vnode.c

Solaris操作系统下的过滤驱动程序, C源码程序.

  FIST_OP_LOOKUP_POSTCALL;

#if 0
  if ((thiscnp->cn_flags & PARAMASK) != (lowercnp->cn_flags & PARAMASK)) {
    char buf[256];
    strcpy(buf, fist_cn_flags(thiscnp->cn_flags));
    panic("LK: this 0x%x <%s>, lower 0x%x <%s>",
	  thiscnp->cn_flags,
	  buf,
	  lowercnp->cn_flags,
	  fist_cn_flags(lowercnp->cn_flags));
  }
#endif
#ifdef FIST_FILTER_NAME
  thiscnp->cn_flags = lowercnp->cn_flags;
#endif /* FIST_FILTER_NAME */
#if 0
  if (had_flags)
    thiscnp->cn_flags |= had_flags;
  else
    thiscnp->cn_flags &= ~had_flags;
#endif
  CNP_AFTER(ap->a_dvp);

#ifdef FIST_FILTER_NAME
  if (!error &&
      (thiscnp->cn_flags & MAKEENTRY) &&
      thiscnp->cn_nameiop != CREATE) {
    fist_dprint(1, "LK cache_enter\n");
    cache_enter(ap->a_dvp, *(ap->a_vpp), thiscnp);
  }
#endif /* FIST_FILTER_NAME */

  fist_dprint(1, "LK1 %d (dvp=0x%x)\n", counter, ap->a_dvp);

  if (error == EJUSTRETURN && (flags & ISLASTCN) &&
      (ap->a_dvp->v_mount->mnt_flag & MNT_RDONLY) &&
      (cnp->cn_nameiop == CREATE || cnp->cn_nameiop == RENAME))
    error = EROFS;
  /*
   * We must do the same locking and unlocking at this layer as
   * is done in the layers below us. We could figure this out
   * based on the error return and the LASTCN, LOCKPARENT, and
   * LOCKLEAF flags. However, it is more expidient to just find
   * out the state of the lower level vnodes and set ours to the
   * same state.
   */
  dvp = ap->a_dvp;
  vp = *ap->a_vpp;
  if (dvp == vp)
    return (error);
  if (!VOP_ISLOCKED(dvp, NULL)) {
    unlockargs.a_vp = dvp;
    unlockargs.a_flags = 0;
    unlockargs.a_p = p;
    vop_nounlock(&unlockargs);
  }
  if (vp != NULLVP && VOP_ISLOCKED(vp, NULL)) {
    lockargs.a_vp = vp;
    lockargs.a_flags = LK_SHARED;
    lockargs.a_p = p;
    vop_nolock(&lockargs);
  }
  return (error);
}

/*
 * Setattr call. Disallow write attempts if the layer is mounted read-only.
 */
static int
wrapfs_setattr(ap)
     struct vop_setattr_args /* {
				struct vnodeop_desc *a_desc;
				struct vnode *a_vp;
				struct vattr *a_vap;
				struct ucred *a_cred;
				struct proc *a_p;
				} */ *ap;
{
  struct vnode *vp = ap->a_vp;
  struct vattr *vap = ap->a_vap;

  fist_dprint(4, "FXN=%s FILE=%s LINE=%d\n",__FUNCTION__,__FILE__,__LINE__);

  if ((vap->va_flags != VNOVAL || vap->va_uid != (uid_t) VNOVAL ||
       vap->va_gid != (gid_t) VNOVAL || vap->va_atime.tv_sec != VNOVAL ||
       vap->va_mtime.tv_sec != VNOVAL || vap->va_mode != (mode_t) VNOVAL) &&
      (vp->v_mount->mnt_flag & MNT_RDONLY))
    return (EROFS);
  if (vap->va_size != VNOVAL) {
    switch (vp->v_type) {
    case VDIR:
      return (EISDIR);
    case VCHR:
    case VBLK:
    case VSOCK:
    case VFIFO:
      if (vap->va_flags != VNOVAL)
	return (EOPNOTSUPP);
      return (0);
    case VREG:
    case VLNK:
    default:
      /*
       * Disallow write attempts if the filesystem is
       * mounted read-only.
       */
      if (vp->v_mount->mnt_flag & MNT_RDONLY)
	return (EROFS);
    }
  }
  return (wrapfs_bypass((struct vop_generic_args *) ap));
}

/*
 *  We handle getattr only to change the fsid.
 */
static int
wrapfs_getattr(ap)
     struct vop_getattr_args /* {
				struct vnode *a_vp;
				struct vattr *a_vap;
				struct ucred *a_cred;
				struct proc *a_p;
				} */ *ap;
{
  int error;

  fist_dprint(4, "WRAPFS_GETATTR: vp=0x%x\n", (int) ap->a_vp);

  if ((error = wrapfs_bypass((struct vop_generic_args *) ap)) != 0) {
    printf("bypass getattr returned error %d\n", error);
    return (error);
  }
  return (0);
}

static int
wrapfs_access(ap)
     struct vop_access_args /* {
			       struct vnode *a_vp;
			       int a_mode;
			       struct ucred *a_cred;
			       struct proc *a_p;
			       } */ *ap;
{
  struct vnode *vp = ap->a_vp;
  mode_t mode = ap->a_mode;

  fist_dprint(4, "FXN=%s FILE=%s LINE=%d\n",__FUNCTION__,__FILE__,__LINE__);

