[linux-2.6-block.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;
	if (!pgd_present(*pgd)) {
		VM_BUG_ON(!pgd_present(*pgd));
		return 1;
	}
	pud = pud_offset(pgd, addr);
	if (!pud_present(*pud)) {
		VM_BUG_ON(!pud_present(*pud));
		return 1;
	}
	pmd = pmd_offset(pud, addr);
	if (!pmd_present(*pmd)) {
		VM_BUG_ON(!pmd_present(*pmd));
		return 1;
	}
	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 = NULL;
	pte_t *pte = NULL;

	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_present(*pud))
		return NULL;

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

	if (pte_cont(pmd_pte(*pmd))) {
		pmd = pmd_offset(
			pud, (addr & CONT_PMD_MASK));
		return (pte_t *)pmd;
	}
	if (pmd_huge(*pmd))
		return (pte_t *)pmd;
	pte = pte_offset_kernel(pmd, addr);
	if (pte_present(*pte) && pte_cont(*pte)) {
		pte = pte_offset_kernel(
			pmd, (addr & CONT_PTE_MASK));
		return pte;
	}
	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; i < ncontig; ++i) {
			/*
			 * 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) {
			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)
			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)
			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_PMD_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;
	54	if (!pgd_present(*pgd)) {
	55	VM_BUG_ON(!pgd_present(*pgd));
	56	return 1;
	57	}
	58	pud = pud_offset(pgd, addr);
	59	if (!pud_present(*pud)) {
	60	VM_BUG_ON(!pud_present(*pud));
	61	return 1;
	62	}
	63	pmd = pmd_offset(pud, addr);
	64	if (!pmd_present(*pmd)) {
	65	VM_BUG_ON(!pmd_present(*pmd));
	66	return 1;
	67	}
	68	if ((pte_t *)pmd == ptep) {
	69	*pgsize = PMD_SIZE;
	70	return CONT_PMDS;
	71	}
	72	return CONT_PTES;
	73	}
	74
	75	void set_huge_pte_at(struct mm_struct *mm, unsigned long addr,
	76	pte_t *ptep, pte_t pte)
	77	{
	78	size_t pgsize;
	79	int i;
	80	int ncontig = find_num_contig(mm, addr, ptep, pte, &pgsize);
	81	unsigned long pfn;
	82	pgprot_t hugeprot;
	83
	84	if (ncontig == 1) {
	85	set_pte_at(mm, addr, ptep, pte);
	86	return;
	87	}
	88
	89	pfn = pte_pfn(pte);
	90	hugeprot = __pgprot(pte_val(pfn_pte(pfn, __pgprot(0))) ^ pte_val(pte));
	91	for (i = 0; i < ncontig; i++) {
	92	pr_debug("%s: set pte %p to 0x%llx\n", __func__, ptep,
	93	pte_val(pfn_pte(pfn, hugeprot)));
	94	set_pte_at(mm, addr, ptep, pfn_pte(pfn, hugeprot));
	95	ptep++;
	96	pfn += pgsize >> PAGE_SHIFT;
	97	addr += pgsize;
	98	}
	99	}
	100
	101	pte_t huge_pte_alloc(struct mm_struct mm,
	102	unsigned long addr, unsigned long sz)
	103	{
	104	pgd_t *pgd;
	105	pud_t *pud;
	106	pte_t *pte = NULL;
	107
108	pr_debug("%s: addr:0x%lx sz:0x%lx\n", __func__, addr, sz);
109	pgd = pgd_offset(mm, addr);
110	pud = pud_alloc(mm, pgd, addr);
111	if (!pud)
112	return NULL;
113
114	if (sz == PUD_SIZE) {
115	pte = (pte_t *)pud;
116	} else if (sz == (PAGE_SIZE * CONT_PTES)) {
117	pmd_t *pmd = pmd_alloc(mm, pud, addr);
118
119	WARN_ON(addr & (sz - 1));
120	/*
121	* Note that if this code were ever ported to the
122	* 32-bit arm platform then it will cause trouble in
123	* the case where CONFIG_HIGHPTE is set, since there
124	* will be no pte_unmap() to correspond with this
125	* pte_alloc_map().
