[mirror_ubuntu-artful-kernel.git] / arch / powerpc / mm / subpage-prot.c

/*
 * Copyright 2007-2008 Paul Mackerras, IBM Corp.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public License
 * as published by the Free Software Foundation; either version
 * 2 of the License, or (at your option) any later version.
 */

#include <linux/errno.h>
#include <linux/kernel.h>
#include <linux/gfp.h>
#include <linux/types.h>
#include <linux/mm.h>
#include <linux/hugetlb.h>

#include <asm/pgtable.h>
#include <linux/uaccess.h>
#include <asm/tlbflush.h>

/*
 * Free all pages allocated for subpage protection maps and pointers.
 * Also makes sure that the subpage_prot_table structure is
 * reinitialized for the next user.
 */
void subpage_prot_free(struct mm_struct *mm)
{
	struct subpage_prot_table *spt = &mm->context.spt;
	unsigned long i, j, addr;
	u32 **p;

	for (i = 0; i < 4; ++i) {
		if (spt->low_prot[i]) {
			free_page((unsigned long)spt->low_prot[i]);
			spt->low_prot[i] = NULL;
		}
	}
	addr = 0;
	for (i = 0; i < (TASK_SIZE_USER64 >> 43); ++i) {
		p = spt->protptrs[i];
		if (!p)
			continue;
		spt->protptrs[i] = NULL;
		for (j = 0; j < SBP_L2_COUNT && addr < spt->maxaddr;
		     ++j, addr += PAGE_SIZE)
			if (p[j])
				free_page((unsigned long)p[j]);
		free_page((unsigned long)p);
	}
	spt->maxaddr = 0;
}

void subpage_prot_init_new_context(struct mm_struct *mm)
{
	struct subpage_prot_table *spt = &mm->context.spt;

	memset(spt, 0, sizeof(*spt));
}

static void hpte_flush_range(struct mm_struct *mm, unsigned long addr,
			     int npages)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *pte;
	spinlock_t *ptl;

	pgd = pgd_offset(mm, addr);
	if (pgd_none(*pgd))
		return;
	pud = pud_offset(pgd, addr);
	if (pud_none(*pud))
		return;
	pmd = pmd_offset(pud, addr);
	if (pmd_none(*pmd))
		return;
	pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	arch_enter_lazy_mmu_mode();
	for (; npages > 0; --npages) {
		pte_update(mm, addr, pte, 0, 0, 0);
		addr += PAGE_SIZE;
		++pte;
	}
	arch_leave_lazy_mmu_mode();
	pte_unmap_unlock(pte - 1, ptl);
}

/*
 * Clear the subpage protection map for an address range, allowing
 * all accesses that are allowed by the pte permissions.
 */
static void subpage_prot_clear(unsigned long addr, unsigned long len)
{
	struct mm_struct *mm = current->mm;
	struct subpage_prot_table *spt = &mm->context.spt;
	u32 **spm, *spp;
	unsigned long i;
	size_t nw;
	unsigned long next, limit;

	down_write(&mm->mmap_sem);
	limit = addr + len;
	if (limit > spt->maxaddr)
		limit = spt->maxaddr;
	for (; addr < limit; addr = next) {
		next = pmd_addr_end(addr, limit);
		if (addr < 0x100000000UL) {
			spm = spt->low_prot;
		} else {
			spm = spt->protptrs[addr >> SBP_L3_SHIFT];
			if (!spm)
				continue;
		}
		spp = spm[(addr >> SBP_L2_SHIFT) & (SBP_L2_COUNT - 1)];
		if (!spp)
			continue;
		spp += (addr >> PAGE_SHIFT) & (SBP_L1_COUNT - 1);

		i = (addr >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
		nw = PTRS_PER_PTE - i;
		if (addr + (nw << PAGE_SHIFT) > next)
			nw = (next - addr) >> PAGE_SHIFT;

		memset(spp, 0, nw * sizeof(u32));

