[mirror_ubuntu-zesty-kernel.git] / arch / x86 / xen / p2m.c

/*
 * Xen leaves the responsibility for maintaining p2m mappings to the
 * guests themselves, but it must also access and update the p2m array
 * during suspend/resume when all the pages are reallocated.
 *
 * The p2m table is logically a flat array, but we implement it as a
 * three-level tree to allow the address space to be sparse.
 *
 *                               Xen
 *                                |
 *     p2m_top              p2m_top_mfn
 *       /  \                   /   \
 * p2m_mid p2m_mid	p2m_mid_mfn p2m_mid_mfn
 *    / \      / \         /           /
 *  p2m p2m p2m p2m p2m p2m p2m ...
 *
 * The p2m_mid_mfn pages are mapped by p2m_top_mfn_p.
 *
 * The p2m_top and p2m_top_mfn levels are limited to 1 page, so the
 * maximum representable pseudo-physical address space is:
 *  P2M_TOP_PER_PAGE * P2M_MID_PER_PAGE * P2M_PER_PAGE pages
 *
 * P2M_PER_PAGE depends on the architecture, as a mfn is always
 * unsigned long (8 bytes on 64-bit, 4 bytes on 32), leading to
 * 512 and 1024 entries respectively. 
 */

#include <linux/init.h>
#include <linux/module.h>
#include <linux/list.h>
#include <linux/hash.h>
#include <linux/sched.h>

#include <asm/cache.h>
#include <asm/setup.h>

#include <asm/xen/page.h>
#include <asm/xen/hypercall.h>
#include <asm/xen/hypervisor.h>

#include "xen-ops.h"

static void __init m2p_override_init(void);

unsigned long xen_max_p2m_pfn __read_mostly;

#define P2M_PER_PAGE		(PAGE_SIZE / sizeof(unsigned long))
#define P2M_MID_PER_PAGE	(PAGE_SIZE / sizeof(unsigned long *))
#define P2M_TOP_PER_PAGE	(PAGE_SIZE / sizeof(unsigned long **))

#define MAX_P2M_PFN		(P2M_TOP_PER_PAGE * P2M_MID_PER_PAGE * P2M_PER_PAGE)

/* Placeholders for holes in the address space */
static RESERVE_BRK_ARRAY(unsigned long, p2m_missing, P2M_PER_PAGE);
static RESERVE_BRK_ARRAY(unsigned long *, p2m_mid_missing, P2M_MID_PER_PAGE);
static RESERVE_BRK_ARRAY(unsigned long, p2m_mid_missing_mfn, P2M_MID_PER_PAGE);

static RESERVE_BRK_ARRAY(unsigned long **, p2m_top, P2M_TOP_PER_PAGE);
static RESERVE_BRK_ARRAY(unsigned long, p2m_top_mfn, P2M_TOP_PER_PAGE);
static RESERVE_BRK_ARRAY(unsigned long *, p2m_top_mfn_p, P2M_TOP_PER_PAGE);

RESERVE_BRK(p2m_mid, PAGE_SIZE * (MAX_DOMAIN_PAGES / (P2M_PER_PAGE * P2M_MID_PER_PAGE)));
RESERVE_BRK(p2m_mid_mfn, PAGE_SIZE * (MAX_DOMAIN_PAGES / (P2M_PER_PAGE * P2M_MID_PER_PAGE)));

static inline unsigned p2m_top_index(unsigned long pfn)
{
	BUG_ON(pfn >= MAX_P2M_PFN);
	return pfn / (P2M_MID_PER_PAGE * P2M_PER_PAGE);
}

static inline unsigned p2m_mid_index(unsigned long pfn)
{
	return (pfn / P2M_PER_PAGE) % P2M_MID_PER_PAGE;
}

static inline unsigned p2m_index(unsigned long pfn)
{
	return pfn % P2M_PER_PAGE;
}

static void p2m_top_init(unsigned long ***top)
{
	unsigned i;

	for (i = 0; i < P2M_TOP_PER_PAGE; i++)
		top[i] = p2m_mid_missing;
}

static void p2m_top_mfn_init(unsigned long *top)
{
	unsigned i;

	for (i = 0; i < P2M_TOP_PER_PAGE; i++)
		top[i] = virt_to_mfn(p2m_mid_missing_mfn);
}

static void p2m_top_mfn_p_init(unsigned long **top)
{
	unsigned i;

	for (i = 0; i < P2M_TOP_PER_PAGE; i++)
		top[i] = p2m_mid_missing_mfn;
}

static void p2m_mid_init(unsigned long **mid)
{
	unsigned i;

	for (i = 0; i < P2M_MID_PER_PAGE; i++)
		mid[i] = p2m_missing;
}

static void p2m_mid_mfn_init(unsigned long *mid)
{
	unsigned i;

	for (i = 0; i < P2M_MID_PER_PAGE; i++)
		mid[i] = virt_to_mfn(p2m_missing);
}

static void p2m_init(unsigned long *p2m)
{
	unsigned i;

	for (i = 0; i < P2M_MID_PER_PAGE; i++)
		p2m[i] = INVALID_P2M_ENTRY;
}

/*
 * Build the parallel p2m_top_mfn and p2m_mid_mfn structures
 *
 * This is called both at boot time, and after resuming from suspend:
 * - At boot time we're called very early, and must use extend_brk()
 *   to allocate memory.
 *
 * - After resume we're called from within stop_machine, but the mfn
 *   tree should alreay be completely allocated.
 */
void xen_build_mfn_list_list(void)
{
	unsigned long pfn;

