freebsd-3.2.il-kernel.patch

这是一个同样来自贝尔实验室的和UNIX有着渊源的操作系统, 其简洁的设计和实现易于我们学习和理解

+ 		/*
+ 		 * Start the read ahead(s), as required.
+ 		 */
+ 		if (u9fs_numasync > 0 && nmp->nm_readahead > 0) {
+ 		    for (nra = 0; nra < nmp->nm_readahead &&
+ 			(off_t)(lbn + 1 + nra) * biosize < np->n_size; nra++) {
+ 			rabn = lbn + 1 + nra;
+ 			if (!incore(vp, rabn)) {
+ 			    rabp = u9fs_getcacheblk(vp, rabn, biosize, p);
+ 			    if (!rabp)
+ 				return (EINTR);
+ 			    if ((rabp->b_flags & (B_CACHE|B_DELWRI)) == 0) {
+ 				rabp->b_flags |= (B_READ | B_ASYNC);
+ 				vfs_busy_pages(rabp, 0);
+ 				if (u9fs_asyncio(rabp, cred)) {
+ 				    rabp->b_flags |= B_INVAL|B_ERROR;
+ 				    vfs_unbusy_pages(rabp);
+ 				    brelse(rabp);
+ 				}
+ 			    } else
+ 				brelse(rabp);
+ 			}
+ 		    }
+ 		}
+ #endif
+ 
+ 		/*
+ 		 * If the block is in the cache and has the required data
+ 		 * in a valid region, just copy it out.
+ 		 * Otherwise, get the block and write back/read in,
+ 		 * as required.
+ 		 */
+ again:
+ 		bufsize = biosize;
+ 		if ((off_t)(lbn + 1) * biosize > np->n_size && 
+ 		    (off_t)(lbn + 1) * biosize - np->n_size < biosize) {
+ 			bufsize = np->n_size - (off_t)lbn * biosize;
+ 			bufsize = (bufsize + DEV_BSIZE - 1) & ~(DEV_BSIZE - 1);
+ 		}
+ 		bp = u9fs_getcacheblk(vp, lbn, bufsize, p);
+ 		if (!bp)
+ 			return (EINTR);
+ 		/*
+ 		 * If we are being called from u9fs_getpages, we must
+ 		 * make sure the buffer is a vmio buffer.  The vp will
+ 		 * already be setup for vmio but there may be some old
+ 		 * non-vmio buffers attached to it.
+ 		 */
+ 		if (getpages && !(bp->b_flags & B_VMIO)) {
+ #ifdef DIAGNOSTIC
+ 			printf("u9fs_bioread: non vmio buf found, discarding\n");
+ #endif
+ 			bp->b_flags |= B_NOCACHE;
+ 			if (bp->b_dirtyend > 0) {
+ 				if ((bp->b_flags & B_DELWRI) == 0)
+ 					panic("u9fsbioread");
+ 				if (VOP_BWRITE(bp) == EINTR)
+ 					return (EINTR);
+ 			} else
+ 				brelse(bp);
+ 			goto again;
+ 		}
+ 		if ((bp->b_flags & B_CACHE) == 0) {
+ 		    bp->b_flags |= B_READ;
+ 		    bp->b_flags &= ~(B_DONE | B_ERROR | B_INVAL);
+ 		    not_readin = 0;
+ 		    vfs_busy_pages(bp, 0);
+ 		    error = u9fs_doio(bp, cred, p);
+ 		    if (error) {
+ 			brelse(bp);
+ 			return (error);
+ 		    }
+ 		    np->n_qid.vers = np->n_dir.dir_qid.vers;
+ 		}
+ 		if (bufsize > on) {
+ 			n = min((unsigned)(bufsize - on), uio->uio_resid);
+ 		} else {
+ 			n = 0;
+ 		}
+ 		diff = np->n_size - uio->uio_offset;
+ 		if (diff < n)
+ 			n = diff;
+ 		if (not_readin && n > 0) {
+ 			if (on < bp->b_validoff || (on + n) > bp->b_validend) {
+ 				bp->b_flags |= B_NOCACHE;
+ 				if (bp->b_dirtyend > 0) {
+ 				    if ((bp->b_flags & B_DELWRI) == 0)
+ 					panic("u9fsbioread");
+ 				    if (VOP_BWRITE(bp) == EINTR)
+ 					return (EINTR);
+ 				} else
+ 				    brelse(bp);
+ 				goto again;
+ 			}
+ 		}
+ 		vp->v_lastr = lbn;
+ 		diff = (on >= bp->b_validend) ? 0 : (bp->b_validend - on);
+ 		if (diff < n)
+ 			n = diff;
+ 		break;
