[mirror_ubuntu-zesty-kernel.git] / arch / arm64 / mm / hugetlbpage.c

/*
 * arch/arm64/mm/hugetlbpage.c
 *
 * Copyright (C) 2013 Linaro Ltd.
 *
 * Based on arch/x86/mm/hugetlbpage.c.
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License version 2 as
 * published by the Free Software Foundation.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 */

#include <linux/init.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/hugetlb.h>
#include <linux/pagemap.h>
#include <linux/err.h>
#include <linux/sysctl.h>
#include <asm/mman.h>
#include <asm/tlb.h>
#include <asm/tlbflush.h>
#include <asm/pgalloc.h>

int pmd_huge(pmd_t pmd)
{
	return pmd_val(pmd) && !(pmd_val(pmd) & PMD_TABLE_BIT);
}

int pud_huge(pud_t pud)
{
#ifndef __PAGETABLE_PMD_FOLDED
	return pud_val(pud) && !(pud_val(pud) & PUD_TABLE_BIT);
#else
	return 0;
#endif
}

static int find_num_contig(struct mm_struct *mm, unsigned long addr,
			   pte_t *ptep, pte_t pte, size_t *pgsize)
{
	pgd_t *pgd = pgd_offset(mm, addr);
	pud_t *pud;
	pmd_t *pmd;

	*pgsize = PAGE_SIZE;
	if (!pte_cont(pte))
		return 1;
	pud = pud_offset(pgd, addr);
	pmd = pmd_offset(pud, addr);
	if ((pte_t *)pmd == ptep) {
		*pgsize = PMD_SIZE;
		return CONT_PMDS;
	}
	return CONT_PTES;
}

void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
			    pte_t *ptep, pte_t pte)
{
	size_t pgsize;
	int i;
	int ncontig = find_num_contig(mm, addr, ptep, pte, &pgsize);
	unsigned long pfn;
	pgprot_t hugeprot;

	if (ncontig == 1) {
		set_pte_at(mm, addr, ptep, pte);
		return;
	}

	pfn = pte_pfn(pte);
	hugeprot = __pgprot(pte_val(pfn_pte(pfn, __pgprot(0))) ^ pte_val(pte));
	for (i = 0; i < ncontig; i++) {
		pr_debug("%s: set pte %p to 0x%llx\n", __func__, ptep,
			 pte_val(pfn_pte(pfn, hugeprot)));
		set_pte_at(mm, addr, ptep, pfn_pte(pfn, hugeprot));
		ptep++;
		pfn += pgsize >> PAGE_SHIFT;
		addr += pgsize;
	}
}

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

	pr_debug("%s: addr:0x%lx sz:0x%lx\n", __func__, addr, sz);
	pgd = pgd_offset(mm, addr);
	pud = pud_alloc(mm, pgd, addr);
	if (!pud)
		return NULL;

	if (sz == PUD_SIZE) {
		pte = (pte_t *)pud;
	} else if (sz == (PAGE_SIZE * CONT_PTES)) {
		pmd_t *pmd = pmd_alloc(mm, pud, addr);

		WARN_ON(addr & (sz - 1));
		/*
		 * Note that if this code were ever ported to the
		 * 32-bit arm platform then it will cause trouble in
		 * the case where CONFIG_HIGHPTE is set, since there
		 * will be no pte_unmap() to correspond with this
		 * pte_alloc_map().
		 */
		pte = pte_alloc_map(mm, pmd, addr);
	} else if (sz == PMD_SIZE) {
		if (IS_ENABLED(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) &&
		    pud_none(*pud))
			pte = huge_pmd_share(mm, addr, pud);
		else
			pte = (pte_t *)pmd_alloc(mm, pud, addr);
	} else if (sz == (PMD_SIZE * CONT_PMDS)) {
		pmd_t *pmd;

		pmd = pmd_alloc(mm, pud, addr);
		WARN_ON(addr & (sz - 1));
		return (pte_t *)pmd;
	}

	pr_debug("%s: addr:0x%lx sz:0x%lx ret pte=%p/0x%llx\n", __func__, addr,
	       sz, pte, pte_val(*pte));
	return pte;
}

