6();
                    355:        for (ep = &bfreelist[BQUEUES-1]; ep > bfreelist; ep--)
                    356:                if (ep->av_forw != ep)
                    357:                        break;
                    358:        if (ep == bfreelist) {          /* no free blocks at all */
                    359:                ep->b_flags |= B_WANTED;
                    360:                sleep((caddr_t)ep, PRIBIO+1);
                    361:                goto loop;
                    362:        }
                    363:        (void) spl0();
                    364:        bp = ep->av_forw;
                    365:        notavail(bp);
                    366:        if (bp->b_flags & B_DELWRI) {
                    367:                bp->b_flags |= B_ASYNC;
                    368:                bwrite(bp);
                    369:                goto loop;
                    370:        }
                    371:        bp->b_flags = B_BUSY;
                    372:        bp->b_back->b_forw = bp->b_forw;
                    373:        bp->b_forw->b_back = bp->b_back;
                    374:        bp->b_forw = dp->b_forw;
                    375:        bp->b_back = dp;
                    376:        dp->b_forw->b_back = bp;
                    377:        dp->b_forw = bp;
                    378:        bp->b_dev = dev;
                    379:        bp->b_blkno = dblkno;
                    380:        return(bp);
                    381: }
                    382: 
                    383: /*
                    384:  * get an empty block,
                    385:  * not assigned to any particular device
                    386:  */
                    387: struct buf *
                    388: geteblk()
                    389: {
                    390:        register struct buf *bp, *dp;
                    391:        register int s;
                    392: 
                    393: loop:
                    394:        s = spl6();
                    395:        for (dp = &bfreelist[BQUEUES-1]; dp > bfreelist; dp--)
                    396:                if (dp->av_forw != dp)
                    397:                        break;
                    398:        if (dp == bfreelist) {          /* no free blocks */
                    399:                dp->b_flags |= B_WANTED;
                    400:                sleep((caddr_t)dp, PRIBIO+1);
                    401:                goto loop;
                    402:        }
                    403:        (void) splx(s);
                    404:        bp = dp->av_forw;
                    405:        notavail(bp);
                    406:        if (bp->b_flags & B_DELWRI) {
                    407:                bp->b_flags |= B_ASYNC;
                    408:                bwrite(bp);
                    409:                goto loop;
                    410:        }
                    411:        bp->b_flags = B_BUSY|B_INVAL;
                    412:        bp->b_back->b_forw = bp->b_forw;
                    413:        bp->b_forw->b_back = bp->b_back;
                    414:        bp->b_forw = dp->b_forw;
                    415:        bp->b_back = dp;
                    416:        dp->b_forw->b_back = bp;
                    417:        dp->b_forw = bp;
                    418:        bp->b_dev = (dev_t)NODEV;
                    419:        return(bp);
                    420: }
                    421: 
                    422: /*
                    423:  * Wait for I/O completion on the buffer; return errors
                    424:  * to the user.
                    425:  */
                    426: iowait(bp)
                    427: register struct buf *bp;
                    428: {
                    429: 
                    430:        (void) spl6();
                    431:        while ((bp->b_flags&B_DONE)==0)
                    432:                sleep((caddr_t)bp, PRIBIO);
                    433:        (void) spl0();
                    434:        geterror(bp);
                    435: }
                    436: 
                    437: #ifdef UNFAST
                    438: /*
                    439:  * Unlink a buffer from the available list and mark it busy.
                    440:  * (internal interface)
                    441:  */
                    442: notavail(bp)
                    443: register struct buf *bp;
                    444: {
                    445:        register s;
                    446: 
                    447:        s = spl6();
                    448:        bp->av_back->av_forw = bp->av_forw;
                    449:        bp->av_forw->av_back = bp->av_back;
                    450:        bp->b_flags |= B_BUSY;
                    451:        splx(s);
                    452: }
                    453: #endif
                    454: 
                    455: /*
                    456:  * Mark I/O complete on a buffer. If the header
                    457:  * indicates a dirty page push completion, the
                    458:  * header is inserted into the ``cleaned'' list
                    459:  * to be processed by the pageout daemon. Otherwise
                    460:  * release it if I/O is asynchronous, and wake 
                    461:  * up anyone waiting for it.