+ 	    case VDIR:
+ 	        biosize = nmp->nm_readdirsize;
+ 		lbn = (uoff_t)uio->uio_offset / biosize;
+ 		on = uio->uio_offset % biosize;
+ 		bp = u9fs_getcacheblk(vp, lbn, biosize, p);
+ 		if (!bp)
+ 		    return (EINTR);
+ 		if ((bp->b_flags & B_CACHE) == 0) {
+ 		    bp->b_flags |= B_READ;
+ 		    vfs_busy_pages(bp, 0);
+ 		    error = u9fs_doio(bp, cred, p);
+ 		    if (error) {
+ 			    brelse(bp);
+ 		    }
+ 		    if (error)
+ 			    return (error);
+ 		    np->n_qid.vers = np->n_dir.dir_qid.vers;
+ 		}
+ 
+ 		/*
+ 		 * Make sure we use a signed variant of min() since
+ 		 * the second term may be negative.
+ 		 */
+ 		n = lmin(uio->uio_resid, biosize - bp->b_resid - on);
+ 		break;
+ 	    default:
+ 		printf(" u9fs_bioread: type %x unexpected\n",vp->v_type);
+ 		break;
+ 	    };
+ 
+ 	    if (n > 0) {
+ 		    error = uiomove(bp->b_data + on, (int)n, uio);
+ 	    }
+ 	    brelse(bp);
+ 	} while (error == 0 && uio->uio_resid > 0 && n > 0);
+ 	return (error);
+ }
+ 
+ /*
+  * Vnode op for write using bio
+  */
+ int
+ u9fs_biowrite(vp, uio, ioflag, cred)
+      register struct vnode *vp;
+      register struct uio *uio;
+      register int  ioflag;
+      register struct ucred *cred;
+ {
+   register int biosize;
+   struct proc *p = uio->uio_procp;
+   struct u9fsnode *np = VTOU9FS(vp);
+   struct buf *bp;
+   struct vattr vattr;
+   struct u9fsmount *nmp = VFSTOU9FS(vp->v_mount);
+   daddr_t lbn;
+   int bufsize;
+   int n, on, error = 0;
+ 
+   if (ioflag & (IO_APPEND | IO_SYNC)) {
+     if (ioflag & IO_APPEND) {
+       error = VOP_GETATTR(vp, &vattr, cred, p);
+       if (error)
+ 	return (error);
+       uio->uio_offset = np->n_size;
+     }
+   }
+   if (uio->uio_offset < 0)
+     return (EINVAL);
+   if ((uio->uio_offset + uio->uio_resid) > nmp->nm_maxfilesize)
+     return (EFBIG);
+   if (uio->uio_resid == 0)
+     return (0);
+ 
+ 	/*
+ 	 * I use nm_rsize, not nm_wsize so that all buffer cache blocks
+ 	 * will be the same size within a filesystem. nfs_writerpc will
+ 	 * still use nm_wsize when sizing the rpc's.
+ 	 */
+ 	biosize = vp->v_mount->mnt_stat.f_iosize;
+ 	do {
+ 		lbn = uio->uio_offset / biosize;
+ 		on = uio->uio_offset & (biosize-1);
+ 		n = min((unsigned)(biosize - on), uio->uio_resid);
+ 		if (uio->uio_offset + n > np->n_size) {
+ 			np->n_size = uio->uio_offset + n;
+ 			vnode_pager_setsize(vp, np->n_size);
+ 		}
+ 		bufsize = biosize;
+ 		if ((off_t)(lbn + 1) * biosize > np->n_size) {
+ 			bufsize = np->n_size - (off_t)lbn * biosize;
+ 			bufsize = (bufsize + DEV_BSIZE - 1) & ~(DEV_BSIZE - 1);
+ 		}
+ 		bp = u9fs_getcacheblk(vp, lbn, bufsize, p);
+ 		if (!bp)
+ 			return (EINTR);
+ 		if (bp->b_wcred == NOCRED) {
+ 			crhold(cred);
+ 			bp->b_wcred = cred;
+ 		}
+ 
+ 		error = uiomove((char *)bp->b_data + on, n, uio);
+ 		if (error) {
+ 			bp->b_flags |= B_ERROR;
+ 			brelse(bp);
+ 			return (error);
+ 		}
+ 
+ 		/*
+ 		 * This will keep the buffer and mmaped regions more coherent.