	/* Pre-initialize p2m_top_mfn to be completely missing */
	if (p2m_top_mfn == NULL) {
		p2m_mid_missing_mfn = extend_brk(PAGE_SIZE, PAGE_SIZE);
		p2m_mid_mfn_init(p2m_mid_missing_mfn);

		p2m_top_mfn_p = extend_brk(PAGE_SIZE, PAGE_SIZE);
		p2m_top_mfn_p_init(p2m_top_mfn_p);

		p2m_top_mfn = extend_brk(PAGE_SIZE, PAGE_SIZE);
		p2m_top_mfn_init(p2m_top_mfn);
	} else {
		/* Reinitialise, mfn's all change after migration */
		p2m_mid_mfn_init(p2m_mid_missing_mfn);
	}

	for (pfn = 0; pfn < xen_max_p2m_pfn; pfn += P2M_PER_PAGE) {
		unsigned topidx = p2m_top_index(pfn);
		unsigned mididx = p2m_mid_index(pfn);
		unsigned long **mid;
		unsigned long *mid_mfn_p;

		mid = p2m_top[topidx];
		mid_mfn_p = p2m_top_mfn_p[topidx];

		/* Don't bother allocating any mfn mid levels if
		 * they're just missing, just update the stored mfn,
		 * since all could have changed over a migrate.
		 */
		if (mid == p2m_mid_missing) {
			BUG_ON(mididx);
			BUG_ON(mid_mfn_p != p2m_mid_missing_mfn);
			p2m_top_mfn[topidx] = virt_to_mfn(p2m_mid_missing_mfn);
			pfn += (P2M_MID_PER_PAGE - 1) * P2M_PER_PAGE;
			continue;
		}

		if (mid_mfn_p == p2m_mid_missing_mfn) {
			/*
			 * XXX boot-time only!  We should never find
			 * missing parts of the mfn tree after
			 * runtime.  extend_brk() will BUG if we call
			 * it too late.
			 */
			mid_mfn_p = extend_brk(PAGE_SIZE, PAGE_SIZE);
			p2m_mid_mfn_init(mid_mfn_p);

			p2m_top_mfn_p[topidx] = mid_mfn_p;
		}

		p2m_top_mfn[topidx] = virt_to_mfn(mid_mfn_p);
		mid_mfn_p[mididx] = virt_to_mfn(mid[mididx]);
	}
}

void xen_setup_mfn_list_list(void)
{
	BUG_ON(HYPERVISOR_shared_info == &xen_dummy_shared_info);

	HYPERVISOR_shared_info->arch.pfn_to_mfn_frame_list_list =
		virt_to_mfn(p2m_top_mfn);
	HYPERVISOR_shared_info->arch.max_pfn = xen_max_p2m_pfn;
}

/* Set up p2m_top to point to the domain-builder provided p2m pages */
void __init xen_build_dynamic_phys_to_machine(void)
{
	unsigned long *mfn_list = (unsigned long *)xen_start_info->mfn_list;
	unsigned long max_pfn = min(MAX_DOMAIN_PAGES, xen_start_info->nr_pages);
	unsigned long pfn;

	xen_max_p2m_pfn = max_pfn;

	p2m_missing = extend_brk(PAGE_SIZE, PAGE_SIZE);
	p2m_init(p2m_missing);

	p2m_mid_missing = extend_brk(PAGE_SIZE, PAGE_SIZE);
	p2m_mid_init(p2m_mid_missing);

	p2m_top = extend_brk(PAGE_SIZE, PAGE_SIZE);
	p2m_top_init(p2m_top);

	/*
	 * The domain builder gives us a pre-constructed p2m array in
	 * mfn_list for all the pages initially given to us, so we just
	 * need to graft that into our tree structure.
	 */
	for (pfn = 0; pfn < max_pfn; pfn += P2M_PER_PAGE) {
		unsigned topidx = p2m_top_index(pfn);
		unsigned mididx = p2m_mid_index(pfn);

		if (p2m_top[topidx] == p2m_mid_missing) {
			unsigned long **mid = extend_brk(PAGE_SIZE, PAGE_SIZE);
			p2m_mid_init(mid);

			p2m_top[topidx] = mid;
		}

		/*
		 * As long as the mfn_list has enough entries to completely
		 * fill a p2m page, pointing into the array is ok. But if
		 * not the entries beyond the last pfn will be undefined.
		 */
		if (unlikely(pfn + P2M_PER_PAGE > max_pfn)) {
			unsigned long p2midx;