                    462:  */
                    463: iodone(bp)
                    464: register struct buf *bp;
                    465: {
                    466:        register int s;
                    467: 
                    468:        if (bp->b_flags & B_DONE)
                    469:                panic("dup iodone");
                    470:        bp->b_flags |= B_DONE;
                    471:        if (bp->b_flags & B_DIRTY) {
                    472:                if (bp->b_flags & B_ERROR)
                    473:                        panic("IO err in push");
                    474:                s = spl6();
                    475:                bp->av_forw = bclnlist;
                    476:                bp->b_bcount = swsize[bp - swbuf];
                    477:                bp->b_pfcent = swpf[bp - swbuf];
                    478:                cnt.v_pgout++;
                    479:                cnt.v_pgpgout += bp->b_bcount / NBPG;
                    480:                bclnlist = bp;
                    481:                if (bswlist.b_flags & B_WANTED)
                    482:                        wakeup((caddr_t)&proc[2]);
                    483:                splx(s);
                    484:                return;
                    485:        }
                    486:        if (bp->b_flags&B_ASYNC)
                    487:                brelse(bp);
                    488:        else {
                    489:                bp->b_flags &= ~B_WANTED;
                    490:                wakeup((caddr_t)bp);
                    491:        }
                    492: }
                    493: 
                    494: /*
                    495:  * Zero the core associated with a buffer.
                    496:  */
                    497: clrbuf(bp)
                    498: struct buf *bp;
                    499: {
                    500:        register *p;
                    501:        register c;
                    502: 
                    503:        p = bp->b_un.b_words;
                    504:        c = BUFSIZE/sizeof(int);
                    505:        do
                    506:                *p++ = 0;
                    507:        while (--c);
                    508:        bp->b_resid = 0;
                    509: }
                    510: 
                    511: /*
                    512:  * swap I/O -
                    513:  *
                    514:  * If the flag indicates a dirty page push initiated
                    515:  * by the pageout daemon, we map the page into the i th
                    516:  * virtual page of process 2 (the daemon itself) where i is
                    517:  * the index of the swap header that has been allocated.
                    518:  * We simply initialize the header and queue the I/O but
                    519:  * do not wait for completion. When the I/O completes,
                    520:  * iodone() will link the header to a list of cleaned
                    521:  * pages to be processed by the pageout daemon.
                    522:  */
                    523: swap(p, dblkno, addr, nbytes, rdflg, flag, dev, pfcent)
                    524:        struct proc *p;
                    525:        swblk_t dblkno;
                    526:        caddr_t addr;
                    527:        int flag, nbytes;
                    528:        dev_t dev;
                    529:        unsigned pfcent;
                    530: {
                    531:        register struct buf *bp;
                    532:        register int c;
                    533:        int p2dp;
                    534:        register struct pte *dpte, *vpte;
                    535: 
                    536:        (void) spl6();
                    537:        while (bswlist.av_forw == NULL) {
                    538:                bswlist.b_flags |= B_WANTED;
                    539:                sleep((caddr_t)&bswlist, PSWP+1);
                    540:        }
                    541:        bp = bswlist.av_forw;
                    542:        bswlist.av_forw = bp->av_forw;
                    543:        (void) spl0();
                    544: 
                    545:        bp->b_flags = B_BUSY | B_PHYS | rdflg | flag;
                    546:        if ((bp->b_flags & (B_DIRTY|B_PGIN)) == 0)
                    547:                if (rdflg == B_READ)
                    548:                        sum.v_pswpin += btoc(nbytes);
                    549:                else
                    550:                        sum.v_pswpout += btoc(nbytes);
                    551:        bp->b_proc = p;
                    552:        if (flag & B_DIRTY) {
                    553:                p2dp = ((bp - swbuf) * CLSIZE) * KLMAX;
                    554:                dpte = dptopte(&proc[2], p2dp);
                    555:                vpte = vtopte(p, btop(addr));
                    556:                for (c = 0; c < nbytes; c += NBPG) {
                    557:                        if (vpte->pg_pfnum == 0 || vpte->pg_fod)
                    558:                                panic("swap bad pte");
                    559:                        *dpte++ = *vpte++;
                    560:                }
                    561:                bp->b_un.b_addr = (caddr_t)ctob(p2dp);
                    562:        } else
                    563:                bp->b_un.b_addr = addr;
                    564:        while (nbytes > 0) {
                    565:                c = imin(ctob(120), nbytes);
                    566:                bp->b_bcount = c;
                    567:                bp->b_blkno = dblkno;
                    568:                bp->b_dev = dev;
                    569:                if (flag & B_DIRTY) {
