struct vm_struct *xen_alloc_vm_area(unsigned long size)
{
int order;
unsigned long virt;
unsigned long nr_pages;
struct vm_struct *area;
order = get_order(size);
virt = __get_free_pages(GFP_KERNEL, order);
if (virt == 0)
goto err0;
nr_pages = 1 << order;
scrub_pages(virt, nr_pages);
area = kmalloc(sizeof(*area), GFP_KERNEL);
if (area == NULL)
goto err1;
area->flags = VM_IOREMAP;
area->addr = (void *)virt;
area->size = size;
area->pages = NULL;
area->nr_pages = nr_pages;
area->phys_addr = 0; /* xenbus_map_ring_valloc uses this field! */
return area;
err1:
free_pages(virt, order);
err0:
return NULL;
}
static void skbuff_ctor(void *buf, kmem_cache_t *cachep, unsigned long unused)
{
int order = 0;
while (skbuff_order_cachep[order] != cachep)
order++;
/* Do our best to allocate contiguous memory but fall back to IOMMU. */
if (order != 0)
(void)xen_create_contiguous_region(
(unsigned long)buf, order, 0);
scrub_pages(buf, 1 << order);
}
static int dealloc_pte_fn(
pte_t *pte, struct page *pmd_page, unsigned long addr, void *data)
{
unsigned long mfn = pte_mfn(*pte);
int ret;
struct xen_memory_reservation reservation = {
.nr_extents = 1,
.extent_order = 0,
.domid = DOMID_SELF
};
set_xen_guest_handle(reservation.extent_start, &mfn);
set_pte_at(&init_mm, addr, pte, __pte_ma(0));
set_phys_to_machine(__pa(addr) >> PAGE_SHIFT, INVALID_P2M_ENTRY);
ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation);
BUG_ON(ret != 1);
return 0;
}
#endif
struct page **alloc_empty_pages_and_pagevec(int nr_pages)
{
unsigned long vaddr, flags;
struct page *page, **pagevec;
int i, ret;
pagevec = kmalloc(sizeof(page) * nr_pages, GFP_KERNEL);
if (pagevec == NULL)
return NULL;
for (i = 0; i < nr_pages; i++) {
page = pagevec[i] = alloc_page(GFP_KERNEL);
if (page == NULL)
goto err;
vaddr = (unsigned long)page_address(page);
scrub_pages(vaddr, 1);
balloon_lock(flags);
if (xen_feature(XENFEAT_auto_translated_physmap)) {
unsigned long gmfn = page_to_pfn(page);
struct xen_memory_reservation reservation = {
.nr_extents = 1,
.extent_order = 0,
.domid = DOMID_SELF
};
set_xen_guest_handle(reservation.extent_start, &gmfn);
ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation,
&reservation);
if (ret == 1)
ret = 0; /* success */
} else {
#ifdef CONFIG_XEN
ret = apply_to_page_range(&init_mm, vaddr, PAGE_SIZE,
dealloc_pte_fn, NULL);
#else
/* Cannot handle non-auto translate mode. */
ret = 1;
#endif
}
if (ret != 0) {
balloon_unlock(flags);
__free_page(page);
goto err;
}
totalram_pages = --current_pages;
balloon_unlock(flags);
}
out:
schedule_work(&balloon_worker);
#ifdef CONFIG_XEN
flush_tlb_all();
#endif
return pagevec;
err:
balloon_lock(flags);
while (--i >= 0)
balloon_append(pagevec[i]);
balloon_unlock(flags);
kfree(pagevec);
pagevec = NULL;
goto out;
}
static int increase_reservation(unsigned long nr_pages)
{
unsigned long pfn, i, flags;
struct page *page;
long rc;
struct xen_memory_reservation reservation = {
.address_bits = 0,
.extent_order = 0,
.domid = DOMID_SELF
};
if (nr_pages > ARRAY_SIZE(frame_list))
nr_pages = ARRAY_SIZE(frame_list);
balloon_lock(flags);
page = balloon_first_page();
for (i = 0; i < nr_pages; i++) {
BUG_ON(page == NULL);
frame_list[i] = page_to_pfn(page);;
page = balloon_next_page(page);
}
set_xen_guest_handle(reservation.extent_start, frame_list);
reservation.nr_extents = nr_pages;
rc = HYPERVISOR_memory_op(