			p2midx = max_pfn % P2M_PER_PAGE;
			for ( ; p2midx < P2M_PER_PAGE; p2midx++)
				mfn_list[pfn + p2midx] = INVALID_P2M_ENTRY;
		}
		p2m_top[topidx][mididx] = &mfn_list[pfn];
	}

	m2p_override_init();
}

unsigned long get_phys_to_machine(unsigned long pfn)
{
	unsigned topidx, mididx, idx;

	if (unlikely(pfn >= MAX_P2M_PFN))
		return INVALID_P2M_ENTRY;

	topidx = p2m_top_index(pfn);
	mididx = p2m_mid_index(pfn);
	idx = p2m_index(pfn);

	return p2m_top[topidx][mididx][idx];
}
EXPORT_SYMBOL_GPL(get_phys_to_machine);

static void *alloc_p2m_page(void)
{
	return (void *)__get_free_page(GFP_KERNEL | __GFP_REPEAT);
}

static void free_p2m_page(void *p)
{
	free_page((unsigned long)p);
}

/* 
 * Fully allocate the p2m structure for a given pfn.  We need to check
 * that both the top and mid levels are allocated, and make sure the
 * parallel mfn tree is kept in sync.  We may race with other cpus, so
 * the new pages are installed with cmpxchg; if we lose the race then
 * simply free the page we allocated and use the one that's there.
 */
static bool alloc_p2m(unsigned long pfn)
{
	unsigned topidx, mididx;
	unsigned long ***top_p, **mid;
	unsigned long *top_mfn_p, *mid_mfn;

	topidx = p2m_top_index(pfn);
	mididx = p2m_mid_index(pfn);

	top_p = &p2m_top[topidx];
	mid = *top_p;

	if (mid == p2m_mid_missing) {
		/* Mid level is missing, allocate a new one */
		mid = alloc_p2m_page();
		if (!mid)
			return false;

		p2m_mid_init(mid);

		if (cmpxchg(top_p, p2m_mid_missing, mid) != p2m_mid_missing)
			free_p2m_page(mid);
	}

	top_mfn_p = &p2m_top_mfn[topidx];
	mid_mfn = p2m_top_mfn_p[topidx];

	BUG_ON(virt_to_mfn(mid_mfn) != *top_mfn_p);

	if (mid_mfn == p2m_mid_missing_mfn) {
		/* Separately check the mid mfn level */
		unsigned long missing_mfn;
		unsigned long mid_mfn_mfn;

		mid_mfn = alloc_p2m_page();
		if (!mid_mfn)
			return false;

		p2m_mid_mfn_init(mid_mfn);

		missing_mfn = virt_to_mfn(p2m_mid_missing_mfn);
		mid_mfn_mfn = virt_to_mfn(mid_mfn);
		if (cmpxchg(top_mfn_p, missing_mfn, mid_mfn_mfn) != missing_mfn)
			free_p2m_page(mid_mfn);
		else
			p2m_top_mfn_p[topidx] = mid_mfn;
	}

	if (p2m_top[topidx][mididx] == p2m_missing) {
		/* p2m leaf page is missing */
		unsigned long *p2m;

		p2m = alloc_p2m_page();
		if (!p2m)
			return false;

		p2m_init(p2m);

		if (cmpxchg(&mid[mididx], p2m_missing, p2m) != p2m_missing)
			free_p2m_page(p2m);
		else
			mid_mfn[mididx] = virt_to_mfn(p2m);
	}

	return true;
}

/* Try to install p2m mapping; fail if intermediate bits missing */
bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn)
{
	unsigned topidx, mididx, idx;

	if (unlikely(pfn >= MAX_P2M_PFN)) {
		BUG_ON(mfn != INVALID_P2M_ENTRY);
		return true;
	}

	topidx = p2m_top_index(pfn);
	mididx = p2m_mid_index(pfn);
	idx = p2m_index(pfn);

	if (p2m_top[topidx][mididx] == p2m_missing)
		return mfn == INVALID_P2M_ENTRY;

	p2m_top[topidx][mididx][idx] = mfn;

	return true;
}

bool set_phys_to_machine(unsigned long pfn, unsigned long mfn)
{
	if (unlikely(xen_feature(XENFEAT_auto_translated_physmap))) {
		BUG_ON(pfn != mfn && mfn != INVALID_P2M_ENTRY);
		return true;
	}

	if (unlikely(!__set_phys_to_machine(pfn, mfn)))  {
		if (!alloc_p2m(pfn))
			return false;

		if (!__set_phys_to_machine(pfn, mfn))
			return false;
	}

	return true;
}

#define M2P_OVERRIDE_HASH_SHIFT	10
#define M2P_OVERRIDE_HASH	(1 << M2P_OVERRIDE_HASH_SHIFT)

static RESERVE_BRK_ARRAY(struct list_head, m2p_overrides, M2P_OVERRIDE_HASH);
static DEFINE_SPINLOCK(m2p_override_lock);

static void __init m2p_override_init(void)
{
	unsigned i;

	m2p_overrides = extend_brk(sizeof(*m2p_overrides) * M2P_OVERRIDE_HASH,
				   sizeof(unsigned long));

	for (i = 0; i < M2P_OVERRIDE_HASH; i++)
		INIT_LIST_HEAD(&m2p_overrides[i]);
}

static unsigned long mfn_hash(unsigned long mfn)
{
	return hash_long(mfn, M2P_OVERRIDE_HASH_SHIFT);
}

