[deliverable/linux.git] / arch / sparc64 / mm / hugetlbpage.c

/*
 * SPARC64 Huge TLB page support.
 *
 * Copyright (C) 2002, 2003, 2006 David S. Miller (davem@davemloft.net)
 */

#include <linux/init.h>
#include <linux/module.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/hugetlb.h>
#include <linux/pagemap.h>
#include <linux/slab.h>
#include <linux/sysctl.h>

#include <asm/mman.h>
#include <asm/pgalloc.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>
#include <asm/cacheflush.h>
#include <asm/mmu_context.h>

/* Slightly simplified from the non-hugepage variant because by
 * definition we don't have to worry about any page coloring stuff
 */
#define VA_EXCLUDE_START (0x0000080000000000UL - (1UL << 32UL))
#define VA_EXCLUDE_END   (0xfffff80000000000UL + (1UL << 32UL))

static unsigned long hugetlb_get_unmapped_area_bottomup(struct file *filp,
							unsigned long addr,
							unsigned long len,
							unsigned long pgoff,
							unsigned long flags)
{
	struct mm_struct *mm = current->mm;
	struct vm_area_struct * vma;
	unsigned long task_size = TASK_SIZE;
	unsigned long start_addr;

	if (test_thread_flag(TIF_32BIT))
		task_size = STACK_TOP32;
	if (unlikely(len >= VA_EXCLUDE_START))
		return -ENOMEM;

	if (len > mm->cached_hole_size) {
	        start_addr = addr = mm->free_area_cache;
	} else {
	        start_addr = addr = TASK_UNMAPPED_BASE;
	        mm->cached_hole_size = 0;
	}

	task_size -= len;

full_search:
	addr = ALIGN(addr, HPAGE_SIZE);

	for (vma = find_vma(mm, addr); ; vma = vma->vm_next) {
		/* At this point:  (!vma || addr < vma->vm_end). */
		if (addr < VA_EXCLUDE_START &&
		    (addr + len) >= VA_EXCLUDE_START) {
			addr = VA_EXCLUDE_END;
			vma = find_vma(mm, VA_EXCLUDE_END);
		}
		if (unlikely(task_size < addr)) {
			if (start_addr != TASK_UNMAPPED_BASE) {
				start_addr = addr = TASK_UNMAPPED_BASE;
				mm->cached_hole_size = 0;
				goto full_search;
			}
			return -ENOMEM;
		}
		if (likely(!vma || addr + len <= vma->vm_start)) {
			/*
			 * Remember the place where we stopped the search:
			 */
			mm->free_area_cache = addr + len;
			return addr;
		}
		if (addr + mm->cached_hole_size < vma->vm_start)
		        mm->cached_hole_size = vma->vm_start - addr;

		addr = ALIGN(vma->vm_end, HPAGE_SIZE);
	}
}

static unsigned long
hugetlb_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
				  const unsigned long len,
				  const unsigned long pgoff,
				  const unsigned long flags)
{
	struct vm_area_struct *vma;
	struct mm_struct *mm = current->mm;
	unsigned long addr = addr0;

	/* This should only ever run for 32-bit processes.  */
	BUG_ON(!test_thread_flag(TIF_32BIT));

	/* check if free_area_cache is useful for us */
	if (len <= mm->cached_hole_size) {
 	        mm->cached_hole_size = 0;
 		mm->free_area_cache = mm->mmap_base;
 	}

	/* either no address requested or can't fit in requested address hole */
	addr = mm->free_area_cache & HPAGE_MASK;

	/* make sure it can fit in the remaining address space */
	if (likely(addr > len)) {
		vma = find_vma(mm, addr-len);
		if (!vma || addr <= vma->vm_start) {
			/* remember the address as a hint for next time */
			return (mm->free_area_cache = addr-len);
		}
	}

	if (unlikely(mm->mmap_base < len))
		goto bottomup;

	addr = (mm->mmap_base-len) & HPAGE_MASK;

	do {
		/*
		 * Lookup failure means no vma is above this address,
		 * else if new region fits below vma->vm_start,
		 * return with success:
		 */
		vma = find_vma(mm, addr);
		if (likely(!vma || addr+len <= vma->vm_start)) {
			/* remember the address as a hint for next time */
			return (mm->free_area_cache = addr);
		}