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

	pgd = pgd_offset(mm, addr);
	pr_debug("%s: addr:0x%lx pgd:%p\n", __func__, addr, pgd);
	if (!pgd_present(*pgd))
		return NULL;

	pud = pud_offset(pgd, addr);
	if (pud_none(*pud))
		return NULL;
	/* swap or huge page */
	if (!pud_present(*pud) || pud_huge(*pud))
		return (pte_t *)pud;
	/* table; check the next level */

	pmd = pmd_offset(pud, addr);
	if (pmd_none(*pmd))
		return NULL;
	if (!pmd_present(*pmd) || pmd_huge(*pmd))
		return (pte_t *)pmd;

	return NULL;
}

pte_t arch_make_huge_pte(pte_t entry, struct vm_area_struct *vma,
			 struct page *page, int writable)
{
	size_t pagesize = huge_page_size(hstate_vma(vma));

	if (pagesize == CONT_PTE_SIZE) {
		entry = pte_mkcont(entry);
	} else if (pagesize == CONT_PMD_SIZE) {
		entry = pmd_pte(pmd_mkcont(pte_pmd(entry)));
	} else if (pagesize != PUD_SIZE && pagesize != PMD_SIZE) {
		pr_warn("%s: unrecognized huge page size 0x%lx\n",
			__func__, pagesize);
	}
	return entry;
}

pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
			      unsigned long addr, pte_t *ptep)
{
	pte_t pte;

	if (pte_cont(*ptep)) {
		int ncontig, i;
		size_t pgsize;
		pte_t *cpte;
		bool is_dirty = false;

		cpte = huge_pte_offset(mm, addr);
		ncontig = find_num_contig(mm, addr, cpte, *cpte, &pgsize);
		/* save the 1st pte to return */
		pte = ptep_get_and_clear(mm, addr, cpte);
		for (i = 1, addr += pgsize; i < ncontig; ++i, addr += pgsize) {
			/*
			 * If HW_AFDBM is enabled, then the HW could
			 * turn on the dirty bit for any of the page
			 * in the set, so check them all.
			 */
			++cpte;
			if (pte_dirty(ptep_get_and_clear(mm, addr, cpte)))
				is_dirty = true;
		}
		if (is_dirty)
			return pte_mkdirty(pte);
		else
			return pte;
	} else {
		return ptep_get_and_clear(mm, addr, ptep);
	}
}

int huge_ptep_set_access_flags(struct vm_area_struct *vma,
			       unsigned long addr, pte_t *ptep,
			       pte_t pte, int dirty)
{
	pte_t *cpte;

	if (pte_cont(pte)) {
		int ncontig, i, changed = 0;
		size_t pgsize = 0;
		unsigned long pfn = pte_pfn(pte);
		/* Select all bits except the pfn */
		pgprot_t hugeprot =
			__pgprot(pte_val(pfn_pte(pfn, __pgprot(0))) ^
				 pte_val(pte));

		cpte = huge_pte_offset(vma->vm_mm, addr);
		pfn = pte_pfn(*cpte);
		ncontig = find_num_contig(vma->vm_mm, addr, cpte,
					  *cpte, &pgsize);
		for (i = 0; i < ncontig; ++i, ++cpte, addr += pgsize) {
			changed |= ptep_set_access_flags(vma, addr, cpte,
							pfn_pte(pfn,
								hugeprot),
							dirty);
			pfn += pgsize >> PAGE_SHIFT;
		}
		return changed;
	} else {
		return ptep_set_access_flags(vma, addr, ptep, pte, dirty);
	}
}

void huge_ptep_set_wrprotect(struct mm_struct *mm,
			     unsigned long addr, pte_t *ptep)
{
	if (pte_cont(*ptep)) {
		int ncontig, i;
		pte_t *cpte;
		size_t pgsize = 0;