  /*
   * Disallow write attempts on read-only layers;
   * unless the file is a socket, fifo, or a block or
   * character device resident on the file system.
   */
  if (mode & VWRITE) {
    switch (vp->v_type) {
    case VDIR:
    case VLNK:
    case VREG:
      if (vp->v_mount->mnt_flag & MNT_RDONLY)
	return (EROFS);
      break;
    default:
      break;
    }
  }
  return (wrapfs_bypass((struct vop_generic_args *) ap));
}

/*
 * We need to process our own vnode lock and then clear the
 * interlock flag as it applies only to our vnode, not the
 * vnodes below us on the stack.
 */
static int
wrapfs_lock(ap)
     struct vop_lock_args /* {
			     struct vnode *a_vp;
			     int a_flags;
			     struct proc *a_p;
			     } */ *ap;
{

  fist_dprint(4, "FXN=%s FILE=%s LINE=%d\n",__FUNCTION__,__FILE__,__LINE__);
  vop_nolock(ap);
  if ((ap->a_flags & LK_TYPE_MASK) == LK_DRAIN)
    return (0);
  ap->a_flags &= ~LK_INTERLOCK;
  return (wrapfs_bypass((struct vop_generic_args *) ap));
}

/*
 * We need to process our own vnode unlock and then clear the
 * interlock flag as it applies only to our vnode, not the
 * vnodes below us on the stack.
 */
static int
wrapfs_unlock(ap)
     struct vop_unlock_args /* {
			       struct vnode *a_vp;
			       int a_flags;
			       struct proc *a_p;
			       } */ *ap;
{
  /*   struct vnode *vp = ap->a_vp; */

  fist_dprint(4, "FXN=%s FILE=%s LINE=%d\n",__FUNCTION__,__FILE__,__LINE__);
  vop_nounlock(ap);
  ap->a_flags &= ~LK_INTERLOCK;
  return (wrapfs_bypass((struct vop_generic_args *) ap));
}

static int
wrapfs_inactive(ap)
     struct vop_inactive_args /* {
				 struct vnode *a_vp;
				 struct proc *a_p;
				 } */ *ap;
{
  struct vnode *vp = ap->a_vp;
  struct wrapfs_node *xp = VP_TO_WRAPFS(vp);
  struct vnode *lowervp = xp->wrapfs_lowervp;

  fist_dprint(4, "FXN=%s FILE=%s LINE=%d\n",__FUNCTION__,__FILE__,__LINE__);

  /*
   * Do nothing (and _don't_ bypass).
   * Wait to vrele lowervp until reclaim,
   * so that until then our wrapfs_node is in the
   * cache and reusable.
   * We still have to tell the lower layer the vnode
   * is now inactive though.
   *
   * NEEDSWORK: Someday, consider inactive'ing
   * the lowervp and then trying to reactivate it
   * with capabilities (v_id)
   * like they do in the name lookup cache code.
   * That's too much work for now.
   */
  VOP_INACTIVE(lowervp, ap->a_p);
  VOP_UNLOCK(ap->a_vp, 0, ap->a_p);
  return (0);
}

static int
wrapfs_reclaim(ap)
     struct vop_reclaim_args /* {
				struct vnode *a_vp;
				struct proc *a_p;
				} */ *ap;
{
  struct vnode *vp = ap->a_vp;
  struct wrapfs_node *xp = VP_TO_WRAPFS(vp);
  struct vnode *lowervp = xp->wrapfs_lowervp;

  fist_dprint(4, "FXN=%s FILE=%s LINE=%d\n",__FUNCTION__,__FILE__,__LINE__);

  /*
   * Note: in vop_reclaim, vp->v_op == dead_vnodeop_p,
   * so we can't call VOPs on ourself.
   */
  /* After this assignment, this node will not be re-used. */
  xp->wrapfs_lowervp = NULLVP;
  LIST_REMOVE(xp, wrapfs_hash);
  FREE(vp->v_data, M_TEMP);
  vp->v_data = NULL;
  vrele(lowervp);
  return (0);
}

static int
wrapfs_print(ap)
     struct vop_print_args /* {
			      struct vnode *a_vp;
			      } */ *ap;
{
  register struct vnode *vp = ap->a_vp;
  printf("\ttag VT_WRAPFS, vp=%p, lowervp=%p\n", vp, WRAPFS_VP_TO_LOWERVP(vp));
  return (0);
}


/****************************************************************************/
/* verify that lowervp has an v_object.  create it if needed */
void
wrapfs_verify_lower_object(vnode_t *vp, char *fxn)
{
  vnode_t *lowervp = WRAPFS_VP_TO_LOWERVP(vp);