 		/* remember the largest hole we saw so far */
 		if (addr + mm->cached_hole_size < vma->vm_start)
 		        mm->cached_hole_size = vma->vm_start - addr;

		/* try just below the current vma->vm_start */
		addr = (vma->vm_start-len) & HPAGE_MASK;
	} while (likely(len < vma->vm_start));

bottomup:
	/*
	 * A failed mmap() very likely causes application failure,
	 * so fall back to the bottom-up function here. This scenario
	 * can happen with large stack limits and large mmap()
	 * allocations.
	 */
	mm->cached_hole_size = ~0UL;
  	mm->free_area_cache = TASK_UNMAPPED_BASE;
	addr = arch_get_unmapped_area(filp, addr0, len, pgoff, flags);
	/*
	 * Restore the topdown base:
	 */
	mm->free_area_cache = mm->mmap_base;
	mm->cached_hole_size = ~0UL;

	return addr;
}

unsigned long
hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
		unsigned long len, unsigned long pgoff, unsigned long flags)
{
	struct mm_struct *mm = current->mm;
	struct vm_area_struct *vma;
	unsigned long task_size = TASK_SIZE;

	if (test_thread_flag(TIF_32BIT))
		task_size = STACK_TOP32;

	if (len & ~HPAGE_MASK)
		return -EINVAL;
	if (len > task_size)
		return -ENOMEM;

	if (flags & MAP_FIXED) {
		if (prepare_hugepage_range(file, addr, len))
			return -EINVAL;
		return addr;
	}

	if (addr) {
		addr = ALIGN(addr, HPAGE_SIZE);
		vma = find_vma(mm, addr);
		if (task_size - len >= addr &&
		    (!vma || addr + len <= vma->vm_start))
			return addr;
	}
	if (mm->get_unmapped_area == arch_get_unmapped_area)
		return hugetlb_get_unmapped_area_bottomup(file, addr, len,
				pgoff, flags);
	else
		return hugetlb_get_unmapped_area_topdown(file, addr, len,
				pgoff, flags);
}

pte_t *huge_pte_alloc(struct mm_struct *mm,
			unsigned long addr, unsigned long sz)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte = NULL;

	/* We must align the address, because our caller will run
	 * set_huge_pte_at() on whatever we return, which writes out
	 * all of the sub-ptes for the hugepage range.  So we have
	 * to give it the first such sub-pte.
	 */
	addr &= HPAGE_MASK;

	pgd = pgd_offset(mm, addr);
	pud = pud_alloc(mm, pgd, addr);
	if (pud) {
		pmd = pmd_alloc(mm, pud, addr);
		if (pmd)
			pte = pte_alloc_map(mm, pmd, addr);
	}
	return pte;
}

pte_t *huge_pte_offset(struct mm_struct *mm, unsigned long addr)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte = NULL;

	addr &= HPAGE_MASK;

	pgd = pgd_offset(mm, addr);
	if (!pgd_none(*pgd)) {
		pud = pud_offset(pgd, addr);
		if (!pud_none(*pud)) {
			pmd = pmd_offset(pud, addr);
			if (!pmd_none(*pmd))
				pte = pte_offset_map(pmd, addr);
		}
	}
	return pte;
}

int huge_pmd_unshare(struct mm_struct *mm, unsigned long *addr, pte_t *ptep)
{
	return 0;
}

void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
		     pte_t *ptep, pte_t entry)
{
	int i;

	if (!pte_present(*ptep) && pte_present(entry))
		mm->context.huge_pte_count++;

	addr &= HPAGE_MASK;
	for (i = 0; i < (1 << HUGETLB_PAGE_ORDER); i++) {
		set_pte_at(mm, addr, ptep, entry);
		ptep++;
		addr += PAGE_SIZE;
		pte_val(entry) += PAGE_SIZE;
	}
}

pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
			      pte_t *ptep)
{
	pte_t entry;
	int i;

