Differential D37097 Diff 112141 sys/fs/tmpfs/tmpfs_subr.c

Changeset View

Standalone View

sys/fs/tmpfs/tmpfs_subr.c

Show First 20 Lines • Show All 107 Lines • ▼ Show 20 Lines
static void		static void
tmpfs_pager_writecount_recalc(vm_object_t object, vm_offset_t old,		tmpfs_pager_writecount_recalc(vm_object_t object, vm_offset_t old,
vm_offset_t new)		vm_offset_t new)
{		{
struct vnode *vp;		struct vnode *vp;

VM_OBJECT_ASSERT_WLOCKED(object);		VM_OBJECT_ASSERT_WLOCKED(object);

vp = object->un_pager.swp.swp_tmpfs;		vp = VM_TO_TMPFS_VP(object);

/*		/*
* Forced unmount?		* Forced unmount?
*/		*/
if (vp == NULL) {		if (vp == NULL) {
KASSERT((object->flags & OBJ_TMPFS_VREF) == 0,		KASSERT((object->flags & OBJ_TMPFS_VREF) == 0,
("object %p with OBJ_TMPFS_VREF but without vnode",		("object %p with OBJ_TMPFS_VREF but without vnode",
object));		object));
▲ Show 20 Lines • Show All 64 Lines • ▼ Show 20 Lines

static void		static void
tmpfs_pager_getvp(vm_object_t object, struct vnode *vpp, bool vp_heldp)		tmpfs_pager_getvp(vm_object_t object, struct vnode *vpp, bool vp_heldp)
{		{
struct vnode *vp;		struct vnode *vp;

/*		/*
* Tmpfs VREG node, which was reclaimed, has tmpfs_pager_type		* Tmpfs VREG node, which was reclaimed, has tmpfs_pager_type
* type, but not OBJ_TMPFS flag. In this case there is no		* type. In this case there is no v_writecount to adjust.
* v_writecount to adjust.
*/		*/
if (vp_heldp != NULL)		if (vp_heldp != NULL)
VM_OBJECT_RLOCK(object);		VM_OBJECT_RLOCK(object);
else		else
VM_OBJECT_ASSERT_LOCKED(object);		VM_OBJECT_ASSERT_LOCKED(object);
if ((object->flags & OBJ_TMPFS) != 0) {		if ((object->flags & OBJ_TMPFS) != 0) {
vp = object->un_pager.swp.swp_tmpfs;		vp = VM_TO_TMPFS_VP(object);
if (vp != NULL) {		if (vp != NULL) {
*vpp = vp;		*vpp = vp;
if (vp_heldp != NULL) {		if (vp_heldp != NULL) {
vhold(vp);		vhold(vp);
*vp_heldp = true;		*vp_heldp = true;
}		}
}		}
}		}
if (vp_heldp != NULL)		if (vp_heldp != NULL)
VM_OBJECT_RUNLOCK(object);		VM_OBJECT_RUNLOCK(object);
}		}

		static void
		tmpfs_page_inserted(vm_object_t obj, vm_page_t m)
		{
		struct tmpfs_node *node;
		struct tmpfs_mount *tm;

		if ((obj->flags & OBJ_TMPFS) == 0)
		return;

		node = obj->un_pager.swp.swp_priv;
		MPASS(node->tn_type == VREG);
		tm = node->tn_reg.tn_tmp;

		if (!vm_pager_has_page(obj, m->pindex, NULL, NULL)) {
		atomic_add_long(&tm->tm_pages_used, 1);
		node->tn_reg.tn_pages += 1;
		}
		}

		static void
		tmpfs_page_removed(vm_object_t obj, vm_page_t m)
		{
		struct tmpfs_node *node;
		struct tmpfs_mount *tm;

		if ((obj->flags & OBJ_TMPFS) == 0)
		return;

		node = obj->un_pager.swp.swp_priv;
		MPASS(node->tn_type == VREG);
		tm = node->tn_reg.tn_tmp;