                    570:                        swpf[bp - swbuf] = pfcent;
                    571:                        swsize[bp - swbuf] = nbytes;
                    572:                }
                    573:                (*bdevsw[major(dev)].d_strategy)(bp);
                    574:                if (flag & B_DIRTY) {
                    575:                        if (c < nbytes)
                    576:                                panic("big push");
                    577:                        return;
                    578:                }
                    579:                (void) spl6();
                    580:                while((bp->b_flags&B_DONE)==0)
                    581:                        sleep((caddr_t)bp, PSWP);
                    582:                (void) spl0();
                    583:                bp->b_un.b_addr += c;
                    584:                bp->b_flags &= ~B_DONE;
                    585:                if (bp->b_flags & B_ERROR) {
                    586:                        if ((flag & (B_UAREA|B_PAGET)) || rdflg == B_WRITE)
                    587:                                panic("hard IO err in swap");
                    588:                        swkill(p, (char *)0);
                    589:                }
                    590:                nbytes -= c;
                    591:                dblkno += btoc(c);
                    592:        }
                    593:        (void) spl6();
                    594:        bp->b_flags &= ~(B_BUSY|B_WANTED|B_PHYS|B_PAGET|B_UAREA|B_DIRTY);
                    595:        bp->av_forw = bswlist.av_forw;
                    596:        bswlist.av_forw = bp;
                    597:        if (bswlist.b_flags & B_WANTED) {
                    598:                bswlist.b_flags &= ~B_WANTED;
                    599:                wakeup((caddr_t)&bswlist);
                    600:                wakeup((caddr_t)&proc[2]);
                    601:        }
                    602:        (void) spl0();
                    603: }
                    604: 
                    605: /*
                    606:  * If rout == 0 then killed on swap error, else
                    607:  * rout is the name of the routine where we ran out of
                    608:  * swap space.
                    609:  */
                    610: swkill(p, rout)
                    611:        struct proc *p;
                    612:        char *rout;
                    613: {
                    614:        char *mesg;
                    615: 
                    616:        printf("pid %d: ", p->p_pid);
                    617:        if (rout)
                    618:                printf(mesg = "killed due to no swap space\n");
                    619:        else
                    620:                printf(mesg = "killed on swap error\n");
                    621:        uprintf("sorry, pid %d was %s", p->p_pid, mesg);
                    622:        /*
                    623:         * To be sure no looping (e.g. in vmsched trying to
                    624:         * swap out) mark process locked in core (as though
                    625:         * done by user) after killing it so noone will try
                    626:         * to swap it out.
                    627:         */
                    628:        psignal(p, SIGKILL);
                    629:        p->p_flag |= SULOCK;
                    630: }
                    631: 
                    632: /*
                    633:  * make sure all write-behind blocks
                    634:  * on dev (or NODEV for all)
                    635:  * are flushed out.
                    636:  * (from umount and update)
                    637:  */
                    638: bflush(dev)
                    639: dev_t dev;
                    640: {
                    641:        register struct buf *bp;
                    642:        register struct buf *flist;
                    643:        register int s;
                    644: 
                    645: loop:
                    646:        s = spl6();
                    647:        trace(TR_BFIN, dev, 1);
                    648:        for (flist = bfreelist; flist < &bfreelist[BQUEUES]; flist++)
                    649:        for (bp = flist->av_forw; bp != flist; bp = bp->av_forw) {
                    650:                if (bp->b_flags&B_DELWRI && (dev == NODEV||dev==bp->b_dev)) {
                    651:                        bp->b_flags |= B_ASYNC;
                    652:                        notavail(bp);
                    653:                        bwrite(bp);
                    654:                        goto loop;
                    655:                }
                    656:        }
                    657:        trace(TR_BFOUT, s, 1);
                    658:        (void) spl0();
                    659: }
                    660: 
                    661: /*
                    662:  * Raw I/O. The arguments are
                    663:  *     The strategy routine for the device
                    664:  *     A buffer, which will always be a special buffer
                    665:  *       header owned exclusively by the device for this purpose
                    666:  *     The device number
                    667:  *     Read/write flag
                    668:  * Essentially all the work is computing physical addresses and
                    669:  * validating them.
                    670:  * If the user has the proper access privilidges, the process is
                    671:  * marked 'delayed unlock' and the pages involved in the I/O are
                    672:  * faulted and locked. After the completion of the I/O, the above pages
                    673:  * are unlocked.