	entry = *ptep;
	if (pte_present(entry))
		mm->context.huge_pte_count--;

	addr &= HPAGE_MASK;

	for (i = 0; i < (1 << HUGETLB_PAGE_ORDER); i++) {
		pte_clear(mm, addr, ptep);
		addr += PAGE_SIZE;
		ptep++;
	}

	return entry;
}

struct page *follow_huge_addr(struct mm_struct *mm,
			      unsigned long address, int write)
{
	return ERR_PTR(-EINVAL);
}

int pmd_huge(pmd_t pmd)
{
	return 0;
}

int pud_huge(pud_t pud)
{
	return 0;
}

struct page *follow_huge_pmd(struct mm_struct *mm, unsigned long address,
			     pmd_t *pmd, int write)
{
	return NULL;
}

static void context_reload(void *__data)
{
	struct mm_struct *mm = __data;

	if (mm == current->mm)
		load_secondary_context(mm);
}

void hugetlb_prefault_arch_hook(struct mm_struct *mm)
{
	struct tsb_config *tp = &mm->context.tsb_block[MM_TSB_HUGE];

	if (likely(tp->tsb != NULL))
		return;

	tsb_grow(mm, MM_TSB_HUGE, 0);
	tsb_context_switch(mm);
	smp_tsb_sync(mm);

	/* On UltraSPARC-III+ and later, configure the second half of
	 * the Data-TLB for huge pages.
	 */
	if (tlb_type == cheetah_plus) {
		unsigned long ctx;

		spin_lock(&ctx_alloc_lock);
		ctx = mm->context.sparc64_ctx_val;
		ctx &= ~CTX_PGSZ_MASK;
		ctx |= CTX_PGSZ_BASE << CTX_PGSZ0_SHIFT;
		ctx |= CTX_PGSZ_HUGE << CTX_PGSZ1_SHIFT;

		if (ctx != mm->context.sparc64_ctx_val) {
			/* When changing the page size fields, we
			 * must perform a context flush so that no
			 * stale entries match.  This flush must
			 * occur with the original context register
			 * settings.
			 */
			do_flush_tlb_mm(mm);

