/*
* shmfd object management including creation and reference counting
* routines.
*/
static struct shmfd *
shm_alloc(struct ucred *ucred, mode_t mode)
{
struct shmfd *shmfd;
shmfd = malloc(sizeof(*shmfd), M_SHMFD, M_WAITOK | M_ZERO);
shmfd->shm_size = 0;
shmfd->shm_uid = ucred->cr_uid;
shmfd->shm_gid = ucred->cr_gid;
shmfd->shm_mode = mode;
shmfd->shm_object = vm_pager_allocate(OBJT_DEFAULT, NULL,
shmfd->shm_size, VM_PROT_DEFAULT, 0, ucred);
KASSERT(shmfd->shm_object != NULL, ("shm_create: vm_pager_allocate"));
VM_OBJECT_LOCK(shmfd->shm_object);
vm_object_clear_flag(shmfd->shm_object, OBJ_ONEMAPPING);
vm_object_set_flag(shmfd->shm_object, OBJ_NOSPLIT);
VM_OBJECT_UNLOCK(shmfd->shm_object);
vfs_timestamp(&shmfd->shm_birthtime);
shmfd->shm_atime = shmfd->shm_mtime = shmfd->shm_ctime =
shmfd->shm_birthtime;
refcount_init(&shmfd->shm_refs, 1);
#ifdef MAC
mac_posixshm_init(shmfd);
mac_posixshm_create(ucred, shmfd);
#endif
return (shmfd);
}
/*
* Allocate (or lookup) pager for a vnode.
* Handle is a vnode pointer.
*
* MPSAFE
*/
vm_object_t
vnode_pager_alloc(void *handle, vm_ooffset_t size, vm_prot_t prot,
vm_ooffset_t offset, struct ucred *cred)
{
vm_object_t object;
struct vnode *vp;
/*
* Pageout to vnode, no can do yet.
*/
if (handle == NULL)
return (NULL);
vp = (struct vnode *) handle;
/*
* If the object is being terminated, wait for it to
* go away.
*/
retry:
while ((object = vp->v_object) != NULL) {
VM_OBJECT_LOCK(object);
if ((object->flags & OBJ_DEAD) == 0)
break;
vm_object_set_flag(object, OBJ_DISCONNECTWNT);
msleep(object, VM_OBJECT_MTX(object), PDROP | PVM, "vadead", 0);
}
if (vp->v_usecount == 0)
panic("vnode_pager_alloc: no vnode reference");
if (object == NULL) {
/*
* Add an object of the appropriate size
*/
object = vm_object_allocate(OBJT_VNODE, OFF_TO_IDX(round_page(size)));
object->un_pager.vnp.vnp_size = size;
object->un_pager.vnp.writemappings = 0;
object->handle = handle;
VI_LOCK(vp);
if (vp->v_object != NULL) {
/*
* Object has been created while we were sleeping
*/
VI_UNLOCK(vp);
vm_object_destroy(object);
goto retry;
}
vp->v_object = object;
VI_UNLOCK(vp);
} else {
object->ref_count++;
VM_OBJECT_UNLOCK(object);
}
vref(vp);
return (object);
}
/* Create the VM system backing object for this vnode */
int
vnode_create_vobject(struct vnode *vp, off_t isize, struct thread *td)
{
vm_object_t object;
vm_ooffset_t size = isize;
struct vattr va;
if (!vn_isdisk(vp, NULL) && vn_canvmio(vp) == FALSE)
return (0);
while ((object = vp->v_object) != NULL) {
VM_OBJECT_WLOCK(object);
if (!(object->flags & OBJ_DEAD)) {
VM_OBJECT_WUNLOCK(object);
return (0);
}
VOP_UNLOCK(vp, 0);
vm_object_set_flag(object, OBJ_DISCONNECTWNT);
VM_OBJECT_SLEEP(object, object, PDROP | PVM, "vodead", 0);
vn_lock(vp, LK_EXCLUSIVE | LK_RETRY);
}
if (size == 0) {
if (vn_isdisk(vp, NULL)) {
size = IDX_TO_OFF(INT_MAX);
} else {
if (VOP_GETATTR(vp, &va, td->td_ucred))
return (0);
size = va.va_size;
}
}
object = vnode_pager_alloc(vp, size, 0, 0, td->td_ucred);
/*
* Dereference the reference we just created. This assumes
* that the object is associated with the vp.