+ 		 */
+ 		u9fs_prot_buf(bp, on, n);
+ 		bp->b_dirtyoff = on;
+ 		bp->b_dirtyend = on + n;
+ 
+ 		if (bp->b_validend == 0 || bp->b_validend < bp->b_dirtyoff ||
+ 		    bp->b_validoff > bp->b_dirtyend) {
+ 		  /* XXX: destroys our read cache if not overlapping */
+ 		  /* two choice: none implemented
+ 		     1> keep the bigger(smaller) piece
+ 		     2> read the missing segment		     
+ 		  */
+ 			bp->b_validoff = bp->b_dirtyoff;
+ 			bp->b_validend = bp->b_dirtyend;
+ 		} else {
+ 			bp->b_validoff = min(bp->b_validoff, bp->b_dirtyoff);
+ 			bp->b_validend = max(bp->b_validend, bp->b_dirtyend);
+ 		}
+ 
+ 		error = bwrite(bp);
+ 		if( error ) {
+ 		  bp->b_flags |= B_ERROR;
+ 		  /* brelse(bp); */
+ 		  return error;		  
+ 		}
+ 	} while (uio->uio_resid > 0 && n > 0);
+ 	return 0;
+ }
+ 
+ /*
+  * Do an I/O operation to/from a cache block. This may be called
+  * synchronously or from an u9fsiod.
+  */
+ int
+ u9fs_doio(bp, cr, p)
+ 	register struct buf *bp;
+ 	struct ucred *cr;
+ 	struct proc *p;
+ {
+ 	register struct uio *uiop;
+ 	register struct vnode *vp;
+ 	struct u9fsnode *np;
+ 	struct u9fsmount *nmp;
+ 	int error = 0, diff, len;
+ 	struct uio uio;
+ 	struct iovec io;
+ 
+ 	vp = bp->b_vp;
+ 	np = VTOU9FS(vp);
+ 	nmp = VFSTOU9FS(vp->v_mount);
+ 	uiop = &uio;
+ 	uiop->uio_iov = &io;
+ 	uiop->uio_iovcnt = 1;
+ 	uiop->uio_segflg = UIO_SYSSPACE;
+ 	uiop->uio_procp = p;
+ 
+ 	if (bp->b_flags & B_READ ) {
+ 	    io.iov_len = uiop->uio_resid = bp->b_bcount;
+ 	    io.iov_base = bp->b_data;
+ 	    uiop->uio_rw = UIO_READ;
+ 	    switch (vp->v_type) {
+ 	    case VREG:
+ 		uiop->uio_offset = ((off_t)bp->b_blkno) * DEV_BSIZE;
+ 		error = u9fs_readrpc(vp, uiop, cr);
+ 		if (!error) {
+ 		    bp->b_validoff = 0;
+ 		    if (uiop->uio_resid) {
+ 			/*
+ 			 * If len > 0, there is a hole in the file and
+ 			 * no writes after the hole have been pushed to
+ 			 * the server yet.
+ 			 * Just zero fill the rest of the valid area.
