/*
* 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.
*
* NOTE: vp->v_filesize is initialized to NOOFFSET (-1), be sure that
* we do not blow up on the case. nsize will always be >= 0, however.
*/
void
vnode_pager_setsize(struct vnode *vp, vm_ooffset_t nsize)
{
vm_pindex_t nobjsize;
vm_pindex_t oobjsize;
vm_object_t object;
object = vp->v_object;
if (object == NULL)
return;
vm_object_hold(object);
KKASSERT(vp->v_object == object);
/*
* Hasn't changed size
*/
if (nsize == vp->v_filesize) {
vm_object_drop(object);
return;
}
/*
* Has changed size. Adjust the VM object's size and v_filesize
* before we start scanning pages to prevent new pages from being
* allocated during the scan.
*/
nobjsize = OFF_TO_IDX(nsize + PAGE_MASK);
oobjsize = object->size;
object->size = nobjsize;
/*
* File has shrunk. Toss any cached pages beyond the new EOF.
*/
if (nsize < vp->v_filesize) {
vp->v_filesize = nsize;
if (nobjsize < oobjsize) {
vm_object_page_remove(object, nobjsize, oobjsize,
FALSE);
}
/*
* This gets rid of garbage at the end of a page that is now
* only partially backed by the vnode. Since we are setting
* the entire page valid & clean after we are done we have
* to be sure that the portion of the page within the file
* bounds is already valid. If it isn't then making it
* valid would create a corrupt block.
*/
if (nsize & PAGE_MASK) {
vm_offset_t kva;
vm_page_t m;
m = vm_page_lookup_busy_wait(object, OFF_TO_IDX(nsize),
TRUE, "vsetsz");
if (m && m->valid) {
int base = (int)nsize & PAGE_MASK;
int size = PAGE_SIZE - base;
struct lwbuf *lwb;
struct lwbuf lwb_cache;
/*
* Clear out partial-page garbage in case
* the page has been mapped.
*
* This is byte aligned.
*/
lwb = lwbuf_alloc(m, &lwb_cache);
kva = lwbuf_kva(lwb);
bzero((caddr_t)kva + base, size);
lwbuf_free(lwb);
/*
* XXX work around SMP data integrity race
* by unmapping the page from user processes.
* The garbage we just cleared may be mapped
* to a user process running on another cpu
* and this code is not running through normal
* I/O channels which handle SMP issues for
* us, so unmap page to synchronize all cpus.
*
* XXX should vm_pager_unmap_page() have
* dealt with this?
*/
vm_page_protect(m, VM_PROT_NONE);
/*
* Clear out partial-page dirty bits. This
* has the side effect of setting the valid
* bits, but that is ok. There are a bunch
* of places in the VM system where we expected
* m->dirty == VM_PAGE_BITS_ALL. The file EOF
* case is one of them. If the page is still
* partially dirty, make it fully dirty.
*
* NOTE: We do not clear out the valid
* bits. This would prevent bogus_page
* replacement from working properly.
*
* NOTE: We do not want to clear the dirty
* bit for a partial DEV_BSIZE'd truncation!
* This is DEV_BSIZE aligned!
*/
vm_page_clear_dirty_beg_nonincl(m, base, size);
if (m->dirty != 0)
m->dirty = VM_PAGE_BITS_ALL;
vm_page_wakeup(m);
} else if (m) {
vm_page_wakeup(m);
}
}
} else {
vp->v_filesize = nsize;
}
vm_object_drop(object);
}
/*
* vm_contig_pg_alloc:
*
* Allocate contiguous pages from the VM. This function does not
* map the allocated pages into the kernel map, otherwise it is
* impossible to make large allocations (i.e. >2G).
*
* Malloc()'s data structures have been used for collection of
* statistics and for allocations of less than a page.
*/
static int
vm_contig_pg_alloc(unsigned long size, vm_paddr_t low, vm_paddr_t high,
unsigned long alignment, unsigned long boundary, int mflags)
{
int i, q, start, pass;
vm_offset_t phys;
vm_page_t pga = vm_page_array;
vm_page_t m;
int pqtype;
size = round_page(size);
if (size == 0)
panic("vm_contig_pg_alloc: size must not be 0");
if ((alignment & (alignment - 1)) != 0)
panic("vm_contig_pg_alloc: alignment must be a power of 2");
if ((boundary & (boundary - 1)) != 0)
panic("vm_contig_pg_alloc: boundary must be a power of 2");
/*
* See if we can get the pages from the contiguous page reserve
* alist. The returned pages will be allocated and wired but not
* busied.
*/
m = vm_page_alloc_contig(low, high, alignment, boundary, size);
if (m)
return (m - &pga[0]);
/*
* Three passes (0, 1, 2). Each pass scans the VM page list for
* free or cached pages. After each pass if the entire scan failed
* we attempt to flush inactive pages and reset the start index back
* to 0. For passes 1 and 2 we also attempt to flush active pages.