		if (!vm_pager_has_page(obj, m->pindex, NULL, NULL)) {
		atomic_add_long(&tm->tm_pages_used, -1);
		node->tn_reg.tn_pages -= 1;
		}
		markjUnsubmitted Done Inline Actions Suppose I write 4096 bytes to a new tmpfs file, so tn_pages == 1 and the page is dirty. Suppose the page is laundered, so a copy exists on the swap device. Suppose that the page is not freed after this point (but it remains clean). Then I truncate the file to 0 bytes. vm_object_page_remove() will free the page, but here we will not decrement tn_pages because the pager still has a copy. Then vm_object_page_remove() frees swap space in the truncated region, but nothing will decrement tn_pages, I think. markj: Suppose I write 4096 bytes to a new tmpfs file, so tn_pages == 1 and the page is dirty. Suppose…
		kibAuthorUnsubmitted Done Inline Actions So it must be handled in tmpfs_pager_freespace(), thank you for noting this. kib: So it must be handled in tmpfs_pager_freespace(), thank you for noting this.
		markjUnsubmitted Done Inline Actions I'm still confused about the scenario where we have a resident page and a copy exists in the pager. Should we count it once or twice? For tmpfs, it makes more sense to count it twice I think. But then the vm_pager_has_page() checks in tmpfs_page_inserted() and tmpfs_page_removed() should be removed, and tmpfs needs to intercept pageout ops and increment the count if a new swap block is to be allocated. And possibly putpages should return an error if the mount size limit would be exceeded. If we count such a page just once, however, then tmpfs_pager_freespace() should only decrement the count if there is no resident copy of the page. Otherwise: Start with sparse file of size PAGE_SIZE (tn_pages == 0) Write 4KB to the file (tn_pages == 1) The resident page is written to the pager (tn_pages == 1) The page is dirtied again, so freespace is called at some point (tn_pages == 0) The file has a resident page but tn_pages == 0 markj: I'm still confused about the scenario where we have a resident page and a copy exists in the…
		kibAuthorUnsubmitted Done Inline Actions I believe that we should only count the page once in tn_pages, if it exists both in swap and in queue. tn_pages counts all page-sized ranges in the tmpfs node that are not holes. I agree that there is a bug in swap_pager_freespace() then, should be fixed. kib: I believe that we should only count the page once in tn_pages, if it exists both in swap and in…
		}

		static boolean_t
		tmpfs_can_alloc_page(vm_object_t obj, vm_pindex_t pindex)
		{
		struct tmpfs_mount *tm;

		tm = VM_TO_TMPFS_MP(obj);
		if (tm == NULL \|\| vm_pager_has_page(obj, pindex, NULL, NULL) \|\|
		tm->tm_pages_max == 0)
		return (true);
		return (tm->tm_pages_max > atomic_load_long(&tm->tm_pages_used));
		}

struct pagerops tmpfs_pager_ops = {		struct pagerops tmpfs_pager_ops = {
.pgo_kvme_type = KVME_TYPE_VNODE,		.pgo_kvme_type = KVME_TYPE_VNODE,
.pgo_alloc = tmpfs_pager_alloc,		.pgo_alloc = tmpfs_pager_alloc,
.pgo_set_writeable_dirty = vm_object_set_writeable_dirty_,		.pgo_set_writeable_dirty = vm_object_set_writeable_dirty_,
.pgo_update_writecount = tmpfs_pager_update_writecount,		.pgo_update_writecount = tmpfs_pager_update_writecount,
.pgo_release_writecount = tmpfs_pager_release_writecount,		.pgo_release_writecount = tmpfs_pager_release_writecount,
.pgo_mightbedirty = vm_object_mightbedirty_,		.pgo_mightbedirty = vm_object_mightbedirty_,
.pgo_getvp = tmpfs_pager_getvp,		.pgo_getvp = tmpfs_pager_getvp,
		.pgo_page_inserted = tmpfs_page_inserted,
		.pgo_page_removed = tmpfs_page_removed,
		.pgo_can_alloc_page = tmpfs_can_alloc_page,
};		};