		cpte = huge_pte_offset(mm, addr);
		ncontig = find_num_contig(mm, addr, cpte, *cpte, &pgsize);
		for (i = 0; i < ncontig; ++i, ++cpte, addr += pgsize)
			ptep_set_wrprotect(mm, addr, cpte);
	} else {
		ptep_set_wrprotect(mm, addr, ptep);
	}
}

void huge_ptep_clear_flush(struct vm_area_struct *vma,
			   unsigned long addr, pte_t *ptep)
{
	if (pte_cont(*ptep)) {
		int ncontig, i;
		pte_t *cpte;
		size_t pgsize = 0;

		cpte = huge_pte_offset(vma->vm_mm, addr);
		ncontig = find_num_contig(vma->vm_mm, addr, cpte,
					  *cpte, &pgsize);
		for (i = 0; i < ncontig; ++i, ++cpte, addr += pgsize)
			ptep_clear_flush(vma, addr, cpte);
	} else {
		ptep_clear_flush(vma, addr, ptep);
	}
}

static __init int setup_hugepagesz(char *opt)
{
	unsigned long ps = memparse(opt, &opt);

	if (ps == PMD_SIZE) {
		hugetlb_add_hstate(PMD_SHIFT - PAGE_SHIFT);
	} else if (ps == PUD_SIZE) {
		hugetlb_add_hstate(PUD_SHIFT - PAGE_SHIFT);
	} else if (ps == (PAGE_SIZE * CONT_PTES)) {
		hugetlb_add_hstate(CONT_PTE_SHIFT);
	} else if (ps == (PMD_SIZE * CONT_PMDS)) {
		hugetlb_add_hstate((PMD_SHIFT + CONT_PMD_SHIFT) - PAGE_SHIFT);
	} else {
		hugetlb_bad_size();
		pr_err("hugepagesz: Unsupported page size %lu K\n", ps >> 10);
		return 0;
	}
	return 1;
}
__setup("hugepagesz=", setup_hugepagesz);