			/* Reload the context register of all processors
			 * also executing in this address space.
			 */
			mm->context.sparc64_ctx_val = ctx;
			on_each_cpu(context_reload, mm, 0);
		}
		spin_unlock(&ctx_alloc_lock);
	}
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* SPARC64 Huge TLB page support.
	3	*
f6b83f07	4	* Copyright (C) 2002, 2003, 2006 David S. Miller (davem@davemloft.net)
1da177e4 LT	5	*/
1da177e4 LT	6
1da177e4 LT	7	#include <linux/init.h>
	8	#include <linux/module.h>
	9	#include <linux/fs.h>
	10	#include <linux/mm.h>
	11	#include <linux/hugetlb.h>
	12	#include <linux/pagemap.h>
1da177e4 LT	13	#include <linux/slab.h>
	14	#include <linux/sysctl.h>
	15
	16	#include <asm/mman.h>
	17	#include <asm/pgalloc.h>
	18	#include <asm/tlb.h>
	19	#include <asm/tlbflush.h>
	20	#include <asm/cacheflush.h>
	21	#include <asm/mmu_context.h>
	22
f6b83f07 DM	23	/* Slightly simplified from the non-hugepage variant because by
	24	* definition we don't have to worry about any page coloring stuff
	25	*/
	26	#define VA_EXCLUDE_START (0x0000080000000000UL - (1UL << 32UL))
	27	#define VA_EXCLUDE_END (0xfffff80000000000UL + (1UL << 32UL))
	28
	29	static unsigned long hugetlb_get_unmapped_area_bottomup(struct file *filp,
	30	unsigned long addr,
	31	unsigned long len,
	32	unsigned long pgoff,
	33	unsigned long flags)
	34	{
	35	struct mm_struct *mm = current->mm;
	36	struct vm_area_struct * vma;
	37	unsigned long task_size = TASK_SIZE;
	38	unsigned long start_addr;
	39
	40	if (test_thread_flag(TIF_32BIT))
	41	task_size = STACK_TOP32;
	42	if (unlikely(len >= VA_EXCLUDE_START))
	43	return -ENOMEM;
	44
	45	if (len > mm->cached_hole_size) {
	46	start_addr = addr = mm->free_area_cache;
	47	} else {
	48	start_addr = addr = TASK_UNMAPPED_BASE;
	49	mm->cached_hole_size = 0;
	50	}
	51
	52	task_size -= len;
	53
	54	full_search:
	55	addr = ALIGN(addr, HPAGE_SIZE);
	56
	57	for (vma = find_vma(mm, addr); ; vma = vma->vm_next) {
	58	/* At this point: (!vma \|\| addr < vma->vm_end). */
	59	if (addr < VA_EXCLUDE_START &&
	60	(addr + len) >= VA_EXCLUDE_START) {
	61	addr = VA_EXCLUDE_END;
	62	vma = find_vma(mm, VA_EXCLUDE_END);
	63	}
	64	if (unlikely(task_size < addr)) {
	65	if (start_addr != TASK_UNMAPPED_BASE) {
	66	start_addr = addr = TASK_UNMAPPED_BASE;
	67	mm->cached_hole_size = 0;
	68	goto full_search;
	69	}
	70	return -ENOMEM;
	71	}
	72	if (likely(!vma \|\| addr + len <= vma->vm_start)) {
	73	/*
	74	* Remember the place where we stopped the search:
	75	*/
	76	mm->free_area_cache = addr + len;
	77	return addr;
	78	}
	79	if (addr + mm->cached_hole_size < vma->vm_start)
	80	mm->cached_hole_size = vma->vm_start - addr;
	81
	82	addr = ALIGN(vma->vm_end, HPAGE_SIZE);
	83	}
	84	}
	85
	86	static unsigned long
87	hugetlb_get_unmapped_area_topdown(struct file *filp, const unsigned long addr0,
88	const unsigned long len,
89	const unsigned long pgoff,
90	const unsigned long flags)
91	{
92	struct vm_area_struct *vma;
93	struct mm_struct *mm = current->mm;
94	unsigned long addr = addr0;
95
96	/* This should only ever run for 32-bit processes. */
97	BUG_ON(!test_thread_flag(TIF_32BIT));
98
99	/* check if free_area_cache is useful for us */
100	if (len <= mm->cached_hole_size) {
101	mm->cached_hole_size = 0;
102	mm->free_area_cache = mm->mmap_base;
103	}
104
105	/* either no address requested or can't fit in requested address hole */
106	addr = mm->free_area_cache & HPAGE_MASK;
107
108	/* make sure it can fit in the remaining address space */