                    674:  */
                    675: physio(strat, bp, dev, rw, mincnt)
                    676: int (*strat)(); 
                    677: register struct buf *bp;
                    678: unsigned (*mincnt)();
                    679: {
                    680:        register int c;
                    681:        char *a;
                    682:        register int s;
                    683: 
                    684:        if (useracc(u.u_base,u.u_count,rw==B_READ?B_WRITE:B_READ) == NULL) {
                    685:                u.u_error = EFAULT;
                    686:                return;
                    687:        }
                    688:        s = spl6();
                    689:        while (bp->b_flags&B_BUSY) {
                    690:                bp->b_flags |= B_WANTED;
                    691:                /*sleep((caddr_t)bp, PRIBIO+1);*/
                    692:                switch(tsleep((caddr_t)bp, PRIBIO+1, 20)) {
                    693:                case TS_OK:
                    694:                        continue;
                    695:                case TS_SIG:    /* can't happen at PRIBIO+1*/
                    696:                        continue;
                    697:                case TS_TIME:
                    698:                        u.u_error = EIO;
                    699:                        (void) splx(s);
                    700:                        return;
                    701:                }
                    702:        }
                    703:        (void) splx(s);
                    704:        bp->b_error = 0;
                    705:        bp->b_proc = u.u_procp;
                    706:        bp->b_un.b_addr = u.u_base;
                    707:        while (u.u_count != 0) {
                    708:                bp->b_flags = B_BUSY | B_PHYS | rw;
                    709:                bp->b_dev = dev;
                    710:                bp->b_blkno = u.u_offset >> PGSHIFT;
                    711:                bp->b_bcount = u.u_count;
                    712:                (*mincnt)(bp);
                    713:                c = bp->b_bcount;
                    714:                u.u_procp->p_flag |= SPHYSIO;
                    715:                vslock(a = bp->b_un.b_addr, c);
                    716:                (*strat)(bp);
                    717:                s = spl6();
                    718:                while ((bp->b_flags&B_DONE) == 0)
                    719:                        sleep((caddr_t)bp, PRIBIO);
                    720:                vsunlock(a, c, rw);
                    721:                u.u_procp->p_flag &= ~SPHYSIO;
                    722:                if (bp->b_flags&B_WANTED)
                    723:                        wakeup((caddr_t)bp);
                    724:                (void) splx(s);
                    725:                bp->b_un.b_addr += c;
                    726:                u.u_count -= c;
                    727:                u.u_offset += c;
                    728:                if (bp->b_flags&B_ERROR)
                    729:                        break;
                    730:        }
                    731:        bp->b_flags &= ~(B_BUSY|B_WANTED|B_PHYS);
                    732:        u.u_count = bp->b_resid;
                    733:        geterror(bp);
                    734: }
                    735: 
                    736: /*ARGSUSED*/
                    737: unsigned
                    738: minphys(bp)
                    739: struct buf *bp;
                    740: {
                    741: 
                    742:        if (bp->b_bcount > 60 * 1024)
                    743:                bp->b_bcount = 60 * 1024;
                    744: }
                    745: 
                    746: /*
                    747:  * Pick up the device's error number and pass it to the user;
                    748:  * if there is an error but the number is 0 set a generalized
                    749:  * code.  Actually the latter is always true because devices
                    750:  * don't yet return specific errors.
                    751:  */
                    752: geterror(bp)
                    753: register struct buf *bp;
                    754: {
                    755: 
                    756:        if (bp->b_flags&B_ERROR)
                    757:                if ((u.u_error = bp->b_error)==0)
                    758:                        u.u_error = EIO;
                    759: }
                    760: 
                    761: /*
                    762:  * Invalidate in core blocks belonging to closed or umounted filesystem
                    763:  *
                    764:  * This is not nicely done at all - the buffer ought to be removed from the
                    765:  * hash chains & have its dev/blkno fields clobbered, but unfortunately we
                    766:  * can't do that here, as it is quite possible that the block is still
                    767:  * being used for i/o. Eventually, all disc drivers should be forced to
                    768:  * have a close routine, which ought ensure that the queue is empty, then
                    769:  * properly flush the queues. Until that happy day, this suffices for
                    770:  * correctness.                                                ... kre
                    771:  */
                    772: binval(dev)
                    773: dev_t dev;
                    774: {
                    775:        register struct buf *bp;
                    776:        register struct bufhd *hp;
                    777: #define dp ((struct buf *)hp)
                    778: 
                    779:        for (hp = bufhash; hp < &bufhash[BUFHSZ]; hp++)
                    780:                for (bp = dp->b_forw; bp != dp; bp = bp->b_forw)
                    781:                        if (bp->b_dev == dev)
                    782:                                bp->b_flags |= B_INVAL;
                    783: }

unix.superglobalmegacorp.com

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