*/
VM_OBJECT_WLOCK(object);
object->ref_count--;
VM_OBJECT_WUNLOCK(object);
vrele(vp);
KASSERT(vp->v_object != NULL, ("vnode_create_vobject: NULL object"));
return (0);
}
/*
* Allocate a VM object for a vnode, typically a regular file vnode.
*
* Some additional information is required to generate a properly sized
* object which covers the entire buffer cache buffer straddling the file
* EOF. Userland does not see the extra pages as the VM fault code tests
* against v_filesize.
*/
vm_object_t
vnode_pager_alloc(void *handle, off_t length, vm_prot_t prot, off_t offset,
int blksize, int boff)
{
vm_object_t object;
struct vnode *vp;
off_t loffset;
vm_pindex_t lsize;
/*
* Pageout to vnode, no can do yet.
*/
if (handle == NULL)
return (NULL);
/*
* XXX hack - This initialization should be put somewhere else.
*/
if (vnode_pbuf_freecnt < 0) {
vnode_pbuf_freecnt = nswbuf / 2 + 1;
}
/*
* Serialize potential vnode/object teardowns and interlocks
*/
vp = (struct vnode *)handle;
lwkt_gettoken(&vp->v_token);
/*
* If the object is being terminated, wait for it to
* go away.
*/
object = vp->v_object;
if (object) {
vm_object_hold(object);
KKASSERT((object->flags & OBJ_DEAD) == 0);
}
if (VREFCNT(vp) <= 0)
panic("vnode_pager_alloc: no vnode reference");
/*
* Round up to the *next* block, then destroy the buffers in question.
* Since we are only removing some of the buffers we must rely on the
* scan count to determine whether a loop is necessary.
*
* Destroy any pages beyond the last buffer.
*/
if (boff < 0)
boff = (int)(length % blksize);
if (boff)
loffset = length + (blksize - boff);
else
loffset = length;
lsize = OFF_TO_IDX(round_page64(loffset));
if (object == NULL) {
/*
* And an object of the appropriate size
*/
object = vm_object_allocate_hold(OBJT_VNODE, lsize);
object->handle = handle;
vp->v_object = object;
vp->v_filesize = length;
if (vp->v_mount && (vp->v_mount->mnt_kern_flag & MNTK_NOMSYNC))
vm_object_set_flag(object, OBJ_NOMSYNC);
vref(vp);
} else {
vm_object_reference_quick(object); /* also vref's */
if (object->size != lsize) {
kprintf("vnode_pager_alloc: Warning, objsize "
"mismatch %jd/%jd vp=%p obj=%p\n",
(intmax_t)object->size,
(intmax_t)lsize,
vp, object);
}
if (vp->v_filesize != length) {
kprintf("vnode_pager_alloc: Warning, filesize "
"mismatch %jd/%jd vp=%p obj=%p\n",
(intmax_t)vp->v_filesize,
(intmax_t)length,
vp, object);
}
}
vm_object_drop(object);
lwkt_reltoken(&vp->v_token);
return (object);
}
/*
* Allocates a new node of type 'type' inside the 'tmp' mount point, with
* its owner set to 'uid', its group to 'gid' and its mode set to 'mode',
* using the credentials of the process 'p'.
*
* If the node type is set to 'VDIR', then the parent parameter must point
* to the parent directory of the node being created. It may only be NULL
* while allocating the root node.
*
* If the node type is set to 'VBLK' or 'VCHR', then the rdev parameter
* specifies the device the node represents.
*
* If the node type is set to 'VLNK', then the parameter target specifies
* the file name of the target file for the symbolic link that is being
* created.
*
* Note that new nodes are retrieved from the available list if it has
* items or, if it is empty, from the node pool as long as there is enough
* space to create them.
*
* Returns zero on success or an appropriate error code on failure.