#ifdef CONFIG_ARM64_64K_PAGES
static __init int add_default_hugepagesz(void)
{
	if (size_to_hstate(CONT_PTES * PAGE_SIZE) == NULL)
		hugetlb_add_hstate(CONT_PTE_SHIFT);
	return 0;
}
arch_initcall(add_default_hugepagesz);
#endif
Commit	Line	Data
084bd298 SC	1	/*
	2	* arch/arm64/mm/hugetlbpage.c
	3	*
	4	* Copyright (C) 2013 Linaro Ltd.
	5	*
	6	* Based on arch/x86/mm/hugetlbpage.c.
	7	*
	8	* This program is free software; you can redistribute it and/or modify
	9	* it under the terms of the GNU General Public License version 2 as
	10	* published by the Free Software Foundation.
	11	*
	12	* This program is distributed in the hope that it will be useful,
	13	* but WITHOUT ANY WARRANTY; without even the implied warranty of
	14	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
	15	* GNU General Public License for more details.
084bd298 SC	16	*/
	17
	18	#include <linux/init.h>
	19	#include <linux/fs.h>
	20	#include <linux/mm.h>
	21	#include <linux/hugetlb.h>
	22	#include <linux/pagemap.h>
	23	#include <linux/err.h>
	24	#include <linux/sysctl.h>
	25	#include <asm/mman.h>
	26	#include <asm/tlb.h>
	27	#include <asm/tlbflush.h>
	28	#include <asm/pgalloc.h>
	29
084bd298 SC	30	int pmd_huge(pmd_t pmd)
084bd298 SC	31	{
fd28f5d4	32	return pmd_val(pmd) && !(pmd_val(pmd) & PMD_TABLE_BIT);
084bd298 SC	33	}
	34
	35	int pud_huge(pud_t pud)
	36	{
4797ec2d	37	#ifndef __PAGETABLE_PMD_FOLDED
fd28f5d4	38	return pud_val(pud) && !(pud_val(pud) & PUD_TABLE_BIT);
4797ec2d MS	39	#else
	40	return 0;
	41	#endif
084bd298 SC	42	}
084bd298 SC	43
66b3923a DW	44	static int find_num_contig(struct mm_struct *mm, unsigned long addr,
	45	pte_t ptep, pte_t pte, size_t pgsize)
	46	{
	47	pgd_t *pgd = pgd_offset(mm, addr);
	48	pud_t *pud;
	49	pmd_t *pmd;
	50
	51	*pgsize = PAGE_SIZE;
	52	if (!pte_cont(pte))
	53	return 1;
66b3923a	54	pud = pud_offset(pgd, addr);
66b3923a	55	pmd = pmd_offset(pud, addr);
66b3923a DW	56	if ((pte_t *)pmd == ptep) {
	57	*pgsize = PMD_SIZE;
	58	return CONT_PMDS;
	59	}
	60	return CONT_PTES;
	61	}
	62
	63	void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
	64	pte_t *ptep, pte_t pte)
	65	{
	66	size_t pgsize;
	67	int i;
	68	int ncontig = find_num_contig(mm, addr, ptep, pte, &pgsize);
	69	unsigned long pfn;
	70	pgprot_t hugeprot;
	71
	72	if (ncontig == 1) {
	73	set_pte_at(mm, addr, ptep, pte);
	74	return;
	75	}
	76
	77	pfn = pte_pfn(pte);
	78	hugeprot = __pgprot(pte_val(pfn_pte(pfn, __pgprot(0))) ^ pte_val(pte));
	79	for (i = 0; i < ncontig; i++) {
	80	pr_debug("%s: set pte %p to 0x%llx\n", __func__, ptep,
	81	pte_val(pfn_pte(pfn, hugeprot)));
	82	set_pte_at(mm, addr, ptep, pfn_pte(pfn, hugeprot));
	83	ptep++;
	84	pfn += pgsize >> PAGE_SHIFT;
	85	addr += pgsize;
	86	}
	87	}
	88
	89	pte_t huge_pte_alloc(struct mm_struct mm,
	90	unsigned long addr, unsigned long sz)
	91	{
	92	pgd_t *pgd;
	93	pud_t *pud;
	94	pte_t *pte = NULL;
	95
	96	pr_debug("%s: addr:0x%lx sz:0x%lx\n", __func__, addr, sz);
	97	pgd = pgd_offset(mm, addr);
	98	pud = pud_alloc(mm, pgd, addr);
	99	if (!pud)
	100	return NULL;
	101
	102	if (sz == PUD_SIZE) {
	103	pte = (pte_t *)pud;
	104	} else if (sz == (PAGE_SIZE * CONT_PTES)) {