  /* allocate a duplicate vnode pager for lower vp if needed */
  if (lowervp->v_object) {
    return;
  }
  lowervp->v_object = vm_pager_allocate(vp->v_object->type,
					lowervp,
					vp->v_object->size,
					VM_PROT_ALL,
					0LL);
  fist_dprint(2, "VERIFY_OBJECT: fxn=%s size=%d obj=0x%x lowervp=0x%x vp=0x%x\n",
	      fxn,
	      vp->v_object->size,
	      (int) lowervp->v_object,
	      (int) lowervp,
	      (int) vp);
}

/*
 * Create a page in the vnode and insert data into it.
 * This is used to synchronize data between the VM and read/write interface.
 * That that was written using write() gets pages containing the same data, so
 * that subsequent mmap() ops get valid data.
 */
void
wrapfs_fill_page(vnode_t *vp, char *buf, long long offset)
{
  vm_page_t pp;
  vm_offset_t kva;
  caddr_t ca;

  fist_dprint(4, "FILL_PAGE: vp=0x%x, buf=0x%x, offset=0x%x\n",
	      (int) vp, (int) buf, buf[0], buf[1], buf[2], (int) offset);
  fist_dprint(1, "FILL_PAGE: vp=0x%x, buf=0x%x [%d,%d,%d], offset=0x%x\n",
	      (int) vp, (int) buf, buf[0], buf[1], buf[2], (int) offset);

  pp = vm_page_grab(vp->v_object, OFF_TO_IDX(offset),
		    VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
  if (!pp) {
    printf("vm_page_grab returned NULL for offset 0x%x!\n", (int) offset);
    return;
  }
  kva = vm_pager_map_page(pp);
  ca = (caddr_t) kva;
  bcopy(buf, ca, PAGE_SIZE);
  vm_pager_unmap_page(kva);
  vm_page_set_validclean(pp, 0, PAGE_SIZE);
  pp->flags &= ~PG_BUSY;
}

void
wrapfs_fill_lowerpage(vnode_t *lowervp, char *buf, long long offset)
{
  vm_page_t pp;
  vm_offset_t kva;
  caddr_t ca;

  fist_dprint(4, "FILL_LOWERPAGE: lowervp=0x%x, buf=0x%x, offset=0x%x\n",
	      (int) lowervp, (int) buf, buf[0], buf[1], buf[2], (int) offset);
  fist_dprint(1, "FILL_LOWERPAGE: vp=0x%x, buf=0x%x [%d,%d,%d], offset=0x%x\n",
	      (int) lowervp, (int) buf, buf[0], buf[1], buf[2], (int) offset);

  pp = vm_page_grab(lowervp->v_object, OFF_TO_IDX(offset),
		    VM_ALLOC_NORMAL | VM_ALLOC_RETRY);
  if (!pp) {
    printf("vm_page_grab2 returned NULL for offset 0x%x!\n", (int) offset);
    return;
  }
  kva = vm_pager_map_page(pp);
  ca = (caddr_t) kva;
  bcopy(buf, ca, PAGE_SIZE);
  vm_pager_unmap_page(kva);
  vm_page_set_validclean(pp, 0, PAGE_SIZE);
  pp->flags &= ~PG_BUSY;
}

static int
wrapfs_read(ap)
     struct vop_read_args /* {
			     struct vnode *a_vp;
			     struct uio *a_uio;
			     int  a_ioflag;
			     struct ucred *a_cred;
			     } */ *ap;
{
  /* easy mappings */
  vnode_t *vp = ap->a_vp;
  uio_t *uiop = ap->a_uio;
  int ioflag = ap->a_ioflag;
  cred_t *cr = ap->a_cred;

  int error = EPERM;
  vnode_t *hidden_vp;
  uio_t temp_uio;
  iovec_t *temp_iovec;
  caddr_t current_base;
  int i, bytes_read;
  int num_pages, resid;
  long long start_loffset, end_loffset;
  long long cleartext_start_loffset, cleartext_end_loffset, current_loffset;

  fist_dprint(4, "fist_wrapfs_read vp %x\n", (int) vp);

#ifdef FIST_DEBUG
  fist_print_uios("fist_wrapfs_read", uiop);
#endif /* FIST_DEBUG */

  cleartext_start_loffset = uiop->uio_offset;
  cleartext_end_loffset = uiop->uio_offset + uiop->uio_resid;
  start_loffset = cleartext_start_loffset & ~(PAGE_SIZE - 1);
  end_loffset = cleartext_end_loffset & ~(PAGE_SIZE - 1);
  /* if not multiple of PAGE_SIZE, then the above formula loses one page.
   * adjust for it */
  if (cleartext_end_loffset > end_loffset)
    end_loffset += PAGE_SIZE;
  resid = end_loffset - start_loffset;
  num_pages = resid >> PAGE_SHIFT;

  fist_dprint(6,
	      "READ: so=%d eo=%d cs=%d es=%d res=%d np=%d ps=%d\n",
	      (int) start_loffset,
	      (int) end_loffset,
	      (int) cleartext_start_loffset,