		/* now flush any existing HPTEs for the range */
		hpte_flush_range(mm, addr, nw);
	}
	up_write(&mm->mmap_sem);
}

#ifdef CONFIG_TRANSPARENT_HUGEPAGE
static int subpage_walk_pmd_entry(pmd_t *pmd, unsigned long addr,
				  unsigned long end, struct mm_walk *walk)
{
	struct vm_area_struct *vma = walk->vma;
	split_huge_pmd(vma, pmd, addr);
	return 0;
}

static void subpage_mark_vma_nohuge(struct mm_struct *mm, unsigned long addr,
				    unsigned long len)
{
	struct vm_area_struct *vma;
	struct mm_walk subpage_proto_walk = {
		.mm = mm,
		.pmd_entry = subpage_walk_pmd_entry,
	};

	/*
	 * We don't try too hard, we just mark all the vma in that range
	 * VM_NOHUGEPAGE and split them.
	 */
	vma = find_vma(mm, addr);
	/*
	 * If the range is in unmapped range, just return
	 */
	if (vma && ((addr + len) <= vma->vm_start))
		return;

	while (vma) {
		if (vma->vm_start >= (addr + len))
			break;
		vma->vm_flags |= VM_NOHUGEPAGE;
		walk_page_vma(vma, &subpage_proto_walk);
		vma = vma->vm_next;
	}
}
#else
static void subpage_mark_vma_nohuge(struct mm_struct *mm, unsigned long addr,
				    unsigned long len)
{
	return;
}
#endif

/*
 * Copy in a subpage protection map for an address range.
 * The map has 2 bits per 4k subpage, so 32 bits per 64k page.
 * Each 2-bit field is 0 to allow any access, 1 to prevent writes,
 * 2 or 3 to prevent all accesses.
 * Note that the normal page protections also apply; the subpage
 * protection mechanism is an additional constraint, so putting 0
 * in a 2-bit field won't allow writes to a page that is otherwise
 * write-protected.
 */
long sys_subpage_prot(unsigned long addr, unsigned long len, u32 __user *map)
{
	struct mm_struct *mm = current->mm;
	struct subpage_prot_table *spt = &mm->context.spt;
	u32 **spm, *spp;
	unsigned long i;
	size_t nw;
	unsigned long next, limit;
	int err;

	/* Check parameters */
	if ((addr & ~PAGE_MASK) || (len & ~PAGE_MASK) ||
	    addr >= mm->task_size || len >= mm->task_size ||
	    addr + len > mm->task_size)
		return -EINVAL;

	if (is_hugepage_only_range(mm, addr, len))
		return -EINVAL;

	if (!map) {
		/* Clear out the protection map for the address range */
		subpage_prot_clear(addr, len);
		return 0;
	}

	if (!access_ok(VERIFY_READ, map, (len >> PAGE_SHIFT) * sizeof(u32)))
		return -EFAULT;

	down_write(&mm->mmap_sem);
	subpage_mark_vma_nohuge(mm, addr, len);
	for (limit = addr + len; addr < limit; addr = next) {
		next = pmd_addr_end(addr, limit);
		err = -ENOMEM;
		if (addr < 0x100000000UL) {
			spm = spt->low_prot;
		} else {
			spm = spt->protptrs[addr >> SBP_L3_SHIFT];
			if (!spm) {
				spm = (u32 **)get_zeroed_page(GFP_KERNEL);
				if (!spm)
					goto out;
				spt->protptrs[addr >> SBP_L3_SHIFT] = spm;
			}
		}
		spm += (addr >> SBP_L2_SHIFT) & (SBP_L2_COUNT - 1);
		spp = *spm;
		if (!spp) {
			spp = (u32 *)get_zeroed_page(GFP_KERNEL);
			if (!spp)
				goto out;
			*spm = spp;
		}
		spp += (addr >> PAGE_SHIFT) & (SBP_L1_COUNT - 1);

		local_irq_disable();
		demote_segment_4k(mm, addr);
		local_irq_enable();

		i = (addr >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
		nw = PTRS_PER_PTE - i;
		if (addr + (nw << PAGE_SHIFT) > next)
			nw = (next - addr) >> PAGE_SHIFT;

		up_write(&mm->mmap_sem);
		if (__copy_from_user(spp, map, nw * sizeof(u32)))
			return -EFAULT;
		map += nw;
		down_write(&mm->mmap_sem);

