[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/config.h>
#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;
}

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++;

	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--;

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