126	*/
3ed3a4f0	127	pte = pte_alloc_map(mm, pmd, addr);
66b3923a DW	128	} else if (sz == PMD_SIZE) {
	129	if (IS_ENABLED(CONFIG_ARCH_WANT_HUGE_PMD_SHARE) &&
	130	pud_none(*pud))
	131	pte = huge_pmd_share(mm, addr, pud);
	132	else
	133	pte = (pte_t *)pmd_alloc(mm, pud, addr);
	134	} else if (sz == (PMD_SIZE * CONT_PMDS)) {
	135	pmd_t *pmd;
	136
	137	pmd = pmd_alloc(mm, pud, addr);
	138	WARN_ON(addr & (sz - 1));
	139	return (pte_t *)pmd;
	140	}
	141
	142	pr_debug("%s: addr:0x%lx sz:0x%lx ret pte=%p/0x%llx\n", __func__, addr,
	143	sz, pte, pte_val(*pte));
	144	return pte;
	145	}
	146
	147	pte_t huge_pte_offset(struct mm_struct mm, unsigned long addr)
	148	{
	149	pgd_t *pgd;
	150	pud_t *pud;
	151	pmd_t *pmd = NULL;
	152	pte_t *pte = NULL;
	153
	154	pgd = pgd_offset(mm, addr);
	155	pr_debug("%s: addr:0x%lx pgd:%p\n", __func__, addr, pgd);
	156	if (!pgd_present(*pgd))
	157	return NULL;
	158	pud = pud_offset(pgd, addr);
	159	if (!pud_present(*pud))
	160	return NULL;
	161
	162	if (pud_huge(*pud))
	163	return (pte_t *)pud;
	164	pmd = pmd_offset(pud, addr);
	165	if (!pmd_present(*pmd))
	166	return NULL;
	167
	168	if (pte_cont(pmd_pte(*pmd))) {
	169	pmd = pmd_offset(
	170	pud, (addr & CONT_PMD_MASK));
	171	return (pte_t *)pmd;
	172	}
	173	if (pmd_huge(*pmd))
	174	return (pte_t *)pmd;
	175	pte = pte_offset_kernel(pmd, addr);
	176	if (pte_present(pte) && pte_cont(pte)) {
	177	pte = pte_offset_kernel(
	178	pmd, (addr & CONT_PTE_MASK));
	179	return pte;
	180	}
	181	return NULL;
	182	}
	183
	184	pte_t arch_make_huge_pte(pte_t entry, struct vm_area_struct *vma,
	185	struct page *page, int writable)
	186	{
	187	size_t pagesize = huge_page_size(hstate_vma(vma));
	188
	189	if (pagesize == CONT_PTE_SIZE) {
	190	entry = pte_mkcont(entry);
	191	} else if (pagesize == CONT_PMD_SIZE) {
192	entry = pmd_pte(pmd_mkcont(pte_pmd(entry)));
193	} else if (pagesize != PUD_SIZE && pagesize != PMD_SIZE) {
194	pr_warn("%s: unrecognized huge page size 0x%lx\n",
195	__func__, pagesize);
196	}
197	return entry;
198	}
199
200	pte_t huge_ptep_get_and_clear(struct mm_struct *mm,
201	unsigned long addr, pte_t *ptep)
202	{
203	pte_t pte;
204
205	if (pte_cont(*ptep)) {
206	int ncontig, i;
207	size_t pgsize;
208	pte_t *cpte;
209	bool is_dirty = false;
210
211	cpte = huge_pte_offset(mm, addr);
212	ncontig = find_num_contig(mm, addr, cpte, *cpte, &pgsize);
213	/* save the 1st pte to return */
214	pte = ptep_get_and_clear(mm, addr, cpte);
215	for (i = 1; i < ncontig; ++i) {
216	/*
217	* If HW_AFDBM is enabled, then the HW could
218	* turn on the dirty bit for any of the page
219	* in the set, so check them all.
