void
vnode_pager_undirty_pages(vm_page_t *ma, int *rtvals, int written)
{
vm_object_t obj;
int i, pos;
if (written == 0)
return;
obj = ma[0]->object;
VM_OBJECT_WLOCK(obj);
for (i = 0, pos = 0; pos < written; i++, pos += PAGE_SIZE) {
if (pos < trunc_page(written)) {
rtvals[i] = VM_PAGER_OK;
vm_page_undirty(ma[i]);
} else {
/* Partially written page. */
rtvals[i] = VM_PAGER_AGAIN;
vm_page_clear_dirty(ma[i], 0, written & PAGE_MASK);
}
}
VM_OBJECT_WUNLOCK(obj);
}
/*
* Lets the VM system know about a change in size for a file.
* We adjust our own internal size and flush any cached pages in
* the associated object that are affected by the size change.
*
* Note: this routine may be invoked as a result of a pager put
* operation (possibly at object termination time), so we must be careful.
*/
void
vnode_pager_setsize(struct vnode *vp, vm_ooffset_t nsize)
{
vm_object_t object;
vm_page_t m;
vm_pindex_t nobjsize;
if ((object = vp->v_object) == NULL)
return;
/* ASSERT_VOP_ELOCKED(vp, "vnode_pager_setsize and not locked vnode"); */
VM_OBJECT_WLOCK(object);
if (object->type == OBJT_DEAD) {
VM_OBJECT_WUNLOCK(object);
return;
}
KASSERT(object->type == OBJT_VNODE,
("not vnode-backed object %p", object));
if (nsize == object->un_pager.vnp.vnp_size) {
/*
* Hasn't changed size
*/
VM_OBJECT_WUNLOCK(object);
return;
}
nobjsize = OFF_TO_IDX(nsize + PAGE_MASK);
if (nsize < object->un_pager.vnp.vnp_size) {
/*
* File has shrunk. Toss any cached pages beyond the new EOF.
*/
if (nobjsize < object->size)
vm_object_page_remove(object, nobjsize, object->size,
0);
/*
* this gets rid of garbage at the end of a page that is now
* only partially backed by the vnode.
*
* XXX for some reason (I don't know yet), if we take a
* completely invalid page and mark it partially valid
* it can screw up NFS reads, so we don't allow the case.
*/
if ((nsize & PAGE_MASK) &&
(m = vm_page_lookup(object, OFF_TO_IDX(nsize))) != NULL &&
m->valid != 0) {
int base = (int)nsize & PAGE_MASK;
int size = PAGE_SIZE - base;
/*
* Clear out partial-page garbage in case
* the page has been mapped.
*/
pmap_zero_page_area(m, base, size);
/*
* Update the valid bits to reflect the blocks that
* have been zeroed. Some of these valid bits may
* have already been set.
*/
vm_page_set_valid_range(m, base, size);
/*
* Round "base" to the next block boundary so that the
* dirty bit for a partially zeroed block is not
* cleared.
*/
base = roundup2(base, DEV_BSIZE);
/*
* Clear out partial-page dirty bits.
*
* note that we do not clear out the valid
* bits. This would prevent bogus_page
* replacement from working properly.
*/
vm_page_clear_dirty(m, base, PAGE_SIZE - base);
}
}
object->un_pager.vnp.vnp_size = nsize;
object->size = nobjsize;
VM_OBJECT_WUNLOCK(object);
}
static int
shm_dotruncate(struct shmfd *shmfd, off_t length)
{
vm_object_t object;
vm_page_t m;
vm_pindex_t nobjsize;
vm_ooffset_t delta;
object = shmfd->shm_object;
VM_OBJECT_LOCK(object);
if (length == shmfd->shm_size) {
VM_OBJECT_UNLOCK(object);
return (0);
}
nobjsize = OFF_TO_IDX(length + PAGE_MASK);
/* Are we shrinking? If so, trim the end. */
if (length < shmfd->shm_size) {
delta = ptoa(object->size - nobjsize);
/* Toss in memory pages. */
if (nobjsize < object->size)
vm_object_page_remove(object, nobjsize, object->size,
FALSE);
/* Toss pages from swap. */
if (object->type == OBJT_SWAP)
swap_pager_freespace(object, nobjsize, delta);
/* Free the swap accounted for shm */
swap_release_by_uid(delta, object->uip);
object->charge -= delta;
/*
* If the last page is partially mapped, then zero out
* the garbage at the end of the page. See comments
* in vnode_pager_setsize() for more details.
*
* XXXJHB: This handles in memory pages, but what about
* a page swapped out to disk?
*/
if ((length & PAGE_MASK) &&
(m = vm_page_lookup(object, OFF_TO_IDX(length))) != NULL &&
m->valid != 0) {
int base = (int)length & PAGE_MASK;
int size = PAGE_SIZE - base;
pmap_zero_page_area(m, base, size);
/*
* Update the valid bits to reflect the blocks that
* have been zeroed. Some of these valid bits may
* have already been set.