		/* now flush any existing HPTEs for the range */
		hpte_flush_range(mm, addr, nw);
	}
	if (limit > spt->maxaddr)
		spt->maxaddr = limit;
	err = 0;
 out:
	up_write(&mm->mmap_sem);
	return err;
}
Commit	Line	Data
fa28237c PM	1	/*
	2	* Copyright 2007-2008 Paul Mackerras, IBM Corp.
	3	*
	4	* This program is free software; you can redistribute it and/or
	5	* modify it under the terms of the GNU General Public License
	6	* as published by the Free Software Foundation; either version
	7	* 2 of the License, or (at your option) any later version.
	8	*/
	9
	10	#include <linux/errno.h>
	11	#include <linux/kernel.h>
	12	#include <linux/gfp.h>
fa28237c PM	13	#include <linux/types.h>
	14	#include <linux/mm.h>
	15	#include <linux/hugetlb.h>
	16
	17	#include <asm/pgtable.h>
7c0f6ba6	18	#include <linux/uaccess.h>
fa28237c PM	19	#include <asm/tlbflush.h>
	20
	21	/*
	22	* Free all pages allocated for subpage protection maps and pointers.
	23	* Also makes sure that the subpage_prot_table structure is
	24	* reinitialized for the next user.
	25	*/
d28513bc	26	void subpage_prot_free(struct mm_struct *mm)
fa28237c	27	{
d28513bc	28	struct subpage_prot_table *spt = &mm->context.spt;
fa28237c PM	29	unsigned long i, j, addr;
	30	u32 **p;
	31
	32	for (i = 0; i < 4; ++i) {
	33	if (spt->low_prot[i]) {
	34	free_page((unsigned long)spt->low_prot[i]);
	35	spt->low_prot[i] = NULL;
	36	}
	37	}
	38	addr = 0;
0da12a7a	39	for (i = 0; i < (TASK_SIZE_USER64 >> 43); ++i) {
fa28237c PM	40	p = spt->protptrs[i];
	41	if (!p)
	42	continue;
	43	spt->protptrs[i] = NULL;
	44	for (j = 0; j < SBP_L2_COUNT && addr < spt->maxaddr;
	45	++j, addr += PAGE_SIZE)
	46	if (p[j])
	47	free_page((unsigned long)p[j]);
	48	free_page((unsigned long)p);
	49	}
	50	spt->maxaddr = 0;
	51	}
	52
d28513bc DG	53	void subpage_prot_init_new_context(struct mm_struct *mm)
	54	{
	55	struct subpage_prot_table *spt = &mm->context.spt;
	56
	57	memset(spt, 0, sizeof(*spt));
	58	}
	59
fa28237c PM	60	static void hpte_flush_range(struct mm_struct *mm, unsigned long addr,
	61	int npages)
	62	{
	63	pgd_t *pgd;
	64	pud_t *pud;
	65	pmd_t *pmd;
	66	pte_t *pte;
	67	spinlock_t *ptl;
	68
	69	pgd = pgd_offset(mm, addr);
	70	if (pgd_none(*pgd))
	71	return;
	72	pud = pud_offset(pgd, addr);
	73	if (pud_none(*pud))
	74	return;
	75	pmd = pmd_offset(pud, addr);
	76	if (pmd_none(*pmd))
	77	return;
	78	pte = pte_offset_map_lock(mm, pmd, addr, &ptl);
	79	arch_enter_lazy_mmu_mode();
	80	for (; npages > 0; --npages) {
88247e8d	81	pte_update(mm, addr, pte, 0, 0, 0);
fa28237c PM	82	addr += PAGE_SIZE;
	83	++pte;
	84	}
	85	arch_leave_lazy_mmu_mode();
	86	pte_unmap_unlock(pte - 1, ptl);
	87	}
	88
	89	/*
	90	* Clear the subpage protection map for an address range, allowing