/* Add an MFN override for a particular page */
int m2p_add_override(unsigned long mfn, struct page *page)
{
	unsigned long flags;
	unsigned long pfn;
	unsigned long address;
	unsigned level;
	pte_t *ptep = NULL;

	pfn = page_to_pfn(page);
	if (!PageHighMem(page)) {
		address = (unsigned long)__va(pfn << PAGE_SHIFT);
		ptep = lookup_address(address, &level);

		if (WARN(ptep == NULL || level != PG_LEVEL_4K,
					"m2p_add_override: pfn %lx not mapped", pfn))
			return -EINVAL;
	}

	page->private = mfn;
	page->index = pfn_to_mfn(pfn);

	__set_phys_to_machine(pfn, FOREIGN_FRAME(mfn));
	if (!PageHighMem(page))
		/* Just zap old mapping for now */
		pte_clear(&init_mm, address, ptep);

	spin_lock_irqsave(&m2p_override_lock, flags);
	list_add(&page->lru,  &m2p_overrides[mfn_hash(mfn)]);
	spin_unlock_irqrestore(&m2p_override_lock, flags);

	return 0;
}

int m2p_remove_override(struct page *page)
{
	unsigned long flags;
	unsigned long mfn;
	unsigned long pfn;
	unsigned long address;
	unsigned level;
	pte_t *ptep = NULL;

	pfn = page_to_pfn(page);
	mfn = get_phys_to_machine(pfn);
	if (mfn == INVALID_P2M_ENTRY || !(mfn & FOREIGN_FRAME_BIT))
		return -EINVAL;

	if (!PageHighMem(page)) {
		address = (unsigned long)__va(pfn << PAGE_SHIFT);
		ptep = lookup_address(address, &level);

		if (WARN(ptep == NULL || level != PG_LEVEL_4K,
					"m2p_remove_override: pfn %lx not mapped", pfn))
			return -EINVAL;
	}

	spin_lock_irqsave(&m2p_override_lock, flags);
	list_del(&page->lru);
	spin_unlock_irqrestore(&m2p_override_lock, flags);
	__set_phys_to_machine(pfn, page->index);

	if (!PageHighMem(page))
		set_pte_at(&init_mm, address, ptep,
				pfn_pte(pfn, PAGE_KERNEL));
		/* No tlb flush necessary because the caller already
		 * left the pte unmapped. */

	return 0;
}

struct page *m2p_find_override(unsigned long mfn)
{
	unsigned long flags;
	struct list_head *bucket = &m2p_overrides[mfn_hash(mfn)];
	struct page *p, *ret;

	ret = NULL;

	spin_lock_irqsave(&m2p_override_lock, flags);

	list_for_each_entry(p, bucket, lru) {
		if (p->private == mfn) {
			ret = p;
			break;
		}
	}

	spin_unlock_irqrestore(&m2p_override_lock, flags);

	return ret;
}

unsigned long m2p_find_override_pfn(unsigned long mfn, unsigned long pfn)
{
	struct page *p = m2p_find_override(mfn);
	unsigned long ret = pfn;

	if (p)
		ret = page_to_pfn(p);