+ 			 */
+ 			diff = bp->b_bcount - uiop->uio_resid;
+ 			len = np->n_size - (((u_quad_t)bp->b_blkno) * DEV_BSIZE
+ 				+ diff);
+ 			if (len > 0) {
+ 			    len = min(len, uiop->uio_resid);
+ 			    bzero((char *)bp->b_data + diff, len);
+ 			    bp->b_validend = diff + len;
+ 			} else
+ 			    bp->b_validend = diff;
+ 		    } else
+ 			bp->b_validend = bp->b_bcount;
+ 		}
+ 		break;
+ 	    case VDIR:
+ 		uiop->uio_offset = ((u_quad_t)bp->b_lblkno) * nmp->nm_readdirsize;
+ 		error = u9fs_readdirrpc(vp, uiop, cr);
+ 		if (error == 0 && uiop->uio_resid == bp->b_bcount)
+ 		  bp->b_flags |= B_INVAL;
+ 		break;
+ 	    default:
+ 		printf("u9fs_doio:  type %x unexpected\n",vp->v_type);
+ 		break;
+ 	    };
+ 	    if (error) {
+ 		bp->b_flags |= B_ERROR;
+ 		bp->b_error = error;
+ 	    }
+ 	} else {
+ 	    if ((off_t)bp->b_blkno * DEV_BSIZE + bp->b_dirtyend > np->n_size)
+ 		bp->b_dirtyend = np->n_size - (off_t)bp->b_blkno * DEV_BSIZE;
+ 
+ 	    if (bp->b_dirtyend > bp->b_dirtyoff) {
+ 		io.iov_len = uiop->uio_resid = bp->b_dirtyend
+ 		    - bp->b_dirtyoff;
+ 		uiop->uio_offset = (off_t)bp->b_blkno * DEV_BSIZE
+ 		    + bp->b_dirtyoff;
+ 		io.iov_base = (char *)bp->b_data + bp->b_dirtyoff;
+ 		uiop->uio_rw = UIO_WRITE;
+ 		bp->b_flags |= B_WRITEINPROG;
+ 		error = u9fs_writerpc(vp, uiop, cr);
+ 		bp->b_flags &= ~B_WRITEINPROG;
+ 
+ 		if (error) {
+ 		  bp->b_flags |= B_ERROR;
+ 		  bp->b_error = np->n_error = error;
+ 		  np->n_flag |= NWRITEERR;
+ 		}
+ 		bp->b_dirtyoff = bp->b_dirtyend = 0;
+ 	    } else {
+ 		bp->b_resid = 0;
+ 		biodone(bp);
+ 		return (0);
+ 	    }
+ 	}
+ 	bp->b_resid = uiop->uio_resid;
+ 	biodone(bp);
+ 	return error;
+ }
+ 
+ /*
+  * Get an u9fs cache block.
+  * Allocate a new one if the block isn't currently in the cache
+  * and return the block marked busy. If the calling process is
+  * interrupted by a signal for an interruptible mount point, return
+  * NULL.
+  */
+ static struct buf *
+ u9fs_getcacheblk(vp, bn, size, p)
+ 	struct vnode *vp;
+ 	daddr_t bn;
+ 	int size;
+ 	struct proc *p;
+ {
+ 	register struct buf *bp;
+ 	struct mount *mp;
+ 	struct u9fsmount *nmp;
+ 
+ 	mp = vp->v_mount;
+ 	nmp = VFSTOU9FS(mp);
+ 
+ 	if (nmp->nm_flag & U9FSMNT_INT) {
+ 		bp = getblk(vp, bn, size, PCATCH, 0);
+ 		while (bp == (struct buf *)0) {
+ 			if (u9fs_sigintr(nmp, p))
+ 				return ((struct buf *)0);
+ 			bp = getblk(vp, bn, size, 0, 2 * hz);
+ 		}
+ 	} else
+ 		bp = getblk(vp, bn, size, 0, 0);
+ 
+ 	if (vp->v_type == VREG) {
+ 		int biosize;
+ 		biosize = mp->mnt_stat.f_iosize;
+ 		bp->b_blkno = bn * (biosize / DEV_BSIZE);
+ 	}
+ 
+ 	return (bp);
+ }
+ 
+ static void
+ u9fs_prot_buf(bp, off, n)
+ 	struct buf *bp;
+ 	int off;
+ 	int n;
+ {
+ 	int pindex, boff, end;
+ 
+ 	if ((bp->b_flags & B_VMIO) == 0)
+ 		return;
+ 
+ 	end = round_page(off + n);
+ 	for (boff = trunc_page(off); boff < end; boff += PAGE_SIZE) {
+ 		pindex = boff >> PAGE_SHIFT;
+ 		vm_page_protect(bp->b_pages[pindex], VM_PROT_NONE);
+ 	}
+ }