220	*/
221	++cpte;
222	if (pte_dirty(ptep_get_and_clear(mm, addr, cpte)))
223	is_dirty = true;
224	}
225	if (is_dirty)
226	return pte_mkdirty(pte);
227	else
228	return pte;
229	} else {
230	return ptep_get_and_clear(mm, addr, ptep);
231	}
232	}
233
234	int huge_ptep_set_access_flags(struct vm_area_struct *vma,
235	unsigned long addr, pte_t *ptep,
236	pte_t pte, int dirty)
237	{
238	pte_t *cpte;
239
240	if (pte_cont(pte)) {
241	int ncontig, i, changed = 0;
242	size_t pgsize = 0;
243	unsigned long pfn = pte_pfn(pte);
244	/* Select all bits except the pfn */
245	pgprot_t hugeprot =
246	__pgprot(pte_val(pfn_pte(pfn, __pgprot(0))) ^
247	pte_val(pte));
248
249	cpte = huge_pte_offset(vma->vm_mm, addr);
250	pfn = pte_pfn(*cpte);
251	ncontig = find_num_contig(vma->vm_mm, addr, cpte,
252	*cpte, &pgsize);
253	for (i = 0; i < ncontig; ++i, ++cpte) {
254	changed = ptep_set_access_flags(vma, addr, cpte,
255	pfn_pte(pfn,
256	hugeprot),
257	dirty);
258	pfn += pgsize >> PAGE_SHIFT;
259	}
260	return changed;
261	} else {
262	return ptep_set_access_flags(vma, addr, ptep, pte, dirty);
263	}
264	}
265
266	void huge_ptep_set_wrprotect(struct mm_struct *mm,
267	unsigned long addr, pte_t *ptep)
268	{
269	if (pte_cont(*ptep)) {
270	int ncontig, i;
271	pte_t *cpte;
272	size_t pgsize = 0;
273
274	cpte = huge_pte_offset(mm, addr);
275	ncontig = find_num_contig(mm, addr, cpte, *cpte, &pgsize);
276	for (i = 0; i < ncontig; ++i, ++cpte)
277	ptep_set_wrprotect(mm, addr, cpte);
278	} else {
279	ptep_set_wrprotect(mm, addr, ptep);
280	}
281	}
282
283	void huge_ptep_clear_flush(struct vm_area_struct *vma,
284	unsigned long addr, pte_t *ptep)
285	{
286	if (pte_cont(*ptep)) {
287	int ncontig, i;
288	pte_t *cpte;
289	size_t pgsize = 0;
290
291	cpte = huge_pte_offset(vma->vm_mm, addr);
292	ncontig = find_num_contig(vma->vm_mm, addr, cpte,
293	*cpte, &pgsize);
294	for (i = 0; i < ncontig; ++i, ++cpte)
295	ptep_clear_flush(vma, addr, cpte);
296	} else {
297	ptep_clear_flush(vma, addr, ptep);
298	}
299	}
300
084bd298 SC	301	static __init int setup_hugepagesz(char *opt)
	302	{
	303	unsigned long ps = memparse(opt, &opt);
66b3923a	304
084bd298 SC	305	if (ps == PMD_SIZE) {
	306	hugetlb_add_hstate(PMD_SHIFT - PAGE_SHIFT);
	307	} else if (ps == PUD_SIZE) {
	308	hugetlb_add_hstate(PUD_SHIFT - PAGE_SHIFT);
ab2e1b89 WD	309	} else if (ps == (PAGE_SIZE * CONT_PTES)) {
	310	hugetlb_add_hstate(CONT_PTE_SHIFT);
	311	} else if (ps == (PMD_SIZE * CONT_PMDS)) {
	312	hugetlb_add_hstate((PMD_SHIFT + CONT_PMD_SHIFT) - PAGE_SHIFT);
084bd298	313	} else {
d77e20ce	314	hugetlb_bad_size();
66b3923a	315	pr_err("hugepagesz: Unsupported page size %lu K\n", ps >> 10);
084bd298 SC	316	return 0;
	317	}
	318	return 1;
	319	}
	320	__setup("hugepagesz=", setup_hugepagesz);
ab2e1b89 WD	321
	322	#ifdef CONFIG_ARM64_64K_PAGES
	323	static __init int add_default_hugepagesz(void)
	324	{
	325	if (size_to_hstate(CONT_PTES * PAGE_SIZE) == NULL)
	326	hugetlb_add_hstate(CONT_PMD_SHIFT);
	327	return 0;
	328	}
	329	arch_initcall(add_default_hugepagesz);
	330	#endif