	return ret;
}
EXPORT_SYMBOL_GPL(m2p_find_override_pfn);
Commit	Line	Data
b5eafe92 JF	1	/*
	2	* Xen leaves the responsibility for maintaining p2m mappings to the
	3	* guests themselves, but it must also access and update the p2m array
	4	* during suspend/resume when all the pages are reallocated.
	5	*
	6	* The p2m table is logically a flat array, but we implement it as a
	7	* three-level tree to allow the address space to be sparse.
	8	*
	9	* Xen
	10	* \|
	11	* p2m_top p2m_top_mfn
	12	* / \ / \
	13	* p2m_mid p2m_mid p2m_mid_mfn p2m_mid_mfn
	14	* / \ / \ / /
	15	* p2m p2m p2m p2m p2m p2m p2m ...
	16	*
	17	* The p2m_mid_mfn pages are mapped by p2m_top_mfn_p.
	18	*
	19	* The p2m_top and p2m_top_mfn levels are limited to 1 page, so the
	20	* maximum representable pseudo-physical address space is:
	21	* P2M_TOP_PER_PAGE * P2M_MID_PER_PAGE * P2M_PER_PAGE pages
	22	*
	23	* P2M_PER_PAGE depends on the architecture, as a mfn is always
	24	* unsigned long (8 bytes on 64-bit, 4 bytes on 32), leading to
	25	* 512 and 1024 entries respectively.
	26	*/
	27
	28	#include <linux/init.h>
	29	#include <linux/module.h>
448f2831 JF	30	#include <linux/list.h>
448f2831 JF	31	#include <linux/hash.h>
87f1d40a	32	#include <linux/sched.h>
b5eafe92 JF	33
	34	#include <asm/cache.h>
	35	#include <asm/setup.h>
	36
	37	#include <asm/xen/page.h>
	38	#include <asm/xen/hypercall.h>
	39	#include <asm/xen/hypervisor.h>
	40
	41	#include "xen-ops.h"
	42
448f2831 JF	43	static void __init m2p_override_init(void);
448f2831 JF	44
b5eafe92 JF	45	unsigned long xen_max_p2m_pfn __read_mostly;
	46
	47	#define P2M_PER_PAGE (PAGE_SIZE / sizeof(unsigned long))
	48	#define P2M_MID_PER_PAGE (PAGE_SIZE / sizeof(unsigned long *))
	49	#define P2M_TOP_PER_PAGE (PAGE_SIZE / sizeof(unsigned long **))
	50
	51	#define MAX_P2M_PFN (P2M_TOP_PER_PAGE * P2M_MID_PER_PAGE * P2M_PER_PAGE)
	52
	53	/* Placeholders for holes in the address space */
	54	static RESERVE_BRK_ARRAY(unsigned long, p2m_missing, P2M_PER_PAGE);
	55	static RESERVE_BRK_ARRAY(unsigned long *, p2m_mid_missing, P2M_MID_PER_PAGE);
	56	static RESERVE_BRK_ARRAY(unsigned long, p2m_mid_missing_mfn, P2M_MID_PER_PAGE);
	57
	58	static RESERVE_BRK_ARRAY(unsigned long **, p2m_top, P2M_TOP_PER_PAGE);
	59	static RESERVE_BRK_ARRAY(unsigned long, p2m_top_mfn, P2M_TOP_PER_PAGE);
	60	static RESERVE_BRK_ARRAY(unsigned long *, p2m_top_mfn_p, P2M_TOP_PER_PAGE);
	61
	62	RESERVE_BRK(p2m_mid, PAGE_SIZE * (MAX_DOMAIN_PAGES / (P2M_PER_PAGE * P2M_MID_PER_PAGE)));
	63	RESERVE_BRK(p2m_mid_mfn, PAGE_SIZE * (MAX_DOMAIN_PAGES / (P2M_PER_PAGE * P2M_MID_PER_PAGE)));
	64
	65	static inline unsigned p2m_top_index(unsigned long pfn)
	66	{
	67	BUG_ON(pfn >= MAX_P2M_PFN);
	68	return pfn / (P2M_MID_PER_PAGE * P2M_PER_PAGE);
	69	}
	70
	71	static inline unsigned p2m_mid_index(unsigned long pfn)
	72	{
	73	return (pfn / P2M_PER_PAGE) % P2M_MID_PER_PAGE;
	74	}
	75
	76	static inline unsigned p2m_index(unsigned long pfn)
	77	{
	78	return pfn % P2M_PER_PAGE;
	79	}
	80
	81	static void p2m_top_init(unsigned long ***top)
	82	{
	83	unsigned i;
	84
	85	for (i = 0; i < P2M_TOP_PER_PAGE; i++)
	86	top[i] = p2m_mid_missing;
	87	}
	88
	89	static void p2m_top_mfn_init(unsigned long *top)
	90	{
	91	unsigned i;
	92
	93	for (i = 0; i < P2M_TOP_PER_PAGE; i++)
	94	top[i] = virt_to_mfn(p2m_mid_missing_mfn);
	95	}
	96
	97	static void p2m_top_mfn_p_init(unsigned long **top)
	98	{
	99	unsigned i;
	100
	101	for (i = 0; i < P2M_TOP_PER_PAGE; i++)
	102	top[i] = p2m_mid_missing_mfn;
	103	}
	104
	105	static void p2m_mid_init(unsigned long **mid)
	106	{
	107	unsigned i;
	108
109	for (i = 0; i < P2M_MID_PER_PAGE; i++)
110	mid[i] = p2m_missing;
111	}
112
113	static void p2m_mid_mfn_init(unsigned long *mid)
114	{
115	unsigned i;
116
117	for (i = 0; i < P2M_MID_PER_PAGE; i++)
118	mid[i] = virt_to_mfn(p2m_missing);
119	}
120
