4.2 distribution

[unix-history] / usr / src / sys / kern / vfs_cluster.c

diff --git a/usr/src/sys/kern/vfs_cluster.c b/usr/src/sys/kern/vfs_cluster.c
index 2682baf..ea5781a 100644 (file)
--- a/usr/src/sys/kern/vfs_cluster.c
+++ b/usr/src/sys/kern/vfs_cluster.c
@@ -1,4 +1,6 @@
-/*     vfs_cluster.c   4.35    82/08/13        */
+/*     vfs_cluster.c   6.1     83/07/29        */
+
+#include "../machine/pte.h"

 
 #include "../h/param.h"
 #include "../h/systm.h"
 
 #include "../h/param.h"
 #include "../h/systm.h"


@@ -8,12 +10,9 @@
 #include "../h/conf.h"
 #include "../h/proc.h"
 #include "../h/seg.h"
 #include "../h/conf.h"
 #include "../h/proc.h"
 #include "../h/seg.h"

-#include "../h/pte.h"

 #include "../h/vm.h"
 #include "../h/trace.h"
 
 #include "../h/vm.h"
 #include "../h/trace.h"
 

-int bioprintfs = 0;
-

 /*
  * Read in (if necessary) the block and return a buffer pointer.
  */
 /*
  * Read in (if necessary) the block and return a buffer pointer.
  */


@@ -25,15 +24,19 @@ bread(dev, blkno, size)
 {
        register struct buf *bp;
 
 {
        register struct buf *bp;
 

+       if (size == 0)
+               panic("bread: size 0");

        bp = getblk(dev, blkno, size);
        if (bp->b_flags&B_DONE) {
                trace(TR_BREADHIT, dev, blkno);
                return(bp);
        }
        bp->b_flags |= B_READ;
        bp = getblk(dev, blkno, size);
        if (bp->b_flags&B_DONE) {
                trace(TR_BREADHIT, dev, blkno);
                return(bp);
        }
        bp->b_flags |= B_READ;

+       if (bp->b_bcount > bp->b_bufsize)
+               panic("bread");

        (*bdevsw[major(dev)].d_strategy)(bp);
        trace(TR_BREADMISS, dev, blkno);
        (*bdevsw[major(dev)].d_strategy)(bp);
        trace(TR_BREADMISS, dev, blkno);

-       u.u_vm.vm_inblk++;              /* pay for read */
+       u.u_ru.ru_inblock++;            /* pay for read */

        biowait(bp);
        return(bp);
 }
        biowait(bp);
        return(bp);
 }


@@ -43,10 +46,10 @@ bread(dev, blkno, size)
  * read-ahead block (which is not allocated to the caller)
  */
 struct buf *
  * read-ahead block (which is not allocated to the caller)
  */
 struct buf *

-breada(dev, blkno, size, rablkno, rasize)
+breada(dev, blkno, size, rablkno, rabsize)

        dev_t dev;
        daddr_t blkno; int size;
        dev_t dev;
        daddr_t blkno; int size;

-       daddr_t rablkno; int rasize;
+       daddr_t rablkno; int rabsize;

 {
        register struct buf *bp, *rabp;
 
 {
        register struct buf *bp, *rabp;
 


@@ -60,9 +63,11 @@ breada(dev, blkno, size, rablkno, rasize)
                bp = getblk(dev, blkno, size);
                if ((bp->b_flags&B_DONE) == 0) {
                        bp->b_flags |= B_READ;
                bp = getblk(dev, blkno, size);
                if ((bp->b_flags&B_DONE) == 0) {
                        bp->b_flags |= B_READ;

+                       if (bp->b_bcount > bp->b_bufsize)
+                               panic("breada");

                        (*bdevsw[major(dev)].d_strategy)(bp);
                        trace(TR_BREADMISS, dev, blkno);
                        (*bdevsw[major(dev)].d_strategy)(bp);
                        trace(TR_BREADMISS, dev, blkno);

