void balloon_release_driver_page(struct page *page)
{
unsigned long flags;
balloon_lock(flags);
balloon_append(page);
driver_pages--;
balloon_unlock(flags);
schedule_work(&balloon_worker);
}
static int __init balloon_init(void)
{
#if defined(CONFIG_X86) && defined(CONFIG_XEN)
unsigned long pfn;
struct page *page;
#endif
if (!is_running_on_xen())
return -ENODEV;
IPRINTK("Initialising balloon driver.\n");
#ifdef CONFIG_XEN
current_pages = min(xen_start_info->nr_pages, max_pfn);
totalram_pages = current_pages;
#else
current_pages = totalram_pages;
#endif
target_pages = current_pages;
balloon_low = 0;
balloon_high = 0;
driver_pages = 0UL;
hard_limit = ~0UL;
init_timer(&balloon_timer);
balloon_timer.data = 0;
balloon_timer.function = balloon_alarm;
/* don't include on bare-metal & FV guest */
#if defined(CONFIG_PROC_FS) && defined(CONFIG_XEN)
if ((balloon_pde = create_xen_proc_entry("balloon", 0644)) == NULL) {
WPRINTK("Unable to create /proc/xen/balloon.\n");
return -1;
}
balloon_pde->read_proc = balloon_read;
balloon_pde->write_proc = balloon_write;
#endif
#if defined(CONFIG_X86) && defined(CONFIG_XEN)
/* Initialise the balloon with excess memory space. */
for (pfn = xen_start_info->nr_pages; pfn < max_pfn; pfn++) {
page = pfn_to_page(pfn);
balloon_append(page);
}
#endif
target_watch.callback = watch_target;
xenstore_notifier.notifier_call = balloon_init_watcher;
register_xenstore_notifier(&xenstore_notifier);
return 0;
}
static int __init balloon_init(void)
{
unsigned long pfn,num_physpages,max_pfn;
struct page *page;
if (!xen_domain())
return -ENODEV;
pr_info("xen_balloon: Initialising balloon driver.\n");
num_physpages = get_num_physpages();
if (xen_pv_domain())
max_pfn = xen_start_info->nr_pages;
else
max_pfn = num_physpages;
balloon_stats.current_pages = min(num_physpages,max_pfn);
totalram_bias = balloon_stats.current_pages - totalram_pages;
old_totalram_pages = totalram_pages;
balloon_stats.target_pages = balloon_stats.current_pages;
balloon_stats.balloon_low = 0;
balloon_stats.balloon_high = 0;
balloon_stats.driver_pages = 0UL;
pr_info("current_pages=%luKB, totalram_pages=%luKB, totalram_bias=%luKB\n",balloon_stats.current_pages*4, totalram_pages*4, totalram_bias*4);
init_timer(&balloon_timer);
balloon_timer.data = 0;
balloon_timer.function = balloon_alarm;
register_balloon(&balloon_sysdev);
/* Initialise the balloon with excess memory space. */
#ifdef CONFIG_PVM
for (pfn = xen_start_info->nr_pages; pfn < max_pfn; pfn++) {
page = pfn_to_page(pfn);
if (!PageReserved(page))
balloon_append(page);
}
#endif
target_watch.callback = watch_target;
xenstore_notifier.notifier_call = balloon_init_watcher;
register_xenstore_notifier(&xenstore_notifier);
return 0;
}
void free_empty_pages_and_pagevec(struct page **pagevec, int nr_pages)
{
unsigned long flags;
int i;
if (pagevec == NULL)
return;
balloon_lock(flags);
for (i = 0; i < nr_pages; i++) {
BUG_ON(page_count(pagevec[i]) != 1);
balloon_append(pagevec[i]);
}
balloon_unlock(flags);
kfree(pagevec);
schedule_work(&balloon_worker);
}
static int __init balloon_init(void)
{
unsigned long pfn;
struct page *page;
if (!xen_pv_domain())
return -ENODEV;
pr_info("xen_balloon: Initialising balloon driver.\n");
balloon_stats.current_pages = min(xen_start_info->nr_pages, max_pfn);
totalram_pages = balloon_stats.current_pages;
balloon_stats.target_pages = balloon_stats.current_pages;
balloon_stats.balloon_low = 0;
balloon_stats.balloon_high = 0;
balloon_stats.driver_pages = 0UL;
balloon_stats.hard_limit = ~0UL;
init_timer(&balloon_timer);
balloon_timer.data = 0;
balloon_timer.function = balloon_alarm;
register_balloon(&balloon_sysdev);
/* Initialise the balloon with excess memory space. */
for (pfn = xen_start_info->nr_pages; pfn < max_pfn; pfn++) {
page = pfn_to_page(pfn);