	105	pmd_t *pmd = pmd_alloc(mm, pud, addr);
	106
	107	WARN_ON(addr & (sz - 1));
	108	/*
	109	* Note that if this code were ever ported to the
	110	* 32-bit arm platform then it will cause trouble in
	111	* the case where CONFIG_HIGHPTE is set, since there
	112	* will be no pte_unmap() to correspond with this
	113	* pte_alloc_map().
	114	*/
3ed3a4f0	115	pte = pte_alloc_map(mm, pmd, addr);
66b3923a DW	116	} else if (sz == PMD_SIZE) {
	117	if (IS_ENABLED(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) &&
	118	pud_none(*pud))
	119	pte = huge_pmd_share(mm, addr, pud);
	120	else
	121	pte = (pte_t *)pmd_alloc(mm, pud, addr);
	122	} else if (sz == (PMD_SIZE * CONT_PMDS)) {
	123	pmd_t *pmd;
	124
	125	pmd = pmd_alloc(mm, pud, addr);
	126	WARN_ON(addr & (sz - 1));
	127	return (pte_t *)pmd;
	128	}
	129
	130	pr_debug("%s: addr:0x%lx sz:0x%lx ret pte=%p/0x%llx\n", __func__, addr,
	131	sz, pte, pte_val(*pte));
	132	return pte;
	133	}
	134
	135	pte_t huge_pte_offset(struct mm_struct mm, unsigned long addr)
	136	{
	137	pgd_t *pgd;
	138	pud_t *pud;
0f3c2bcf	139	pmd_t *pmd;
66b3923a DW	140
	141	pgd = pgd_offset(mm, addr);
	142	pr_debug("%s: addr:0x%lx pgd:%p\n", __func__, addr, pgd);
	143	if (!pgd_present(*pgd))
	144	return NULL;
0f3c2bcf	145
66b3923a	146	pud = pud_offset(pgd, addr);
0f3c2bcf	147	if (pud_none(*pud))
66b3923a	148	return NULL;
0f3c2bcf PA	149	/* swap or huge page */
0f3c2bcf PA	150	if (!pud_present(pud) \|\| pud_huge(pud))
66b3923a	151	return (pte_t *)pud;
0f3c2bcf PA	152	/* table; check the next level */
0f3c2bcf PA	153
66b3923a	154	pmd = pmd_offset(pud, addr);
0f3c2bcf	155	if (pmd_none(*pmd))
66b3923a	156	return NULL;
0f3c2bcf	157	if (!pmd_present(pmd) \|\| pmd_huge(pmd))
66b3923a	158	return (pte_t *)pmd;
0f3c2bcf	159
66b3923a DW	160	return NULL;
	161	}
	162
	163	pte_t arch_make_huge_pte(pte_t entry, struct vm_area_struct *vma,
	164	struct page *page, int writable)
	165	{
	166	size_t pagesize = huge_page_size(hstate_vma(vma));
	167
	168	if (pagesize == CONT_PTE_SIZE) {
	169	entry = pte_mkcont(entry);
	170	} else if (pagesize == CONT_PMD_SIZE) {
	171	entry = pmd_pte(pmd_mkcont(pte_pmd(entry)));
	172	} else if (pagesize != PUD_SIZE && pagesize != PMD_SIZE) {
	173	pr_warn("%s: unrecognized huge page size 0x%lx\n",
	174	__func__, pagesize);
	175	}
	176	return entry;
	177	}
	178
	179	pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
	180	unsigned long addr, pte_t *ptep)
	181	{
	182	pte_t pte;
	183
	184	if (pte_cont(*ptep)) {
	185	int ncontig, i;
	186	size_t pgsize;
	187	pte_t *cpte;
	188	bool is_dirty = false;
	189
	190	cpte = huge_pte_offset(mm, addr);
	191	ncontig = find_num_contig(mm, addr, cpte, *cpte, &pgsize);
	192	/* save the 1st pte to return */
	193	pte = ptep_get_and_clear(mm, addr, cpte);
0c2f0afe	194	for (i = 1, addr += pgsize; i < ncontig; ++i, addr += pgsize) {
66b3923a DW	195	/*
	196	* If HW_AFDBM is enabled, then the HW could
	197	* turn on the dirty bit for any of the page
	198	* in the set, so check them all.
	199	*/
	200	++cpte;
	201	if (pte_dirty(ptep_get_and_clear(mm, addr, cpte)))
	202	is_dirty = true;
	203	}
	204	if (is_dirty)