*/
int
tmpfs_alloc_node(struct tmpfs_mount *tmp, enum vtype type,
uid_t uid, gid_t gid, mode_t mode,
char *target, int rmajor, int rminor,
struct tmpfs_node **node)
{
struct tmpfs_node *nnode;
struct timespec ts;
udev_t rdev;
KKASSERT(IFF(type == VLNK, target != NULL));
KKASSERT(IFF(type == VBLK || type == VCHR, rmajor != VNOVAL));
if (tmp->tm_nodes_inuse >= tmp->tm_nodes_max)
return (ENOSPC);
nnode = objcache_get(tmp->tm_node_pool, M_WAITOK | M_NULLOK);
if (nnode == NULL)
return (ENOSPC);
/* Generic initialization. */
nnode->tn_type = type;
vfs_timestamp(&ts);
nnode->tn_ctime = nnode->tn_mtime = nnode->tn_atime
= ts.tv_sec;
nnode->tn_ctimensec = nnode->tn_mtimensec = nnode->tn_atimensec
= ts.tv_nsec;
nnode->tn_uid = uid;
nnode->tn_gid = gid;
nnode->tn_mode = mode;
nnode->tn_id = tmpfs_fetch_ino(tmp);
nnode->tn_advlock.init_done = 0;
KKASSERT(nnode->tn_links == 0);
/* Type-specific initialization. */
switch (nnode->tn_type) {
case VBLK:
case VCHR:
rdev = makeudev(rmajor, rminor);
if (rdev == NOUDEV) {
objcache_put(tmp->tm_node_pool, nnode);
return(EINVAL);
}
nnode->tn_rdev = rdev;
break;
case VDIR:
RB_INIT(&nnode->tn_dir.tn_dirtree);
RB_INIT(&nnode->tn_dir.tn_cookietree);
nnode->tn_size = 0;
break;
case VFIFO:
/* FALLTHROUGH */
case VSOCK:
break;
case VLNK:
nnode->tn_size = strlen(target);
nnode->tn_link = kmalloc(nnode->tn_size + 1, tmp->tm_name_zone,
M_WAITOK | M_NULLOK);
if (nnode->tn_link == NULL) {
objcache_put(tmp->tm_node_pool, nnode);
return (ENOSPC);
}
bcopy(target, nnode->tn_link, nnode->tn_size);
nnode->tn_link[nnode->tn_size] = '\0';
break;
case VREG:
nnode->tn_reg.tn_aobj = swap_pager_alloc(NULL, 0,
VM_PROT_DEFAULT, 0);
nnode->tn_reg.tn_aobj_pages = 0;
nnode->tn_size = 0;
vm_object_set_flag(nnode->tn_reg.tn_aobj, OBJ_NOPAGEIN);
break;
default:
panic("tmpfs_alloc_node: type %p %d", nnode, (int)nnode->tn_type);
}
TMPFS_NODE_LOCK(nnode);
TMPFS_LOCK(tmp);
LIST_INSERT_HEAD(&tmp->tm_nodes_used, nnode, tn_entries);
tmp->tm_nodes_inuse++;
TMPFS_UNLOCK(tmp);
TMPFS_NODE_UNLOCK(nnode);
*node = nnode;
return 0;
}
/*
* The caller must hold proc_token.
*/
static int
do_vmtotal_callback(struct proc *p, void *data)
{
struct vmtotal *totalp = data;
struct lwp *lp;
vm_map_entry_t entry;
vm_map_t map;
int paging;
if (p->p_flag & P_SYSTEM)
return(0);
FOREACH_LWP_IN_PROC(lp, p) {
switch (lp->lwp_stat) {
case LSSTOP:
case LSSLEEP:
if ((p->p_flag & P_SWAPPEDOUT) == 0) {
if ((lp->lwp_flag & LWP_SINTR) == 0)
totalp->t_dw++;
else if (lp->lwp_slptime < maxslp)
totalp->t_sl++;
} else if (lp->lwp_slptime < maxslp) {
totalp->t_sw++;
}
if (lp->lwp_slptime >= maxslp)
return(0);
break;
case LSRUN:
if (p->p_flag & P_SWAPPEDOUT)
totalp->t_sw++;
else
totalp->t_rq++;
if (p->p_stat == SIDL)
return(0);
break;
default:
return (0);
}
}
/*
* Note active objects.
*/
paging = 0;
lwkt_gettoken(&vm_token);
if (p->p_vmspace) {
map = &p->p_vmspace->vm_map;
vm_map_lock_read(map);
for (entry = map->header.next;
entry != &map->header; entry = entry->next) {
if (entry->maptype != VM_MAPTYPE_NORMAL &&
entry->maptype != VM_MAPTYPE_VPAGETABLE) {
continue;
}
if (entry->object.vm_object == NULL)
continue;
vm_object_set_flag(entry->object.vm_object, OBJ_ACTIVE);
paging |= entry->object.vm_object->paging_in_progress;
}
vm_map_unlock_read(map);
}
lwkt_reltoken(&vm_token);
if (paging)
totalp->t_pw++;
return(0);
}