	91	* all accesses that are allowed by the pte permissions.
	92	*/
	93	static void subpage_prot_clear(unsigned long addr, unsigned long len)
	94	{
	95	struct mm_struct *mm = current->mm;
d28513bc	96	struct subpage_prot_table *spt = &mm->context.spt;
fa28237c	97	u32 *spm, spp;
6b5e7229 JM	98	unsigned long i;
6b5e7229 JM	99	size_t nw;
fa28237c PM	100	unsigned long next, limit;
	101
	102	down_write(&mm->mmap_sem);
	103	limit = addr + len;
	104	if (limit > spt->maxaddr)
	105	limit = spt->maxaddr;
	106	for (; addr < limit; addr = next) {
	107	next = pmd_addr_end(addr, limit);
b0d436c7	108	if (addr < 0x100000000UL) {
fa28237c PM	109	spm = spt->low_prot;
	110	} else {
	111	spm = spt->protptrs[addr >> SBP_L3_SHIFT];
	112	if (!spm)
	113	continue;
	114	}
	115	spp = spm[(addr >> SBP_L2_SHIFT) & (SBP_L2_COUNT - 1)];
	116	if (!spp)
	117	continue;
	118	spp += (addr >> PAGE_SHIFT) & (SBP_L1_COUNT - 1);
	119
	120	i = (addr >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
	121	nw = PTRS_PER_PTE - i;
	122	if (addr + (nw << PAGE_SHIFT) > next)
	123	nw = (next - addr) >> PAGE_SHIFT;
	124
	125	memset(spp, 0, nw * sizeof(u32));
	126
	127	/* now flush any existing HPTEs for the range */
	128	hpte_flush_range(mm, addr, nw);
	129	}
	130	up_write(&mm->mmap_sem);
	131	}
	132
d8e355a2 AK	133	#ifdef CONFIG_TRANSPARENT_HUGEPAGE
	134	static int subpage_walk_pmd_entry(pmd_t *pmd, unsigned long addr,
	135	unsigned long end, struct mm_walk *walk)
	136	{
1757bbd9	137	struct vm_area_struct *vma = walk->vma;
78ddc534	138	split_huge_pmd(vma, pmd, addr);
d8e355a2 AK	139	return 0;
	140	}
	141
	142	static void subpage_mark_vma_nohuge(struct mm_struct *mm, unsigned long addr,
	143	unsigned long len)
	144	{
	145	struct vm_area_struct *vma;
	146	struct mm_walk subpage_proto_walk = {
	147	.mm = mm,
	148	.pmd_entry = subpage_walk_pmd_entry,
	149	};
	150
	151	/*
	152	* We don't try too hard, we just mark all the vma in that range
	153	* VM_NOHUGEPAGE and split them.
	154	*/
	155	vma = find_vma(mm, addr);
	156	/*
	157	* If the range is in unmapped range, just return
	158	*/
	159	if (vma && ((addr + len) <= vma->vm_start))
	160	return;
	161
	162	while (vma) {
	163	if (vma->vm_start >= (addr + len))
	164	break;
	165	vma->vm_flags \|= VM_NOHUGEPAGE;
1757bbd9	166	walk_page_vma(vma, &subpage_proto_walk);
d8e355a2 AK	167	vma = vma->vm_next;
	168	}
	169	}
	170	#else
	171	static void subpage_mark_vma_nohuge(struct mm_struct *mm, unsigned long addr,
	172	unsigned long len)
	173	{
	174	return;
	175	}
	176	#endif
	177
fa28237c PM	178	/*
	179	* Copy in a subpage protection map for an address range.
	180	* The map has 2 bits per 4k subpage, so 32 bits per 64k page.