121	static void p2m_init(unsigned long *p2m)
122	{
123	unsigned i;
124
125	for (i = 0; i < P2M_MID_PER_PAGE; i++)
126	p2m[i] = INVALID_P2M_ENTRY;
127	}
128
129	/*
130	* Build the parallel p2m_top_mfn and p2m_mid_mfn structures
131	*
132	* This is called both at boot time, and after resuming from suspend:
133	* - At boot time we're called very early, and must use extend_brk()
134	* to allocate memory.
135	*
136	* - After resume we're called from within stop_machine, but the mfn
137	* tree should alreay be completely allocated.
138	*/
139	void xen_build_mfn_list_list(void)
140	{
141	unsigned long pfn;
142
143	/* Pre-initialize p2m_top_mfn to be completely missing */
144	if (p2m_top_mfn == NULL) {
145	p2m_mid_missing_mfn = extend_brk(PAGE_SIZE, PAGE_SIZE);
146	p2m_mid_mfn_init(p2m_mid_missing_mfn);
147
148	p2m_top_mfn_p = extend_brk(PAGE_SIZE, PAGE_SIZE);
149	p2m_top_mfn_p_init(p2m_top_mfn_p);
150
151	p2m_top_mfn = extend_brk(PAGE_SIZE, PAGE_SIZE);
152	p2m_top_mfn_init(p2m_top_mfn);
153	} else {
154	/* Reinitialise, mfn's all change after migration */
155	p2m_mid_mfn_init(p2m_mid_missing_mfn);
156	}
157
158	for (pfn = 0; pfn < xen_max_p2m_pfn; pfn += P2M_PER_PAGE) {
159	unsigned topidx = p2m_top_index(pfn);
160	unsigned mididx = p2m_mid_index(pfn);
161	unsigned long **mid;
162	unsigned long *mid_mfn_p;
163
164	mid = p2m_top[topidx];
165	mid_mfn_p = p2m_top_mfn_p[topidx];
166
167	/* Don't bother allocating any mfn mid levels if
168	* they're just missing, just update the stored mfn,
169	* since all could have changed over a migrate.
170	*/
171	if (mid == p2m_mid_missing) {
172	BUG_ON(mididx);
173	BUG_ON(mid_mfn_p != p2m_mid_missing_mfn);
174	p2m_top_mfn[topidx] = virt_to_mfn(p2m_mid_missing_mfn);
175	pfn += (P2M_MID_PER_PAGE - 1) * P2M_PER_PAGE;
176	continue;
177	}
178
179	if (mid_mfn_p == p2m_mid_missing_mfn) {
180	/*
181	* XXX boot-time only! We should never find
182	* missing parts of the mfn tree after
183	* runtime. extend_brk() will BUG if we call
184	* it too late.
185	*/
186	mid_mfn_p = extend_brk(PAGE_SIZE, PAGE_SIZE);
187	p2m_mid_mfn_init(mid_mfn_p);
188
189	p2m_top_mfn_p[topidx] = mid_mfn_p;
190	}
191
192	p2m_top_mfn[topidx] = virt_to_mfn(mid_mfn_p);
193	mid_mfn_p[mididx] = virt_to_mfn(mid[mididx]);
194	}
195	}
196
197	void xen_setup_mfn_list_list(void)
198	{
199	BUG_ON(HYPERVISOR_shared_info == &xen_dummy_shared_info);
200
201	HYPERVISOR_shared_info->arch.pfn_to_mfn_frame_list_list =
202	virt_to_mfn(p2m_top_mfn);
203	HYPERVISOR_shared_info->arch.max_pfn = xen_max_p2m_pfn;
204	}
205
206	/* Set up p2m_top to point to the domain-builder provided p2m pages */
207	void __init xen_build_dynamic_phys_to_machine(void)
208	{
209	unsigned long mfn_list = (unsigned long )xen_start_info->mfn_list;
210	unsigned long max_pfn = min(MAX_DOMAIN_PAGES, xen_start_info->nr_pages);
211	unsigned long pfn;
212
213	xen_max_p2m_pfn = max_pfn;
214
215	p2m_missing = extend_brk(PAGE_SIZE, PAGE_SIZE);
216	p2m_init(p2m_missing);
217
218	p2m_mid_missing = extend_brk(PAGE_SIZE, PAGE_SIZE);
219	p2m_mid_init(p2m_mid_missing);
220
221	p2m_top = extend_brk(PAGE_SIZE, PAGE_SIZE);
222	p2m_top_init(p2m_top);
223
224	/*
225	* The domain builder gives us a pre-constructed p2m array in
226	* mfn_list for all the pages initially given to us, so we just
227	* need to graft that into our tree structure.
228	*/
229	for (pfn = 0; pfn < max_pfn; pfn += P2M_PER_PAGE) {
230	unsigned topidx = p2m_top_index(pfn);
231	unsigned mididx = p2m_mid_index(pfn);
232
233	if (p2m_top[topidx] == p2m_mid_missing) {
234	unsigned long **mid = extend_brk(PAGE_SIZE, PAGE_SIZE);
235	p2m_mid_init(mid);
236
237	p2m_top[topidx] = mid;
238	}
239
8e1b4cf2 SB	240	/*
	241	* As long as the mfn_list has enough entries to completely
	242	* fill a p2m page, pointing into the array is ok. But if
	243	* not the entries beyond the last pfn will be undefined.