static int		static int
tmpfs_node_ctor(void mem, int size, void arg, int flags)		tmpfs_node_ctor(void mem, int size, void arg, int flags)
{		{
struct tmpfs_node *node;		struct tmpfs_node *node;

node = mem;		node = mem;
▲ Show 20 Lines • Show All 333 Lines • ▼ Show 20 Lines	case VLNK:
atomic_store_ptr(&nnode->tn_link_target, symlink);		atomic_store_ptr(&nnode->tn_link_target, symlink);
atomic_store_char((char *)&nnode->tn_link_smr, symlink_smr);		atomic_store_char((char *)&nnode->tn_link_smr, symlink_smr);
atomic_thread_fence_rel();		atomic_thread_fence_rel();
break;		break;

case VREG:		case VREG:
nnode->tn_reg.tn_aobj =		nnode->tn_reg.tn_aobj =
vm_pager_allocate(tmpfs_pager_type, NULL, 0,		vm_pager_allocate(tmpfs_pager_type, NULL, 0,
VM_PROT_DEFAULT, 0,		VM_PROT_DEFAULT, 0,
NULL /* XXXKIB - tmpfs needs swap reservation */);		NULL /* XXXKIB - tmpfs needs swap reservation */);
/* OBJ_TMPFS is set together with the setting of vp->v_object */		nnode->tn_reg.tn_aobj->un_pager.swp.swp_priv = nnode;
		vm_object_set_flag(nnode->tn_reg.tn_aobj, OBJ_TMPFS);
nnode->tn_reg.tn_tmp = tmp;		nnode->tn_reg.tn_tmp = tmp;
break;		break;

default:		default:
panic("tmpfs_alloc_node: type %p %d", nnode,		panic("tmpfs_alloc_node: type %p %d", nnode,
(int)nnode->tn_type);		(int)nnode->tn_type);
}		}

▲ Show 20 Lines • Show All 74 Lines • ▼ Show 20 Lines	case VMARKER:
panic("%s: bad type %d for node %p", __func__,		panic("%s: bad type %d for node %p", __func__,
(int)node->tn_type, node);		(int)node->tn_type, node);
}		}
#endif		#endif

switch (node->tn_type) {		switch (node->tn_type) {
case VREG:		case VREG:
uobj = node->tn_reg.tn_aobj;		uobj = node->tn_reg.tn_aobj;
if (uobj != NULL && uobj->size != 0)		node->tn_reg.tn_aobj = NULL;
atomic_subtract_long(&tmp->tm_pages_used, uobj->size);

tmpfs_free_tmp(tmp);

if (uobj != NULL) {		if (uobj != NULL) {
KASSERT((uobj->flags & OBJ_TMPFS) == 0,		VM_OBJECT_WLOCK(uobj);
("leaked OBJ_TMPFS node %p vm_obj %p", node, uobj));		vm_object_clear_flag(uobj, OBJ_TMPFS);
		atomic_add_long(&tmp->tm_pages_used,
		node->tn_reg.tn_pages);
		markjUnsubmitted Done Inline Actions Shouldn't this be a subtraction? markj: Shouldn't this be a subtraction?
		kibAuthorUnsubmitted Done Inline Actions Of course, this was non-tested late minute change to switch to tn_pages. kib: Of course, this was non-tested late minute change to switch to tn_pages.
		VM_OBJECT_WUNLOCK(uobj);
vm_object_deallocate(uobj);		vm_object_deallocate(uobj);
}		}
		tmpfs_free_tmp(tmp);
break;		break;
case VLNK:		case VLNK:
tmpfs_free_tmp(tmp);		tmpfs_free_tmp(tmp);