-                       u.u_vm.vm_inblk++;              /* pay for read */
+                       u.u_ru.ru_inblock++;            /* pay for read */

                } else
                        trace(TR_BREADHIT, dev, blkno);
        }
                } else
                        trace(TR_BREADHIT, dev, blkno);
        }


@@ -72,15 +77,17 @@ breada(dev, blkno, size, rablkno, rasize)
         * on it also (as above).
         */
        if (rablkno && !incore(dev, rablkno)) {
         * on it also (as above).
         */
        if (rablkno && !incore(dev, rablkno)) {

-               rabp = getblk(dev, rablkno, rasize);
+               rabp = getblk(dev, rablkno, rabsize);

                if (rabp->b_flags & B_DONE) {
                        brelse(rabp);
                        trace(TR_BREADHITRA, dev, blkno);
                } else {
                        rabp->b_flags |= B_READ|B_ASYNC;
                if (rabp->b_flags & B_DONE) {
                        brelse(rabp);
                        trace(TR_BREADHITRA, dev, blkno);
                } else {
                        rabp->b_flags |= B_READ|B_ASYNC;

+                       if (rabp->b_bcount > rabp->b_bufsize)
+                               panic("breadrabp");

                        (*bdevsw[major(dev)].d_strategy)(rabp);
                        trace(TR_BREADMISSRA, dev, rablock);
                        (*bdevsw[major(dev)].d_strategy)(rabp);
                        trace(TR_BREADMISSRA, dev, rablock);

-                       u.u_vm.vm_inblk++;              /* pay in advance */
+                       u.u_ru.ru_inblock++;            /* pay in advance */

                }
        }
 
                }
        }
 


@@ -105,28 +112,24 @@ bwrite(bp)
        register flag;
 
        flag = bp->b_flags;
        register flag;
 
        flag = bp->b_flags;

-       bp->b_flags &= ~(B_READ | B_DONE | B_ERROR | B_DELWRI | B_AGE);
+       bp->b_flags &= ~(B_READ | B_DONE | B_ERROR | B_DELWRI);

        if ((flag&B_DELWRI) == 0)
        if ((flag&B_DELWRI) == 0)

-               u.u_vm.vm_oublk++;              /* noone paid yet */
+               u.u_ru.ru_oublock++;            /* noone paid yet */

        trace(TR_BWRITE, bp->b_dev, bp->b_blkno);
        trace(TR_BWRITE, bp->b_dev, bp->b_blkno);

-if (bioprintfs)
-printf("write %x blk %d count %d\n", bp->b_dev, bp->b_blkno, bp->b_bcount);
+       if (bp->b_bcount > bp->b_bufsize)
+               panic("bwrite");

        (*bdevsw[major(bp->b_dev)].d_strategy)(bp);
 
        /*
         * If the write was synchronous, then await i/o completion.
         * If the write was "delayed", then we put the buffer on
         * the q of blocks awaiting i/o completion status.
        (*bdevsw[major(bp->b_dev)].d_strategy)(bp);
 
        /*
         * If the write was synchronous, then await i/o completion.
         * If the write was "delayed", then we put the buffer on
         * the q of blocks awaiting i/o completion status.

-        * Otherwise, the i/o must be finished and we check for
-        * an error.

         */
        if ((flag&B_ASYNC) == 0) {
                biowait(bp);
                brelse(bp);
        } else if (flag & B_DELWRI)
                bp->b_flags |= B_AGE;
         */
        if ((flag&B_ASYNC) == 0) {
                biowait(bp);
                brelse(bp);
        } else if (flag & B_DELWRI)
                bp->b_flags |= B_AGE;

-       else
-               u.u_error = geterror(bp);

 }
 
 /*
 }
 
 /*


@@ -143,7 +146,7 @@ bdwrite(bp)
        register int flags;
 
        if ((bp->b_flags&B_DELWRI) == 0)
        register int flags;
 
        if ((bp->b_flags&B_DELWRI) == 0)