109	if (likely(addr > len)) {
110	vma = find_vma(mm, addr-len);
111	if (!vma \|\| addr <= vma->vm_start) {
112	/* remember the address as a hint for next time */
113	return (mm->free_area_cache = addr-len);
114	}
115	}
116
117	if (unlikely(mm->mmap_base < len))
118	goto bottomup;
119
120	addr = (mm->mmap_base-len) & HPAGE_MASK;
121
122	do {
123	/*
124	* Lookup failure means no vma is above this address,
125	* else if new region fits below vma->vm_start,
126	* return with success:
127	*/
128	vma = find_vma(mm, addr);
129	if (likely(!vma \|\| addr+len <= vma->vm_start)) {
130	/* remember the address as a hint for next time */
131	return (mm->free_area_cache = addr);
132	}
133
134	/* remember the largest hole we saw so far */
135	if (addr + mm->cached_hole_size < vma->vm_start)
136	mm->cached_hole_size = vma->vm_start - addr;
137
138	/* try just below the current vma->vm_start */
139	addr = (vma->vm_start-len) & HPAGE_MASK;
140	} while (likely(len < vma->vm_start));
141
142	bottomup:
143	/*
144	* A failed mmap() very likely causes application failure,
145	* so fall back to the bottom-up function here. This scenario
146	* can happen with large stack limits and large mmap()
147	* allocations.
148	*/
149	mm->cached_hole_size = ~0UL;
150	mm->free_area_cache = TASK_UNMAPPED_BASE;
151	addr = arch_get_unmapped_area(filp, addr0, len, pgoff, flags);
152	/*
153	* Restore the topdown base:
154	*/
155	mm->free_area_cache = mm->mmap_base;
156	mm->cached_hole_size = ~0UL;
157
158	return addr;
159	}
160
161	unsigned long
162	hugetlb_get_unmapped_area(struct file *file, unsigned long addr,
163	unsigned long len, unsigned long pgoff, unsigned long flags)
164	{
165	struct mm_struct *mm = current->mm;
166	struct vm_area_struct *vma;
167	unsigned long task_size = TASK_SIZE;
168
169	if (test_thread_flag(TIF_32BIT))
170	task_size = STACK_TOP32;
171
172	if (len & ~HPAGE_MASK)
173	return -EINVAL;
174	if (len > task_size)
175	return -ENOMEM;
176
ac35ee48	177	if (flags & MAP_FIXED) {
a5516438	178	if (prepare_hugepage_range(file, addr, len))
ac35ee48 BH	179	return -EINVAL;
	180	return addr;
	181	}
	182
f6b83f07 DM	183	if (addr) {
	184	addr = ALIGN(addr, HPAGE_SIZE);
	185	vma = find_vma(mm, addr);
	186	if (task_size - len >= addr &&
	187	(!vma \|\| addr + len <= vma->vm_start))
	188	return addr;
	189	}
	190	if (mm->get_unmapped_area == arch_get_unmapped_area)
	191	return hugetlb_get_unmapped_area_bottomup(file, addr, len,
	192	pgoff, flags);
	193	else
	194	return hugetlb_get_unmapped_area_topdown(file, addr, len,
	195	pgoff, flags);
	196	}
	197
a5516438 AK	198	pte_t huge_pte_alloc(struct mm_struct mm,
a5516438 AK	199	unsigned long addr, unsigned long sz)
1da177e4 LT	200	{
	201	pgd_t *pgd;
	202	pud_t *pud;
	203	pmd_t *pmd;
	204	pte_t *pte = NULL;
	205
9df1dab1 DM	206	/* We must align the address, because our caller will run
	207	* set_huge_pte_at() on whatever we return, which writes out
	208	* all of the sub-ptes for the hugepage range. So we have
	209	* to give it the first such sub-pte.
	210	*/
	211	addr &= HPAGE_MASK;
	212
1da177e4	213	pgd = pgd_offset(mm, addr);
dcc1e8dd DM	214	pud = pud_alloc(mm, pgd, addr);
	215	if (pud) {
	216	pmd = pmd_alloc(mm, pud, addr);
	217	if (pmd)
	218	pte = pte_alloc_map(mm, pmd, addr);
1da177e4 LT	219	}
	220	return pte;
	221	}
	222
63551ae0	223	pte_t huge_pte_offset(struct mm_struct mm, unsigned long addr)
1da177e4 LT	224	{
	225	pgd_t *pgd;
	226	pud_t *pud;
	227	pmd_t *pmd;
	228	pte_t *pte = NULL;
	229