8e1b4cf2 SB	244	*/
	245	if (unlikely(pfn + P2M_PER_PAGE > max_pfn)) {
	246	unsigned long p2midx;
cf04d120 SB	247
	248	p2midx = max_pfn % P2M_PER_PAGE;
	249	for ( ; p2midx < P2M_PER_PAGE; p2midx++)
	250	mfn_list[pfn + p2midx] = INVALID_P2M_ENTRY;
	251	}
	252	p2m_top[topidx][mididx] = &mfn_list[pfn];
b5eafe92	253	}
448f2831 JF	254
448f2831 JF	255	m2p_override_init();
b5eafe92 JF	256	}
	257
	258	unsigned long get_phys_to_machine(unsigned long pfn)
	259	{
	260	unsigned topidx, mididx, idx;
	261
	262	if (unlikely(pfn >= MAX_P2M_PFN))
	263	return INVALID_P2M_ENTRY;
	264
	265	topidx = p2m_top_index(pfn);
	266	mididx = p2m_mid_index(pfn);
	267	idx = p2m_index(pfn);
	268
	269	return p2m_top[topidx][mididx][idx];
	270	}
	271	EXPORT_SYMBOL_GPL(get_phys_to_machine);
	272
	273	static void *alloc_p2m_page(void)
	274	{
	275	return (void *)__get_free_page(GFP_KERNEL \| __GFP_REPEAT);
	276	}
	277
	278	static void free_p2m_page(void *p)
	279	{
	280	free_page((unsigned long)p);
	281	}
	282
	283	/*
	284	* Fully allocate the p2m structure for a given pfn. We need to check
	285	* that both the top and mid levels are allocated, and make sure the
	286	* parallel mfn tree is kept in sync. We may race with other cpus, so
	287	* the new pages are installed with cmpxchg; if we lose the race then
	288	* simply free the page we allocated and use the one that's there.
	289	*/
	290	static bool alloc_p2m(unsigned long pfn)
	291	{
	292	unsigned topidx, mididx;
	293	unsigned long *top_p, mid;
	294	unsigned long top_mfn_p, mid_mfn;
	295
	296	topidx = p2m_top_index(pfn);
	297	mididx = p2m_mid_index(pfn);
	298
	299	top_p = &p2m_top[topidx];
	300	mid = *top_p;
	301
	302	if (mid == p2m_mid_missing) {
	303	/* Mid level is missing, allocate a new one */
	304	mid = alloc_p2m_page();
	305	if (!mid)
	306	return false;
	307
	308	p2m_mid_init(mid);
	309
	310	if (cmpxchg(top_p, p2m_mid_missing, mid) != p2m_mid_missing)
	311	free_p2m_page(mid);
	312	}
	313
	314	top_mfn_p = &p2m_top_mfn[topidx];
	315	mid_mfn = p2m_top_mfn_p[topidx];
	316
	317	BUG_ON(virt_to_mfn(mid_mfn) != *top_mfn_p);
	318
	319	if (mid_mfn == p2m_mid_missing_mfn) {
320	/* Separately check the mid mfn level */
321	unsigned long missing_mfn;
322	unsigned long mid_mfn_mfn;
323
324	mid_mfn = alloc_p2m_page();
325	if (!mid_mfn)
326	return false;
327
328	p2m_mid_mfn_init(mid_mfn);
329
330	missing_mfn = virt_to_mfn(p2m_mid_missing_mfn);
331	mid_mfn_mfn = virt_to_mfn(mid_mfn);
332	if (cmpxchg(top_mfn_p, missing_mfn, mid_mfn_mfn) != missing_mfn)
333	free_p2m_page(mid_mfn);
334	else
335	p2m_top_mfn_p[topidx] = mid_mfn;
336	}
337
338	if (p2m_top[topidx][mididx] == p2m_missing) {
339	/* p2m leaf page is missing */
340	unsigned long *p2m;
341
342	p2m = alloc_p2m_page();
343	if (!p2m)
344	return false;
345
346	p2m_init(p2m);
347
348	if (cmpxchg(&mid[mididx], p2m_missing, p2m) != p2m_missing)
349	free_p2m_page(p2m);
350	else
351	mid_mfn[mididx] = virt_to_mfn(p2m);
352	}
353
354	return true;
355	}
356
357	/* Try to install p2m mapping; fail if intermediate bits missing */
358	bool __set_phys_to_machine(unsigned long pfn, unsigned long mfn)
359	{
360	unsigned topidx, mididx, idx;
361
362	if (unlikely(pfn >= MAX_P2M_PFN)) {
363	BUG_ON(mfn != INVALID_P2M_ENTRY);
364	return true;
365	}
366
367	topidx = p2m_top_index(pfn);
368	mididx = p2m_mid_index(pfn);
369	idx = p2m_index(pfn);
370
371	if (p2m_top[topidx][mididx] == p2m_missing)
372	return mfn == INVALID_P2M_ENTRY;
373
374	p2m_top[topidx][mididx][idx] = mfn;
375
376	return true;
377	}
378
379	bool set_phys_to_machine(unsigned long pfn, unsigned long mfn)
380	{
381	if (unlikely(xen_feature(XENFEAT_auto_translated_physmap))) {
382	BUG_ON(pfn != mfn && mfn != INVALID_P2M_ENTRY);
383	return true;
384	}
385
386	if (unlikely(!__set_phys_to_machine(pfn, mfn))) {
387	if (!alloc_p2m(pfn))
388	return false;
389