if (!PageReserved(page))
balloon_append(page);
}
target_watch.callback = watch_target;
xenstore_notifier.notifier_call = balloon_init_watcher;
register_xenstore_notifier(&xenstore_notifier);
return 0;
}
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);
}
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);
spin_lock_irqsave(&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 < 0)
goto out;
for (i = 0; i < rc; 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));
set_phys_to_machine(pfn, frame_list[i]);
/* Link back into the page tables if not highmem. */
#ifdef CONFIG_PVM
if (!xen_hvm_domain() && pfn < max_low_pfn) {
int ret;
ret = HYPERVISOR_update_va_mapping(
(unsigned long)__va(pfn << PAGE_SHIFT),
mfn_pte(frame_list[i], PAGE_KERNEL),
0);
BUG_ON(ret);
}
#endif
/* Relinquish the page back to the allocator. */
ClearPageReserved(page);
init_page_count(page);
__free_page(page);
}
balloon_stats.current_pages += rc;
if (old_totalram_pages + rc < totalram_pages)
{
printk(KERN_INFO "old_totalram=%luKB, totalram_pages=%luKB\n", old_totalram_pages*4, totalram_pages*4);
balloon_stats.current_pages = totalram_pages + totalram_bias;
printk(KERN_INFO "when ballooning, the mem online! totalram=%luKB, current=%luKB\n", totalram_pages*4, balloon_stats.current_pages*4);
}
old_totalram_pages = totalram_pages;
out:
spin_unlock_irqrestore(&balloon_lock, flags);
return rc < 0 ? rc : rc != nr_pages;
}
static int decrease_reservation(unsigned long nr_pages)
{
unsigned long pfn, i, flags;
struct page *page;
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);
scrub_page(page);
if (!xen_hvm_domain() && !PageHighMem(page)) {
ret = HYPERVISOR_update_va_mapping(
(unsigned long)__va(pfn << PAGE_SHIFT),
__pte_ma(0), 0);
BUG_ON(ret);
}
}
/* Ensure that ballooned highmem pages don't have kmaps. */
#ifdef CONFIG_PVM
kmap_flush_unused();
flush_tlb_all();
#endif
spin_lock_irqsave(&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);
balloon_stats.current_pages -= nr_pages;
if(old_totalram_pages < totalram_pages + nr_pages)
{
printk(KERN_INFO "old_totalram=%luKB, totalram_pages=%luKB\n", old_totalram_pages*4, totalram_pages*4);
balloon_stats.current_pages = totalram_pages + totalram_bias;
printk(KERN_INFO "when ballooning, the mem online! totalram=%luKB, current=%luKB\n", totalram_pages*4, balloon_stats.current_pages*4);
}
old_totalram_pages = totalram_pages;
spin_unlock_irqrestore(&balloon_lock, 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(struct work_struct *work)
{
int need_sleep = 0;
long credit;
long total_increase = 0;
char buffer[16];
mutex_lock(&balloon_mutex);
printk(KERN_INFO "totalram_pages=%luKB, current_pages=%luKB,totalram_bias=%luKB\n", totalram_pages*4, balloon_stats.current_pages*4, totalram_bias*4);
if (totalram_pages > old_totalram_pages)
{
//TODO:Just know that totalram_pages will increase.
total_increase = (totalram_pages - old_totalram_pages) % GB2PAGE;
if (totalram_bias > total_increase )
{
totalram_bias = totalram_bias - total_increase;
}
balloon_stats.current_pages = totalram_pages + totalram_bias;
old_totalram_pages = totalram_pages;
}
printk(KERN_INFO "totalram_pages=%luKB, current_pages=%luKB, totalram_bias=%luKB,total_increase=%ld\n", totalram_pages*4, balloon_stats.current_pages*4, totalram_bias*4, total_increase*4);
xenbus_write(XBT_NIL, "control/uvp", "Balloon_flag", "1");
do {
credit = current_target() - balloon_stats.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() != balloon_stats.current_pages)
{
mod_timer(&balloon_timer, jiffies + HZ);
sprintf(buffer,"%lu",balloon_stats.current_pages<<(PAGE_SHIFT-10));
xenbus_write(XBT_NIL, "memory", "target", buffer);
}
xenbus_write(XBT_NIL, "control/uvp", "Balloon_flag", "0");
mutex_unlock(&balloon_mutex);
}