XENMEM_populate_physmap, &reservation);
if (rc < nr_pages) {
int ret;
/* We hit the Xen hard limit: reprobe. */
set_xen_guest_handle(reservation.extent_start, frame_list);
reservation.nr_extents = rc;
ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation,
&reservation);
BUG_ON(ret != rc);
hard_limit = current_pages + rc - driver_pages;
goto out;
}
for (i = 0; i < nr_pages; i++) {
page = balloon_retrieve();
BUG_ON(page == NULL);
pfn = page_to_pfn(page);
BUG_ON(!xen_feature(XENFEAT_auto_translated_physmap) &&
phys_to_machine_mapping_valid(pfn));
/* Update P->M and M->P tables. */
set_phys_to_machine(pfn, frame_list[i]);
#ifdef CONFIG_XEN
xen_machphys_update(frame_list[i], pfn);
/* Link back into the page tables if not highmem. */
if (pfn < max_low_pfn) {
int ret;
ret = HYPERVISOR_update_va_mapping(
(unsigned long)__va(pfn << PAGE_SHIFT),
pfn_pte_ma(frame_list[i], PAGE_KERNEL),
0);
BUG_ON(ret);
}
#endif
/* Relinquish the page back to the allocator. */
ClearPageReserved(page);
set_page_count(page, 1);
__free_page(page);
}
current_pages += nr_pages;
totalram_pages = current_pages;
out:
balloon_unlock(flags);
return 0;
}
static int decrease_reservation(unsigned long nr_pages)
{
unsigned long pfn, i, flags;
struct page *page;
void *v;
int need_sleep = 0;
int ret;
struct xen_memory_reservation reservation = {
.address_bits = 0,
.extent_order = 0,
.domid = DOMID_SELF
};
if (nr_pages > ARRAY_SIZE(frame_list))
nr_pages = ARRAY_SIZE(frame_list);
for (i = 0; i < nr_pages; i++) {
if ((page = alloc_page(GFP_BALLOON)) == NULL) {
nr_pages = i;
need_sleep = 1;
break;
}
pfn = page_to_pfn(page);
frame_list[i] = pfn_to_mfn(pfn);
if (!PageHighMem(page)) {
v = phys_to_virt(pfn << PAGE_SHIFT);
scrub_pages(v, 1);
#ifdef CONFIG_XEN
ret = HYPERVISOR_update_va_mapping(
(unsigned long)v, __pte_ma(0), 0);
BUG_ON(ret);
#endif
}
#ifdef CONFIG_XEN_SCRUB_PAGES
else {
v = kmap(page);
scrub_pages(v, 1);
kunmap(page);
}
#endif
}
#ifdef CONFIG_XEN
/* Ensure that ballooned highmem pages don't have kmaps. */
kmap_flush_unused();
flush_tlb_all();
#endif
balloon_lock(flags);
/* No more mappings: invalidate P2M and add to balloon. */
for (i = 0; i < nr_pages; i++) {
pfn = mfn_to_pfn(frame_list[i]);
set_phys_to_machine(pfn, INVALID_P2M_ENTRY);
balloon_append(pfn_to_page(pfn));
}
set_xen_guest_handle(reservation.extent_start, frame_list);
reservation.nr_extents = nr_pages;
ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation);
BUG_ON(ret != nr_pages);
current_pages -= nr_pages;
totalram_pages = current_pages;
balloon_unlock(flags);
return need_sleep;
}
/*
* We avoid multiple worker processes conflicting via the balloon mutex.
* We may of course race updates of the target counts (which are protected
* by the balloon lock), or with changes to the Xen hard limit, but we will
* recover from these in time.
*/
static void balloon_process(void *unused)
{
int need_sleep = 0;
long credit;
down(&balloon_mutex);
do {
credit = current_target() - current_pages;
if (credit > 0)
need_sleep = (increase_reservation(credit) != 0);
if (credit < 0)
need_sleep = (decrease_reservation(-credit) != 0);
#ifndef CONFIG_PREEMPT
if (need_resched())
schedule();
#endif
} while ((credit != 0) && !need_sleep);
/* Schedule more work if there is some still to be done. */
if (current_target() != current_pages)
mod_timer(&balloon_timer, jiffies + HZ);
up(&balloon_mutex);
}