-               u.u_vm.vm_oublk++;              /* noone paid yet */
+               u.u_ru.ru_oublock++;            /* noone paid yet */

        flags = bdevsw[major(bp->b_dev)].d_flags;
        if(flags & B_TAPE)
                bawrite(bp);
        flags = bdevsw[major(bp->b_dev)].d_flags;
        if(flags & B_TAPE)
                bawrite(bp);


@@ -193,9 +196,13 @@ brelse(bp)
         * Stick the buffer back on a free list.
         */
        s = spl6();
         * Stick the buffer back on a free list.
         */
        s = spl6();

-       if (bp->b_flags & (B_ERROR|B_INVAL)) {
+       if (bp->b_bufsize <= 0) {
+               /* block has no buffer ... put at front of unused buffer list */
+               flist = &bfreelist[BQ_EMPTY];
+               binsheadfree(bp, flist);
+       } else if (bp->b_flags & (B_ERROR|B_INVAL)) {

                /* block has no info ... put at front of most free list */
                /* block has no info ... put at front of most free list */

-               flist = &bfreelist[BQUEUES-1];
+               flist = &bfreelist[BQ_AGE];

                binsheadfree(bp, flist);
        } else {
                if (bp->b_flags & B_LOCKED)
                binsheadfree(bp, flist);
        } else {
                if (bp->b_flags & B_LOCKED)


@@ -256,10 +263,10 @@ getblk(dev, blkno, size)
        daddr_t blkno;
        int size;
 {
        daddr_t blkno;
        int size;
 {

-       register struct buf *bp, *dp, *ep;
+       register struct buf *bp, *dp;

        int s;
 
        int s;
 

-       if ((unsigned)blkno >= 1 << (sizeof(int)*NBBY-PGSHIFT))
+       if ((unsigned)blkno >= 1 << (sizeof(int)*NBBY-PGSHIFT)) /* XXX */

                blkno = 1 << ((sizeof(int)*NBBY-PGSHIFT) + 1);
        /*
         * Search the cache for the block.  If we hit, but
                blkno = 1 << ((sizeof(int)*NBBY-PGSHIFT) + 1);
        /*
         * Search the cache for the block.  If we hit, but


@@ -288,35 +295,13 @@ loop:
        }
        if (major(dev) >= nblkdev)
                panic("blkdev");
        }
        if (major(dev) >= nblkdev)
                panic("blkdev");

-       /*
-        * Not found in the cache, select something from
-        * a free list.  Preference is to LRU list, then AGE list.
-        */
-       s = spl6();
-       for (ep = &bfreelist[BQUEUES-1]; ep > bfreelist; ep--)
-               if (ep->av_forw != ep)
-                       break;
-       if (ep == bfreelist) {          /* no free blocks at all */
-               ep->b_flags |= B_WANTED;
-               sleep((caddr_t)ep, PRIBIO+1);
-               splx(s);
-               goto loop;
-       }
-       splx(s);
-       bp = ep->av_forw;
-       notavail(bp);
-       if (bp->b_flags & B_DELWRI) {
-               bp->b_flags |= B_ASYNC;
-               bwrite(bp);
-               goto loop;
-       }
-       trace(TR_BRELSE, bp->b_dev, bp->b_blkno);
-       bp->b_flags = B_BUSY;
+       bp = getnewbuf();

        bfree(bp);
        bremhash(bp);
        binshash(bp, dp);
        bp->b_dev = dev;
        bp->b_blkno = blkno;
        bfree(bp);
        bremhash(bp);
        binshash(bp, dp);
        bp->b_dev = dev;
        bp->b_blkno = blkno;

+       bp->b_error = 0;

        if (brealloc(bp, size) == 0)
                goto loop;
        return(bp);
        if (brealloc(bp, size) == 0)
                goto loop;
        return(bp);


@@ -330,33 +315,17 @@ struct buf *
 geteblk(size)
        int size;
 {
 geteblk(size)
        int size;
 {

-       register struct buf *bp, *dp;
-       int s;
+       register struct buf *bp, *flist;

 
 loop:
 
 loop:

-       s = spl6();
-       for (dp = &bfreelist[BQUEUES-1]; dp > bfreelist; dp--)
-               if (dp->av_forw != dp)
-                       break;
-       if (dp == bfreelist) {          /* no free blocks */
-               dp->b_flags |= B_WANTED;
-               sleep((caddr_t)dp, PRIBIO+1);
-               goto loop;
-       }
-       splx(s);
-       bp = dp->av_forw;
-       notavail(bp);
-       if (bp->b_flags & B_DELWRI) {
-               bp->b_flags |= B_ASYNC;
-               bwrite(bp);
-               goto loop;
-       }
-       trace(TR_BRELSE, bp->b_dev, bp->b_blkno);
-       bp->b_flags = B_BUSY|B_INVAL;
+       bp = getnewbuf();
+       bp->b_flags |= B_INVAL;

        bfree(bp);
        bremhash(bp);
        bfree(bp);
        bremhash(bp);

-       binshash(bp, dp);
+       flist = &bfreelist[BQ_AGE];
+       binshash(bp, flist);

        bp->b_dev = (dev_t)NODEV;
        bp->b_dev = (dev_t)NODEV;

+       bp->b_error = 0;

        if (brealloc(bp, size) == 0)
                goto loop;
        return(bp);
        if (brealloc(bp, size) == 0)
                goto loop;
        return(bp);


@@ -364,6 +333,7 @@ loop:
 
 /*
  * Allocate space associated with a buffer.
 
 /*
  * Allocate space associated with a buffer.

+ * If can't get space, buffer is released

  */
 brealloc(bp, size)
        register struct buf *bp;
  */
 brealloc(bp, size)
        register struct buf *bp;


@@ -387,11 +357,11 @@ brealloc(bp, size)
                }
                if (bp->b_flags & B_LOCKED)
                        panic("brealloc");
                }
                if (bp->b_flags & B_LOCKED)
                        panic("brealloc");

-               goto allocit;
+               return (allocbuf(bp, size));

        }
        bp->b_flags &= ~B_DONE;
        if (bp->b_dev == NODEV)
        }
        bp->b_flags &= ~B_DONE;
        if (bp->b_dev == NODEV)

-               goto allocit;
+               return (allocbuf(bp, size));

 
        /*
         * Search cache for any buffers that overlap the one that we
 
        /*
         * Search cache for any buffers that overlap the one that we


@@ -402,7 +372,7 @@ brealloc(bp, size)
         * when two buffer are trying to get the same set of disk blocks.
         */
        start = bp->b_blkno;
         * when two buffer are trying to get the same set of disk blocks.
         */
        start = bp->b_blkno;

-       last = start + (size / DEV_BSIZE) - 1;
+       last = start + btodb(size) - 1;

        dp = BUFHASH(bp->b_dev, bp->b_blkno);
 loop:
        for (ep = dp->b_forw; ep != dp; ep = ep->b_forw) {
        dp = BUFHASH(bp->b_dev, bp->b_blkno);
 loop:
        for (ep = dp->b_forw; ep != dp; ep = ep->b_forw) {


@@ -410,55 +380,60 @@ loop:
                        continue;
                /* look for overlap */
                if (ep->b_bcount == 0 || ep->b_blkno > last ||
                        continue;
                /* look for overlap */
                if (ep->b_bcount == 0 || ep->b_blkno > last ||

-                   ep->b_blkno + (ep->b_bcount / DEV_BSIZE) <= start)
+                   ep->b_blkno + btodb(ep->b_bcount) <= start)

                        continue;
                        continue;

-if (bioprintfs)
-if (ep->b_flags&B_BUSY)
-printf("sleeping on:dev 0x%x, blks %d-%d, flg 0%o allocing dev 0x%x, blks %d-%d, flg 0%o\n",
-ep->b_dev, ep->b_blkno, ep->b_blkno + (ep->b_bcount / DEV_BSIZE) - 1,
-ep->b_flags, bp->b_dev, start, last, bp->b_flags);

                s = spl6();
                if (ep->b_flags&B_BUSY) {
                        ep->b_flags |= B_WANTED;
                        sleep((caddr_t)ep, PRIBIO+1);
                s = spl6();
                if (ep->b_flags&B_BUSY) {
                        ep->b_flags |= B_WANTED;
                        sleep((caddr_t)ep, PRIBIO+1);