symlink = node->tn_link_target;		symlink = node->tn_link_target;
atomic_store_ptr(&node->tn_link_target, NULL);		atomic_store_ptr(&node->tn_link_target, NULL);
if (atomic_load_char(&node->tn_link_smr)) {		if (atomic_load_char(&node->tn_link_smr)) {
cache_symlink_free(symlink, node->tn_size + 1);		cache_symlink_free(symlink, node->tn_size + 1);
▲ Show 20 Lines • Show All 125 Lines • ▼ Show 20 Lines	tmpfs_destroy_vobject(struct vnode *vp, vm_object_t obj)
* May be going through forced unmount.		* May be going through forced unmount.
*/		*/
want_vrele = false;		want_vrele = false;
if ((obj->flags & OBJ_TMPFS_VREF) != 0) {		if ((obj->flags & OBJ_TMPFS_VREF) != 0) {
vm_object_clear_flag(obj, OBJ_TMPFS_VREF);		vm_object_clear_flag(obj, OBJ_TMPFS_VREF);
want_vrele = true;		want_vrele = true;
}		}

vm_object_clear_flag(obj, OBJ_TMPFS);
obj->un_pager.swp.swp_tmpfs = NULL;
if (vp->v_writecount < 0)		if (vp->v_writecount < 0)
vp->v_writecount = 0;		vp->v_writecount = 0;
VI_UNLOCK(vp);		VI_UNLOCK(vp);
VM_OBJECT_WUNLOCK(obj);		VM_OBJECT_WUNLOCK(obj);
if (want_vrele) {		if (want_vrele) {
vrele(vp);		vrele(vp);
}		}
}		}
▲ Show 20 Lines • Show All 121 Lines • ▼ Show 20 Lines	KASSERT((object->flags & OBJ_TMPFS_VREF) == 0,
("%s: object %p with OBJ_TMPFS_VREF but without vnode",		("%s: object %p with OBJ_TMPFS_VREF but without vnode",
__func__, object));		__func__, object));
KASSERT(object->un_pager.swp.writemappings == 0,		KASSERT(object->un_pager.swp.writemappings == 0,
("%s: object %p has writemappings",		("%s: object %p has writemappings",
__func__, object));		__func__, object));
VI_LOCK(vp);		VI_LOCK(vp);
KASSERT(vp->v_object == NULL, ("Not NULL v_object in tmpfs"));		KASSERT(vp->v_object == NULL, ("Not NULL v_object in tmpfs"));
vp->v_object = object;		vp->v_object = object;
object->un_pager.swp.swp_tmpfs = vp;
vm_object_set_flag(object, OBJ_TMPFS);
vn_irflag_set_locked(vp, VIRF_PGREAD \| VIRF_TEXT_REF);		vn_irflag_set_locked(vp, VIRF_PGREAD \| VIRF_TEXT_REF);
VI_UNLOCK(vp);		VI_UNLOCK(vp);
VM_OBJECT_WUNLOCK(object);		VM_OBJECT_WUNLOCK(object);
break;		break;
case VDIR:		case VDIR:
MPASS(node->tn_dir.tn_parent != NULL);		MPASS(node->tn_dir.tn_parent != NULL);
if (node->tn_dir.tn_parent == node)		if (node->tn_dir.tn_parent == node)
vp->v_vflag \|= VV_ROOT;		vp->v_vflag \|= VV_ROOT;
▲ Show 20 Lines • Show All 733 Lines • ▼ Show 20 Lines
* size 'newsize'. 'vp' must point to a vnode that represents a regular file.		* size 'newsize'. 'vp' must point to a vnode that represents a regular file.
* 'newsize' must be positive.		* 'newsize' must be positive.
*		*
* Returns zero on success or an appropriate error code on failure.		* Returns zero on success or an appropriate error code on failure.
*/		*/
int		int
tmpfs_reg_resize(struct vnode *vp, off_t newsize, boolean_t ignerr)		tmpfs_reg_resize(struct vnode *vp, off_t newsize, boolean_t ignerr)
{		{
struct tmpfs_mount *tmp;
struct tmpfs_node *node;		struct tmpfs_node *node;
vm_object_t uobj;		vm_object_t uobj;
vm_pindex_t idx, newpages, oldpages;		vm_pindex_t idx, newpages, oldpages;
off_t oldsize;		off_t oldsize;
int base, error;		int base, error;