*/
start = 0;
for (pass = 0; pass < 3; pass++) {
/*
* Find first page in array that is free, within range,
* aligned, and such that the boundary won't be crossed.
*/
again:
for (i = start; i < vmstats.v_page_count; i++) {
m = &pga[i];
phys = VM_PAGE_TO_PHYS(m);
pqtype = m->queue - m->pc;
if (((pqtype == PQ_FREE) || (pqtype == PQ_CACHE)) &&
(phys >= low) && (phys < high) &&
((phys & (alignment - 1)) == 0) &&
(((phys ^ (phys + size - 1)) & ~(boundary - 1)) == 0) &&
m->busy == 0 && m->wire_count == 0 &&
m->hold_count == 0 &&
(m->flags & (PG_BUSY | PG_NEED_COMMIT)) == 0)
{
break;
}
}
/*
* If we cannot find the page in the given range, or we have
* crossed the boundary, call the vm_contig_pg_clean() function
* for flushing out the queues, and returning it back to
* normal state.
*/
if ((i == vmstats.v_page_count) ||
((VM_PAGE_TO_PHYS(&pga[i]) + size) > high)) {
/*
* Best effort flush of all inactive pages.
* This is quite quick, for now stall all
* callers, even if they've specified M_NOWAIT.
*/
for (q = 0; q < PQ_L2_SIZE; ++q) {
vm_contig_pg_clean(PQ_INACTIVE + q,
vmstats.v_inactive_count);
lwkt_yield();
}
/*
* Best effort flush of active pages.
*
* This is very, very slow.
* Only do this if the caller has agreed to M_WAITOK.
*
* If enough pages are flushed, we may succeed on
* next (final) pass, if not the caller, contigmalloc(),
* will fail in the index < 0 case.
*/
if (pass > 0 && (mflags & M_WAITOK)) {
for (q = 0; q < PQ_L2_SIZE; ++q) {
vm_contig_pg_clean(PQ_ACTIVE + q,
vmstats.v_active_count);
}
lwkt_yield();
}
/*
* We're already too high in the address space
* to succeed, reset to 0 for the next iteration.
*/
start = 0;
continue; /* next pass */
}
start = i;
/*
* Check successive pages for contiguous and free.
*
* (still in critical section)
*/
for (i = start + 1; i < (start + size / PAGE_SIZE); i++) {
m = &pga[i];
pqtype = m->queue - m->pc;
if ((VM_PAGE_TO_PHYS(&m[0]) !=
(VM_PAGE_TO_PHYS(&m[-1]) + PAGE_SIZE)) ||
((pqtype != PQ_FREE) && (pqtype != PQ_CACHE)) ||
m->busy || m->wire_count ||
m->hold_count ||
(m->flags & (PG_BUSY | PG_NEED_COMMIT)))
{
start++;
goto again;
}
}
/*
* Try to allocate the pages, wiring them as we go.
*
* (still in critical section)
*/
for (i = start; i < (start + size / PAGE_SIZE); i++) {
m = &pga[i];
if (vm_page_busy_try(m, TRUE)) {
vm_contig_pg_free(start,
(i - start) * PAGE_SIZE);
start++;
goto again;
}
pqtype = m->queue - m->pc;
if (pqtype == PQ_CACHE &&
m->hold_count == 0 &&
m->wire_count == 0 &&
(m->flags & (PG_UNMANAGED | PG_NEED_COMMIT)) == 0) {
vm_page_protect(m, VM_PROT_NONE);
KKASSERT((m->flags & PG_MAPPED) == 0);
KKASSERT(m->dirty == 0);
vm_page_free(m);
--i;
continue; /* retry the page */
}
if (pqtype != PQ_FREE || m->hold_count) {
vm_page_wakeup(m);
vm_contig_pg_free(start,
(i - start) * PAGE_SIZE);
start++;
goto again;
}
KKASSERT((m->valid & m->dirty) == 0);
KKASSERT(m->wire_count == 0);
KKASSERT(m->object == NULL);
vm_page_unqueue_nowakeup(m);
m->valid = VM_PAGE_BITS_ALL;
if (m->flags & PG_ZERO)
vm_page_zero_count--;
KASSERT(m->dirty == 0,
("vm_contig_pg_alloc: page %p was dirty", m));
KKASSERT(m->wire_count == 0);
KKASSERT(m->busy == 0);
/*
* Clear all flags except PG_BUSY, PG_ZERO, and
* PG_WANTED, then unbusy the now allocated page.
*/
vm_page_flag_clear(m, ~(PG_BUSY | PG_SBUSY |
PG_ZERO | PG_WANTED));
vm_page_wire(m);
vm_page_wakeup(m);
}
/*
* Our job is done, return the index page of vm_page_array.
*/
return (start); /* aka &pga[start] */
}
/*
* Failed.
*/
return (-1);
}