390	if (!__set_phys_to_machine(pfn, mfn))
391	return false;
392	}
393
394	return true;
395	}
448f2831 JF	396
	397	#define M2P_OVERRIDE_HASH_SHIFT 10
	398	#define M2P_OVERRIDE_HASH (1 << M2P_OVERRIDE_HASH_SHIFT)
	399
	400	static RESERVE_BRK_ARRAY(struct list_head, m2p_overrides, M2P_OVERRIDE_HASH);
	401	static DEFINE_SPINLOCK(m2p_override_lock);
	402
	403	static void __init m2p_override_init(void)
	404	{
	405	unsigned i;
	406
	407	m2p_overrides = extend_brk(sizeof(m2p_overrides) M2P_OVERRIDE_HASH,
	408	sizeof(unsigned long));
	409
	410	for (i = 0; i < M2P_OVERRIDE_HASH; i++)
	411	INIT_LIST_HEAD(&m2p_overrides[i]);
	412	}
	413
	414	static unsigned long mfn_hash(unsigned long mfn)
	415	{
	416	return hash_long(mfn, M2P_OVERRIDE_HASH_SHIFT);
	417	}
	418
	419	/* Add an MFN override for a particular page */
87f1d40a	420	int m2p_add_override(unsigned long mfn, struct page *page)
448f2831 JF	421	{
448f2831 JF	422	unsigned long flags;
87f1d40a JF	423	unsigned long pfn;
	424	unsigned long address;
	425	unsigned level;
	426	pte_t *ptep = NULL;
	427
	428	pfn = page_to_pfn(page);
	429	if (!PageHighMem(page)) {
	430	address = (unsigned long)__va(pfn << PAGE_SHIFT);
	431	ptep = lookup_address(address, &level);
	432
	433	if (WARN(ptep == NULL \|\| level != PG_LEVEL_4K,
	434	"m2p_add_override: pfn %lx not mapped", pfn))
	435	return -EINVAL;
	436	}
	437
448f2831	438	page->private = mfn;
9b705f0e	439	page->index = pfn_to_mfn(pfn);
448f2831	440
9b705f0e	441	__set_phys_to_machine(pfn, FOREIGN_FRAME(mfn));
87f1d40a JF	442	if (!PageHighMem(page))
	443	/* Just zap old mapping for now */
	444	pte_clear(&init_mm, address, ptep);
	445
448f2831 JF	446	spin_lock_irqsave(&m2p_override_lock, flags);
	447	list_add(&page->lru, &m2p_overrides[mfn_hash(mfn)]);
	448	spin_unlock_irqrestore(&m2p_override_lock, flags);
87f1d40a JF	449
87f1d40a JF	450	return 0;
448f2831 JF	451	}
448f2831 JF	452
87f1d40a	453	int m2p_remove_override(struct page *page)
448f2831 JF	454	{
448f2831 JF	455	unsigned long flags;
9b705f0e SS	456	unsigned long mfn;
9b705f0e SS	457	unsigned long pfn;
87f1d40a JF	458	unsigned long address;
	459	unsigned level;
	460	pte_t *ptep = NULL;
9b705f0e SS	461
	462	pfn = page_to_pfn(page);
	463	mfn = get_phys_to_machine(pfn);
	464	if (mfn == INVALID_P2M_ENTRY \|\| !(mfn & FOREIGN_FRAME_BIT))
87f1d40a JF	465	return -EINVAL;
	466
	467	if (!PageHighMem(page)) {
	468	address = (unsigned long)__va(pfn << PAGE_SHIFT);
	469	ptep = lookup_address(address, &level);
	470
	471	if (WARN(ptep == NULL \|\| level != PG_LEVEL_4K,
	472	"m2p_remove_override: pfn %lx not mapped", pfn))
	473	return -EINVAL;
	474	}
9b705f0e	475
448f2831 JF	476	spin_lock_irqsave(&m2p_override_lock, flags);
	477	list_del(&page->lru);
	478	spin_unlock_irqrestore(&m2p_override_lock, flags);
9b705f0e	479	__set_phys_to_machine(pfn, page->index);
87f1d40a JF	480
	481	if (!PageHighMem(page))
	482	set_pte_at(&init_mm, address, ptep,
	483	pfn_pte(pfn, PAGE_KERNEL));
	484	/* No tlb flush necessary because the caller already
	485	* left the pte unmapped. */
	486
	487	return 0;
448f2831 JF	488	}
	489
	490	struct page *m2p_find_override(unsigned long mfn)
	491	{
	492	unsigned long flags;
	493	struct list_head *bucket = &m2p_overrides[mfn_hash(mfn)];
	494	struct page p, ret;
	495
	496	ret = NULL;
	497
	498	spin_lock_irqsave(&m2p_override_lock, flags);
	499
	500	list_for_each_entry(p, bucket, lru) {
	501	if (p->private == mfn) {
	502	ret = p;
	503	break;
	504	}
	505	}
	506
	507	spin_unlock_irqrestore(&m2p_override_lock, flags);
	508
	509	return ret;
	510	}
	511
	512	unsigned long m2p_find_override_pfn(unsigned long mfn, unsigned long pfn)
	513	{
	514	struct page *p = m2p_find_override(mfn);
	515	unsigned long ret = pfn;
	516
	517	if (p)
	518	ret = page_to_pfn(p);
	519
	520	return ret;
	521	}
e1b478e4	522	EXPORT_SYMBOL_GPL(m2p_find_override_pfn);