-                       (void) splx(s);
+                       splx(s);

                        goto loop;
                }
                        goto loop;
                }

-               (void) splx(s);
+               splx(s);

                notavail(ep);
                if (ep->b_flags & B_DELWRI) {
                notavail(ep);
                if (ep->b_flags & B_DELWRI) {

-if (bioprintfs)
-printf("DELWRI:dev 0x%x, blks %d-%d, flg 0%o allocing dev 0x%x, blks %d-%d, flg 0%o\n",
-ep->b_dev, ep->b_blkno, ep->b_blkno + (ep->b_bcount / DEV_BSIZE) - 1,
-ep->b_flags, bp->b_dev, start, last, bp->b_flags);

                        bwrite(ep);
                        goto loop;
                }
                ep->b_flags |= B_INVAL;
                brelse(ep);
        }
                        bwrite(ep);
                        goto loop;
                }
                ep->b_flags |= B_INVAL;
                brelse(ep);
        }

-allocit:
-       /*
-        * Here the buffer is already available, so all we
-        * need to do is set the size. Someday a better memory
-        * management scheme will be implemented.
-        */
-       bp->b_bcount = size;
-       return (1);
+       return (allocbuf(bp, size));

 }
 
 /*
 }
 
 /*

- * Release space associated with a buffer.
+ * Find a buffer which is available for use.
+ * Select something from a free list.
+ * Preference is to AGE list, then LRU list.

  */
  */

-bfree(bp)
-       struct buf *bp;
+struct buf *
+getnewbuf()

 {
 {

-       /*
-        * Here the buffer does not change, so all we
-        * need to do is set the size. Someday a better memory
-        * management scheme will be implemented.
-        */
-       bp->b_bcount = 0;
+       register struct buf *bp, *dp;
+       int s;
+
+loop:
+       s = spl6();
+       for (dp = &bfreelist[BQ_AGE]; dp > bfreelist; dp--)
+               if (dp->av_forw != dp)
+                       break;
+       if (dp == bfreelist) {          /* no free blocks */
+               dp->b_flags |= B_WANTED;
+               sleep((caddr_t)dp, PRIBIO+1);
+               splx(s);
+               goto loop;
+       }
+       splx(s);
+       bp = dp->av_forw;
+       notavail(bp);
+       if (bp->b_flags & B_DELWRI) {
+               bp->b_flags |= B_ASYNC;
+               bwrite(bp);
+               goto loop;
+       }
+       trace(TR_BRELSE, bp->b_dev, bp->b_blkno);
+       bp->b_flags = B_BUSY;
+       return (bp);

 }
 
 /*
 }
 
 /*


@@ -474,38 +449,26 @@ biowait(bp)
        while ((bp->b_flags&B_DONE)==0)
                sleep((caddr_t)bp, PRIBIO);
        splx(s);
        while ((bp->b_flags&B_DONE)==0)
                sleep((caddr_t)bp, PRIBIO);
        splx(s);

-       u.u_error = geterror(bp);
+       if (u.u_error == 0)                     /* XXX */
+               u.u_error = geterror(bp);

 }
 
 /*
 }
 
 /*

- * Mark I/O complete on a buffer. If the header
- * indicates a dirty page push completion, the
- * header is inserted into the ``cleaned'' list
- * to be processed by the pageout daemon. Otherwise
- * release it if I/O is asynchronous, and wake 
- * up anyone waiting for it.
+ * Mark I/O complete on a buffer.
+ * If someone should be called, e.g. the pageout
+ * daemon, do so.  Otherwise, wake up anyone
+ * waiting for it.