	205	return pte_mkdirty(pte);
	206	else
	207	return pte;
	208	} else {
	209	return ptep_get_and_clear(mm, addr, ptep);
	210	}
	211	}
	212
	213	int huge_ptep_set_access_flags(struct vm_area_struct *vma,
	214	unsigned long addr, pte_t *ptep,
	215	pte_t pte, int dirty)
	216	{
	217	pte_t *cpte;
	218
	219	if (pte_cont(pte)) {
	220	int ncontig, i, changed = 0;
	221	size_t pgsize = 0;
	222	unsigned long pfn = pte_pfn(pte);
	223	/* Select all bits except the pfn */
	224	pgprot_t hugeprot =
	225	__pgprot(pte_val(pfn_pte(pfn, __pgprot(0))) ^
	226	pte_val(pte));
	227
	228	cpte = huge_pte_offset(vma->vm_mm, addr);
	229	pfn = pte_pfn(*cpte);
	230	ncontig = find_num_contig(vma->vm_mm, addr, cpte,
	231	*cpte, &pgsize);
0c2f0afe	232	for (i = 0; i < ncontig; ++i, ++cpte, addr += pgsize) {
69d01234	233	changed \|= ptep_set_access_flags(vma, addr, cpte,
66b3923a DW	234	pfn_pte(pfn,
	235	hugeprot),
	236	dirty);
	237	pfn += pgsize >> PAGE_SHIFT;
	238	}
	239	return changed;
	240	} else {
	241	return ptep_set_access_flags(vma, addr, ptep, pte, dirty);
	242	}
	243	}
	244
	245	void huge_ptep_set_wrprotect(struct mm_struct *mm,
	246	unsigned long addr, pte_t *ptep)
	247	{
	248	if (pte_cont(*ptep)) {
	249	int ncontig, i;
	250	pte_t *cpte;
	251	size_t pgsize = 0;
	252
	253	cpte = huge_pte_offset(mm, addr);
	254	ncontig = find_num_contig(mm, addr, cpte, *cpte, &pgsize);
0c2f0afe	255	for (i = 0; i < ncontig; ++i, ++cpte, addr += pgsize)
66b3923a DW	256	ptep_set_wrprotect(mm, addr, cpte);
	257	} else {
	258	ptep_set_wrprotect(mm, addr, ptep);
	259	}
	260	}
	261
	262	void huge_ptep_clear_flush(struct vm_area_struct *vma,
	263	unsigned long addr, pte_t *ptep)
	264	{
	265	if (pte_cont(*ptep)) {
	266	int ncontig, i;
	267	pte_t *cpte;
	268	size_t pgsize = 0;
	269
	270	cpte = huge_pte_offset(vma->vm_mm, addr);
	271	ncontig = find_num_contig(vma->vm_mm, addr, cpte,
	272	*cpte, &pgsize);
0c2f0afe	273	for (i = 0; i < ncontig; ++i, ++cpte, addr += pgsize)
66b3923a DW	274	ptep_clear_flush(vma, addr, cpte);
	275	} else {
	276	ptep_clear_flush(vma, addr, ptep);
	277	}
	278	}
	279
084bd298 SC	280	static __init int setup_hugepagesz(char *opt)
	281	{
	282	unsigned long ps = memparse(opt, &opt);
66b3923a	283
084bd298 SC	284	if (ps == PMD_SIZE) {
	285	hugetlb_add_hstate(PMD_SHIFT - PAGE_SHIFT);
	286	} else if (ps == PUD_SIZE) {
	287	hugetlb_add_hstate(PUD_SHIFT - PAGE_SHIFT);
ab2e1b89 WD	288	} else if (ps == (PAGE_SIZE * CONT_PTES)) {
	289	hugetlb_add_hstate(CONT_PTE_SHIFT);
	290	} else if (ps == (PMD_SIZE * CONT_PMDS)) {
	291	hugetlb_add_hstate((PMD_SHIFT + CONT_PMD_SHIFT) - PAGE_SHIFT);
084bd298	292	} else {
d77e20ce	293	hugetlb_bad_size();
66b3923a	294	pr_err("hugepagesz: Unsupported page size %lu K\n", ps >> 10);
084bd298 SC	295	return 0;
	296	}
	297	return 1;
	298	}
	299	__setup("hugepagesz=", setup_hugepagesz);
ab2e1b89 WD	300
	301	#ifdef CONFIG_ARM64_64K_PAGES
	302	static __init int add_default_hugepagesz(void)
	303	{
	304	if (size_to_hstate(CONT_PTES * PAGE_SIZE) == NULL)
6ed0038d	305	hugetlb_add_hstate(CONT_PTE_SHIFT);
ab2e1b89 WD	306	return 0;
	307	}
	308	arch_initcall(add_default_hugepagesz);
	309	#endif