[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/smp_lock.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 (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)
{
	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;
}

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