f6b83f07 DM	230	addr &= HPAGE_MASK;
f6b83f07 DM	231
1da177e4	232	pgd = pgd_offset(mm, addr);
f6b83f07	233	if (!pgd_none(*pgd)) {
1da177e4	234	pud = pud_offset(pgd, addr);
f6b83f07	235	if (!pud_none(*pud)) {
1da177e4	236	pmd = pmd_offset(pud, addr);
f6b83f07	237	if (!pmd_none(*pmd))
1da177e4 LT	238	pte = pte_offset_map(pmd, addr);
	239	}
	240	}
	241	return pte;
	242	}
	243
39dde65c CK	244	int huge_pmd_unshare(struct mm_struct mm, unsigned long addr, pte_t *ptep)
	245	{
	246	return 0;
	247	}
	248
63551ae0 DG	249	void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
63551ae0 DG	250	pte_t *ptep, pte_t entry)
1da177e4	251	{
63551ae0 DG	252	int i;
63551ae0 DG	253
dcc1e8dd DM	254	if (!pte_present(*ptep) && pte_present(entry))
	255	mm->context.huge_pte_count++;
	256
bb8236f2	257	addr &= HPAGE_MASK;
63551ae0 DG	258	for (i = 0; i < (1 << HUGETLB_PAGE_ORDER); i++) {
	259	set_pte_at(mm, addr, ptep, entry);
	260	ptep++;
	261	addr += PAGE_SIZE;
	262	pte_val(entry) += PAGE_SIZE;
	263	}
	264	}
1da177e4	265
63551ae0 DG	266	pte_t huge_ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
	267	pte_t *ptep)
	268	{
	269	pte_t entry;
	270	int i;
1da177e4	271
63551ae0	272	entry = *ptep;
dcc1e8dd DM	273	if (pte_present(entry))
dcc1e8dd DM	274	mm->context.huge_pte_count--;
1da177e4	275
bb8236f2 DM	276	addr &= HPAGE_MASK;
bb8236f2 DM	277
1da177e4	278	for (i = 0; i < (1 << HUGETLB_PAGE_ORDER); i++) {
63551ae0	279	pte_clear(mm, addr, ptep);
1da177e4	280	addr += PAGE_SIZE;
63551ae0	281	ptep++;
1da177e4	282	}
63551ae0 DG	283
63551ae0 DG	284	return entry;
1da177e4 LT	285	}
1da177e4 LT	286
1da177e4 LT	287	struct page follow_huge_addr(struct mm_struct mm,
	288	unsigned long address, int write)
	289	{
	290	return ERR_PTR(-EINVAL);
	291	}
	292
	293	int pmd_huge(pmd_t pmd)
	294	{
	295	return 0;
	296	}
	297
ceb86879 AK	298	int pud_huge(pud_t pud)
	299	{
	300	return 0;
	301	}
	302
1da177e4 LT	303	struct page follow_huge_pmd(struct mm_struct mm, unsigned long address,
	304	pmd_t *pmd, int write)
	305	{
	306	return NULL;
	307	}
	308
1da177e4 LT	309	static void context_reload(void *__data)
	310	{
	311	struct mm_struct *mm = __data;
	312
	313	if (mm == current->mm)
	314	load_secondary_context(mm);
	315	}
	316
63551ae0	317	void hugetlb_prefault_arch_hook(struct mm_struct *mm)
1da177e4	318	{
dcc1e8dd DM	319	struct tsb_config *tp = &mm->context.tsb_block[MM_TSB_HUGE];
	320
	321	if (likely(tp->tsb != NULL))
	322	return;
	323
	324	tsb_grow(mm, MM_TSB_HUGE, 0);
	325	tsb_context_switch(mm);
	326	smp_tsb_sync(mm);
	327
1da177e4 LT	328	/* On UltraSPARC-III+ and later, configure the second half of
	329	* the Data-TLB for huge pages.
	330	*/
	331	if (tlb_type == cheetah_plus) {
	332	unsigned long ctx;
	333
	334	spin_lock(&ctx_alloc_lock);
	335	ctx = mm->context.sparc64_ctx_val;
	336	ctx &= ~CTX_PGSZ_MASK;
	337	ctx \|= CTX_PGSZ_BASE << CTX_PGSZ0_SHIFT;
	338	ctx \|= CTX_PGSZ_HUGE << CTX_PGSZ1_SHIFT;
	339
	340	if (ctx != mm->context.sparc64_ctx_val) {
	341	/* When changing the page size fields, we
	342	* must perform a context flush so that no
	343	* stale entries match. This flush must
	344	* occur with the original context register
	345	* settings.
	346	*/
	347	do_flush_tlb_mm(mm);
	348
	349	/* Reload the context register of all processors
	350	* also executing in this address space.
	351	*/
	352	mm->context.sparc64_ctx_val = ctx;
15c8b6c1	353	on_each_cpu(context_reload, mm, 0);
1da177e4 LT	354	}
	355	spin_unlock(&ctx_alloc_lock);
	356	}
1da177e4	357	}