	181	* Each 2-bit field is 0 to allow any access, 1 to prevent writes,
	182	* 2 or 3 to prevent all accesses.
	183	* Note that the normal page protections also apply; the subpage
	184	* protection mechanism is an additional constraint, so putting 0
	185	* in a 2-bit field won't allow writes to a page that is otherwise
	186	* write-protected.
	187	*/
	188	long sys_subpage_prot(unsigned long addr, unsigned long len, u32 __user *map)
	189	{
	190	struct mm_struct *mm = current->mm;
d28513bc	191	struct subpage_prot_table *spt = &mm->context.spt;
fa28237c	192	u32 *spm, spp;
6b5e7229 JM	193	unsigned long i;
6b5e7229 JM	194	size_t nw;
fa28237c PM	195	unsigned long next, limit;
	196	int err;
	197
	198	/* Check parameters */
	199	if ((addr & ~PAGE_MASK) \|\| (len & ~PAGE_MASK) \|\|
be77e999 AK	200	addr >= mm->task_size \|\| len >= mm->task_size \|\|
be77e999 AK	201	addr + len > mm->task_size)
fa28237c PM	202	return -EINVAL;
	203
	204	if (is_hugepage_only_range(mm, addr, len))
	205	return -EINVAL;
	206
	207	if (!map) {
	208	/* Clear out the protection map for the address range */
	209	subpage_prot_clear(addr, len);
	210	return 0;
	211	}
	212
	213	if (!access_ok(VERIFY_READ, map, (len >> PAGE_SHIFT) * sizeof(u32)))
	214	return -EFAULT;
	215
	216	down_write(&mm->mmap_sem);
d8e355a2	217	subpage_mark_vma_nohuge(mm, addr, len);
fa28237c PM	218	for (limit = addr + len; addr < limit; addr = next) {
	219	next = pmd_addr_end(addr, limit);
	220	err = -ENOMEM;
b0d436c7	221	if (addr < 0x100000000UL) {
fa28237c PM	222	spm = spt->low_prot;
	223	} else {
	224	spm = spt->protptrs[addr >> SBP_L3_SHIFT];
	225	if (!spm) {
	226	spm = (u32 **)get_zeroed_page(GFP_KERNEL);
	227	if (!spm)
	228	goto out;
	229	spt->protptrs[addr >> SBP_L3_SHIFT] = spm;
	230	}
	231	}
	232	spm += (addr >> SBP_L2_SHIFT) & (SBP_L2_COUNT - 1);
	233	spp = *spm;
	234	if (!spp) {
	235	spp = (u32 *)get_zeroed_page(GFP_KERNEL);
	236	if (!spp)
	237	goto out;
	238	*spm = spp;
	239	}
	240	spp += (addr >> PAGE_SHIFT) & (SBP_L1_COUNT - 1);
	241
	242	local_irq_disable();
	243	demote_segment_4k(mm, addr);
	244	local_irq_enable();
	245
	246	i = (addr >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
	247	nw = PTRS_PER_PTE - i;
	248	if (addr + (nw << PAGE_SHIFT) > next)
	249	nw = (next - addr) >> PAGE_SHIFT;
	250
	251	up_write(&mm->mmap_sem);
fa28237c	252	if (__copy_from_user(spp, map, nw * sizeof(u32)))
a967f161	253	return -EFAULT;
fa28237c PM	254	map += nw;
	255	down_write(&mm->mmap_sem);
	256
	257	/* now flush any existing HPTEs for the range */
	258	hpte_flush_range(mm, addr, nw);
	259	}
	260	if (limit > spt->maxaddr)
	261	spt->maxaddr = limit;
	262	err = 0;
	263	out:
	264	up_write(&mm->mmap_sem);
fa28237c PM	265	return err;
fa28237c PM	266	}