MPASS(vp->v_type == VREG);		MPASS(vp->v_type == VREG);
MPASS(newsize >= 0);		MPASS(newsize >= 0);

node = VP_TO_TMPFS_NODE(vp);		node = VP_TO_TMPFS_NODE(vp);
uobj = node->tn_reg.tn_aobj;		uobj = node->tn_reg.tn_aobj;
tmp = VFS_TO_TMPFS(vp->v_mount);

/*		/*
* Convert the old and new sizes to the number of pages needed to		* Convert the old and new sizes to the number of pages needed to
* store them. It may happen that we do not need to do anything		* store them. It may happen that we do not need to do anything
* because the last allocated page can accommodate the change on		* because the last allocated page can accommodate the change on
* its own.		* its own.
*/		*/
oldsize = node->tn_size;		oldsize = node->tn_size;
oldpages = OFF_TO_IDX(oldsize + PAGE_MASK);		oldpages = OFF_TO_IDX(oldsize + PAGE_MASK);
MPASS(oldpages == uobj->size);		MPASS(oldpages == uobj->size);
newpages = OFF_TO_IDX(newsize + PAGE_MASK);		newpages = OFF_TO_IDX(newsize + PAGE_MASK);

if (__predict_true(newpages == oldpages && newsize >= oldsize)) {		if (__predict_true(newpages == oldpages && newsize >= oldsize)) {
node->tn_size = newsize;		node->tn_size = newsize;
return (0);		return (0);
}		}

if (newpages > oldpages &&
!tmpfs_pages_check_avail(tmp, newpages - oldpages))
return (ENOSPC);

VM_OBJECT_WLOCK(uobj);		VM_OBJECT_WLOCK(uobj);
if (newsize < oldsize) {		if (newsize < oldsize) {
/*		/*
* Zero the truncated part of the last page.		* Zero the truncated part of the last page.
*/		*/
base = newsize & PAGE_MASK;		base = newsize & PAGE_MASK;
if (base != 0) {		if (base != 0) {
idx = OFF_TO_IDX(newsize);		idx = OFF_TO_IDX(newsize);
error = tmpfs_partial_page_invalidate(uobj, idx, base,		error = tmpfs_partial_page_invalidate(uobj, idx, base,
PAGE_SIZE, ignerr);		PAGE_SIZE, ignerr);
if (error != 0) {		if (error != 0) {
VM_OBJECT_WUNLOCK(uobj);		VM_OBJECT_WUNLOCK(uobj);
return (error);		return (error);
}		}
}		}

/*		/*
* Release any swap space and free any whole pages.		* Release any swap space and free any whole pages.
*/		*/
if (newpages < oldpages)		if (newpages < oldpages)
vm_object_page_remove(uobj, newpages, 0, 0);		vm_object_page_remove(uobj, newpages, 0, 0);
}		}
uobj->size = newpages;		uobj->size = newpages;
VM_OBJECT_WUNLOCK(uobj);		VM_OBJECT_WUNLOCK(uobj);

atomic_add_long(&tmp->tm_pages_used, newpages - oldpages);

node->tn_size = newsize;		node->tn_size = newsize;
return (0);		return (0);
}		}

/*		/*
* Punch hole in the aobj associated with the regular file pointed to by 'vp'.		* Punch hole in the aobj associated with the regular file pointed to by 'vp'.
* Requests completely beyond the end-of-file are converted to no-op.		* Requests completely beyond the end-of-file are converted to no-op.
▲ Show 20 Lines • Show All 501 Lines • Show Last 20 Lines