  */
 biodone(bp)
        register struct buf *bp;
 {
  */
 biodone(bp)
        register struct buf *bp;
 {

-       register int s;

 
        if (bp->b_flags & B_DONE)
                panic("dup biodone");
        bp->b_flags |= B_DONE;
 
        if (bp->b_flags & B_DONE)
                panic("dup biodone");
        bp->b_flags |= B_DONE;

-       if (bp->b_flags & B_DIRTY) {
-               if (bp->b_flags & B_ERROR)
-                       panic("IO err in push");
-               s = spl6();
-               bp->av_forw = bclnlist;
-               bp->b_bcount = swsize[bp - swbuf];
-               bp->b_pfcent = swpf[bp - swbuf];
-               cnt.v_pgout++;
-               cnt.v_pgpgout += bp->b_bcount / NBPG;
-               bclnlist = bp;
-               if (bswlist.b_flags & B_WANTED)
-                       wakeup((caddr_t)&proc[2]);
-               splx(s);
+       if (bp->b_flags & B_CALL) {
+               bp->b_flags &= ~B_CALL;
+               (*bp->b_iodone)(bp);

                return;
        }
        if (bp->b_flags&B_ASYNC)
                return;
        }
        if (bp->b_flags&B_ASYNC)


@@ -517,11 +480,51 @@ biodone(bp)
 }
 
 /*
 }
 
 /*

- * make sure all write-behind blocks
+ * Insure that no part of a specified block is in an incore buffer.
+ */
+blkflush(dev, blkno, size)
+       dev_t dev;
+       daddr_t blkno;
+       long size;
+{
+       register struct buf *ep;
+       struct buf *dp;
+       daddr_t start, last;
+       int s;
+
+       start = blkno;
+       last = start + btodb(size) - 1;
+       dp = BUFHASH(dev, blkno);
+loop:
+       for (ep = dp->b_forw; ep != dp; ep = ep->b_forw) {
+               if (ep->b_dev != dev || (ep->b_flags&B_INVAL))
+                       continue;
+               /* look for overlap */
+               if (ep->b_bcount == 0 || ep->b_blkno > last ||
+                   ep->b_blkno + btodb(ep->b_bcount) <= start)
+                       continue;
+               s = spl6();
+               if (ep->b_flags&B_BUSY) {
+                       ep->b_flags |= B_WANTED;
+                       sleep((caddr_t)ep, PRIBIO+1);
+                       splx(s);
+                       goto loop;
+               }
+               if (ep->b_flags & B_DELWRI) {
+                       splx(s);
+                       notavail(ep);
+                       bwrite(ep);
+                       goto loop;
+               }
+               splx(s);
+       }
+}
+
+/*
+ * Make sure all write-behind blocks

  * on dev (or NODEV for all)
  * are flushed out.
  * (from umount and update)
  * on dev (or NODEV for all)
  * are flushed out.
  * (from umount and update)

- * (and temporarily pagein)

  */
 bflush(dev)
        dev_t dev;
  */
 bflush(dev)
        dev_t dev;


@@ -532,7 +535,7 @@ bflush(dev)
 
 loop:
        s = spl6();
 
 loop:
        s = spl6();

-       for (flist = bfreelist; flist < &bfreelist[BQUEUES]; flist++)
+       for (flist = bfreelist; flist < &bfreelist[BQ_EMPTY]; flist++)

        for (bp = flist->av_forw; bp != flist; bp = bp->av_forw) {
                if ((bp->b_flags & B_DELWRI) == 0)
                        continue;
        for (bp = flist->av_forw; bp != flist; bp = bp->av_forw) {
                if ((bp->b_flags & B_DELWRI) == 0)
                        continue;


@@ -540,6 +543,7 @@ loop:
                        bp->b_flags |= B_ASYNC;
                        notavail(bp);
                        bwrite(bp);
                        bp->b_flags |= B_ASYNC;
                        notavail(bp);
                        bwrite(bp);

+                       splx(s);

                        goto loop;
                }
        }
                        goto loop;
                }
        }