*/
vm_page_set_valid(m, base, size);
/*
* Round "base" to the next block boundary so that the
* dirty bit for a partially zeroed block is not
* cleared.
*/
base = roundup2(base, DEV_BSIZE);
vm_page_lock_queues();
vm_page_clear_dirty(m, base, PAGE_SIZE - base);
vm_page_unlock_queues();
} else if ((length & PAGE_MASK) &&
__predict_false(object->cache != NULL)) {
vm_page_cache_free(object, OFF_TO_IDX(length),
nobjsize);
}
} else {
/* Attempt to reserve the swap */
delta = ptoa(nobjsize - object->size);
if (!swap_reserve_by_uid(delta, object->uip)) {
VM_OBJECT_UNLOCK(object);
return (ENOMEM);
}
object->charge += delta;
}
shmfd->shm_size = length;
mtx_lock(&shm_timestamp_lock);
vfs_timestamp(&shmfd->shm_ctime);
shmfd->shm_mtime = shmfd->shm_ctime;
mtx_unlock(&shm_timestamp_lock);
object->size = nobjsize;
VM_OBJECT_UNLOCK(object);
return (0);
}
/*
* This is now called from local media FS's to operate against their
* own vnodes if they fail to implement VOP_PUTPAGES.
*
* This is typically called indirectly via the pageout daemon and
* clustering has already typically occurred, so in general we ask the
* underlying filesystem to write the data out asynchronously rather
* then delayed.
*/
int
vnode_pager_generic_putpages(struct vnode *vp, vm_page_t *ma, int bytecount,
int flags, int *rtvals)
{
int i;
vm_object_t object;
vm_page_t m;
int count;
int maxsize, ncount;
vm_ooffset_t poffset;
struct uio auio;
struct iovec aiov;
int error;
int ioflags;
int ppscheck = 0;
static struct timeval lastfail;
static int curfail;
object = vp->v_object;
count = bytecount / PAGE_SIZE;
for (i = 0; i < count; i++)
rtvals[i] = VM_PAGER_ERROR;
if ((int64_t)ma[0]->pindex < 0) {
printf("vnode_pager_putpages: attempt to write meta-data!!! -- 0x%lx(%lx)\n",
(long)ma[0]->pindex, (u_long)ma[0]->dirty);
rtvals[0] = VM_PAGER_BAD;
return VM_PAGER_BAD;
}
maxsize = count * PAGE_SIZE;
ncount = count;
poffset = IDX_TO_OFF(ma[0]->pindex);
/*
* If the page-aligned write is larger then the actual file we
* have to invalidate pages occurring beyond the file EOF. However,
* there is an edge case where a file may not be page-aligned where
* the last page is partially invalid. In this case the filesystem
* may not properly clear the dirty bits for the entire page (which
* could be VM_PAGE_BITS_ALL due to the page having been mmap()d).
* With the page locked we are free to fix-up the dirty bits here.
*
* We do not under any circumstances truncate the valid bits, as
* this will screw up bogus page replacement.
*/
VM_OBJECT_WLOCK(object);
if (maxsize + poffset > object->un_pager.vnp.vnp_size) {
if (object->un_pager.vnp.vnp_size > poffset) {
int pgoff;
maxsize = object->un_pager.vnp.vnp_size - poffset;
ncount = btoc(maxsize);
if ((pgoff = (int)maxsize & PAGE_MASK) != 0) {
/*
* If the object is locked and the following
* conditions hold, then the page's dirty
* field cannot be concurrently changed by a
* pmap operation.
*/
m = ma[ncount - 1];
vm_page_assert_sbusied(m);
KASSERT(!pmap_page_is_write_mapped(m),
("vnode_pager_generic_putpages: page %p is not read-only", m));
vm_page_clear_dirty(m, pgoff, PAGE_SIZE -
pgoff);
}
} else {
maxsize = 0;
ncount = 0;
}
if (ncount < count) {
for (i = ncount; i < count; i++) {
rtvals[i] = VM_PAGER_BAD;
}
}
}
VM_OBJECT_WUNLOCK(object);
/*
* pageouts are already clustered, use IO_ASYNC to force a bawrite()
* rather then a bdwrite() to prevent paging I/O from saturating
* the buffer cache. Dummy-up the sequential heuristic to cause
* large ranges to cluster. If neither IO_SYNC or IO_ASYNC is set,
* the system decides how to cluster.
*/
ioflags = IO_VMIO;
if (flags & (VM_PAGER_PUT_SYNC | VM_PAGER_PUT_INVAL))
ioflags |= IO_SYNC;
else if ((flags & VM_PAGER_CLUSTER_OK) == 0)
ioflags |= IO_ASYNC;
ioflags |= (flags & VM_PAGER_PUT_INVAL) ? IO_INVAL: 0;
ioflags |= (flags & VM_PAGER_PUT_NOREUSE) ? IO_NOREUSE : 0;
ioflags |= IO_SEQMAX << IO_SEQSHIFT;
aiov.iov_base = (caddr_t) 0;
aiov.iov_len = maxsize;
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_offset = poffset;
auio.uio_segflg = UIO_NOCOPY;
auio.uio_rw = UIO_WRITE;
auio.uio_resid = maxsize;
auio.uio_td = (struct thread *) 0;
error = VOP_WRITE(vp, &auio, ioflags, curthread->td_ucred);
PCPU_INC(cnt.v_vnodeout);
PCPU_ADD(cnt.v_vnodepgsout, ncount);
if (error) {
if ((ppscheck = ppsratecheck(&lastfail, &curfail, 1)))
printf("vnode_pager_putpages: I/O error %d\n", error);
}
if (auio.uio_resid) {
if (ppscheck || ppsratecheck(&lastfail, &curfail, 1))
printf("vnode_pager_putpages: residual I/O %zd at %lu\n",
auio.uio_resid, (u_long)ma[0]->pindex);
}
for (i = 0; i < ncount; i++) {
rtvals[i] = VM_PAGER_OK;
}
return rtvals[0];
}
/*
* This is now called from local media FS's to operate against their
* own vnodes if they fail to implement VOP_PUTPAGES.
*
* This is typically called indirectly via the pageout daemon and
* clustering has already typically occurred, so in general we ask the
* underlying filesystem to write the data out asynchronously rather
* then delayed.
*/
int
vnode_pager_generic_putpages(struct vnode *vp, vm_page_t *ma, int bytecount,
int flags, int *rtvals)
{
vm_object_t object;
vm_page_t m;
vm_ooffset_t poffset;
struct uio auio;
struct iovec aiov;
int count, error, i, maxsize, ncount, pgoff, ppscheck;
static struct timeval lastfail;
static int curfail;
object = vp->v_object;
count = bytecount / PAGE_SIZE;
for (i = 0; i < count; i++)
rtvals[i] = VM_PAGER_ERROR;
if ((int64_t)ma[0]->pindex < 0) {
printf("vnode_pager_generic_putpages: "
"attempt to write meta-data 0x%jx(%lx)\n",
(uintmax_t)ma[0]->pindex, (u_long)ma[0]->dirty);
rtvals[0] = VM_PAGER_BAD;
return (VM_PAGER_BAD);
}
maxsize = count * PAGE_SIZE;
ncount = count;
poffset = IDX_TO_OFF(ma[0]->pindex);
/*
* If the page-aligned write is larger then the actual file we
* have to invalidate pages occurring beyond the file EOF. However,
* there is an edge case where a file may not be page-aligned where
* the last page is partially invalid. In this case the filesystem
* may not properly clear the dirty bits for the entire page (which
* could be VM_PAGE_BITS_ALL due to the page having been mmap()d).
* With the page locked we are free to fix-up the dirty bits here.
*
* We do not under any circumstances truncate the valid bits, as
* this will screw up bogus page replacement.
*/
VM_OBJECT_WLOCK(object);
if (maxsize + poffset > object->un_pager.vnp.vnp_size) {
if (object->un_pager.vnp.vnp_size > poffset) {
maxsize = object->un_pager.vnp.vnp_size - poffset;
ncount = btoc(maxsize);
if ((pgoff = (int)maxsize & PAGE_MASK) != 0) {
/*
* If the object is locked and the following
* conditions hold, then the page's dirty
* field cannot be concurrently changed by a
* pmap operation.
*/
m = ma[ncount - 1];
vm_page_assert_sbusied(m);
KASSERT(!pmap_page_is_write_mapped(m),
("vnode_pager_generic_putpages: page %p is not read-only", m));
MPASS(m->dirty != 0);
vm_page_clear_dirty(m, pgoff, PAGE_SIZE -
pgoff);
}
} else {
maxsize = 0;
ncount = 0;
}
for (i = ncount; i < count; i++)
rtvals[i] = VM_PAGER_BAD;
}
for (i = 0; i < ncount - ((btoc(maxsize) & PAGE_MASK) != 0); i++)
MPASS(ma[i]->dirty == VM_PAGE_BITS_ALL);
VM_OBJECT_WUNLOCK(object);
aiov.iov_base = NULL;
aiov.iov_len = maxsize;
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_offset = poffset;
auio.uio_segflg = UIO_NOCOPY;
auio.uio_rw = UIO_WRITE;
auio.uio_resid = maxsize;
auio.uio_td = NULL;
error = VOP_WRITE(vp, &auio, vnode_pager_putpages_ioflags(flags),
curthread->td_ucred);
VM_CNT_INC(v_vnodeout);
VM_CNT_ADD(v_vnodepgsout, ncount);
ppscheck = 0;
if (error != 0 && (ppscheck = ppsratecheck(&lastfail, &curfail, 1))
!= 0)
printf("vnode_pager_putpages: I/O error %d\n", error);
if (auio.uio_resid != 0 && (ppscheck != 0 ||
ppsratecheck(&lastfail, &curfail, 1) != 0))
printf("vnode_pager_putpages: residual I/O %zd at %ju\n",
auio.uio_resid, (uintmax_t)ma[0]->pindex);
for (i = 0; i < ncount; i++)
rtvals[i] = VM_PAGER_OK;
return (rtvals[0]);
}
