[deliverable/linux.git] / mm / msync.c

/*
 *	linux/mm/msync.c
 *
 * Copyright (C) 1994-1999  Linus Torvalds
 */

/*
 * The msync() system call.
 */
#include <linux/slab.h>
#include <linux/pagemap.h>
#include <linux/mm.h>
#include <linux/mman.h>
#include <linux/hugetlb.h>
#include <linux/syscalls.h>

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

/*
 * Called with mm->page_table_lock held to protect against other
 * threads/the swapper from ripping pte's out from under us.
 */

static void sync_pte_range(struct vm_area_struct *vma, pmd_t *pmd,
				unsigned long addr, unsigned long end)
{
	pte_t *pte;

	pte = pte_offset_map(pmd, addr);
	do {
		unsigned long pfn;
		struct page *page;

		if (!pte_present(*pte))
			continue;
		if (!pte_maybe_dirty(*pte))
			continue;
		pfn = pte_pfn(*pte);
		if (!pfn_valid(pfn))
			continue;
		page = pfn_to_page(pfn);
		if (PageReserved(page))
			continue;

		if (ptep_clear_flush_dirty(vma, addr, pte) ||
		    page_test_and_clear_dirty(page))
			set_page_dirty(page);
	} while (pte++, addr += PAGE_SIZE, addr != end);
	pte_unmap(pte - 1);
}

static inline void sync_pmd_range(struct vm_area_struct *vma, pud_t *pud,
				unsigned long addr, unsigned long end)
{
	pmd_t *pmd;
	unsigned long next;

	pmd = pmd_offset(pud, addr);
	do {
		next = pmd_addr_end(addr, end);
		if (pmd_none_or_clear_bad(pmd))
			continue;
		sync_pte_range(vma, pmd, addr, next);
	} while (pmd++, addr = next, addr != end);
}

static inline void sync_pud_range(struct vm_area_struct *vma, pgd_t *pgd,
				unsigned long addr, unsigned long end)
{
	pud_t *pud;
	unsigned long next;

	pud = pud_offset(pgd, addr);
	do {
		next = pud_addr_end(addr, end);
		if (pud_none_or_clear_bad(pud))
			continue;
		sync_pmd_range(vma, pud, addr, next);
	} while (pud++, addr = next, addr != end);
}

static void sync_page_range(struct vm_area_struct *vma,
				unsigned long addr, unsigned long end)
{
	struct mm_struct *mm = vma->vm_mm;
	pgd_t *pgd;
	unsigned long next;

	/* For hugepages we can't go walking the page table normally,
	 * but that's ok, hugetlbfs is memory based, so we don't need
	 * to do anything more on an msync() */
	if (is_vm_hugetlb_page(vma))
		return;

	BUG_ON(addr >= end);
	pgd = pgd_offset(mm, addr);
	flush_cache_range(vma, addr, end);
	spin_lock(&mm->page_table_lock);
	do {
		next = pgd_addr_end(addr, end);
		if (pgd_none_or_clear_bad(pgd))
			continue;
		sync_pud_range(vma, pgd, addr, next);
	} while (pgd++, addr = next, addr != end);
	spin_unlock(&mm->page_table_lock);
}

#ifdef CONFIG_PREEMPT
static inline void filemap_sync(struct vm_area_struct *vma,
				unsigned long addr, unsigned long end)
{
	const size_t chunk = 64 * 1024;	/* bytes */
	unsigned long next;

	do {
		next = addr + chunk;
		if (next > end || next < addr)
			next = end;
		sync_page_range(vma, addr, next);
		cond_resched();
	} while (addr = next, addr != end);
}
#else
static inline void filemap_sync(struct vm_area_struct *vma,
				unsigned long addr, unsigned long end)
{
	sync_page_range(vma, addr, end);
}
#endif

/*
 * MS_SYNC syncs the entire file - including mappings.
 *
 * MS_ASYNC does not start I/O (it used to, up to 2.5.67).  Instead, it just
 * marks the relevant pages dirty.  The application may now run fsync() to
 * write out the dirty pages and wait on the writeout and check the result.
 * Or the application may run fadvise(FADV_DONTNEED) against the fd to start
 * async writeout immediately.
 * So my _not_ starting I/O in MS_ASYNC we provide complete flexibility to
 * applications.
 */
static int msync_interval(struct vm_area_struct *vma,
			unsigned long addr, unsigned long end, int flags)
{
	int ret = 0;
	struct file *file = vma->vm_file;

	if ((flags & MS_INVALIDATE) && (vma->vm_flags & VM_LOCKED))
		return -EBUSY;

	if (file && (vma->vm_flags & VM_SHARED)) {
		filemap_sync(vma, addr, end);

		if (flags & MS_SYNC) {
			struct address_space *mapping = file->f_mapping;
			int err;

			ret = filemap_fdatawrite(mapping);
			if (file->f_op && file->f_op->fsync) {
				/*
				 * We don't take i_sem here because mmap_sem
				 * is already held.
				 */
				err = file->f_op->fsync(file,file->f_dentry,1);
				if (err && !ret)
					ret = err;
			}
			err = filemap_fdatawait(mapping);
			if (!ret)
				ret = err;
		}
	}
	return ret;
}

asmlinkage long sys_msync(unsigned long start, size_t len, int flags)
{
	unsigned long end;
	struct vm_area_struct *vma;
	int unmapped_error, error = -EINVAL;

	if (flags & MS_SYNC)
		current->flags |= PF_SYNCWRITE;

	down_read(&current->mm->mmap_sem);
	if (flags & ~(MS_ASYNC | MS_INVALIDATE | MS_SYNC))
		goto out;
	if (start & ~PAGE_MASK)
		goto out;
	if ((flags & MS_ASYNC) && (flags & MS_SYNC))
		goto out;
	error = -ENOMEM;
	len = (len + ~PAGE_MASK) & PAGE_MASK;
	end = start + len;
	if (end < start)
		goto out;
	error = 0;
	if (end == start)
		goto out;
	/*
	 * If the interval [start,end) covers some unmapped address ranges,
	 * just ignore them, but return -ENOMEM at the end.
	 */
	vma = find_vma(current->mm, start);
	unmapped_error = 0;
	for (;;) {
		/* Still start < end. */
		error = -ENOMEM;
		if (!vma)
			goto out;
		/* Here start < vma->vm_end. */
		if (start < vma->vm_start) {
			unmapped_error = -ENOMEM;
			start = vma->vm_start;
		}
		/* Here vma->vm_start <= start < vma->vm_end. */
		if (end <= vma->vm_end) {
			if (start < end) {
				error = msync_interval(vma, start, end, flags);
				if (error)
					goto out;
			}
			error = unmapped_error;
			goto out;
		}
		/* Here vma->vm_start <= start < vma->vm_end < end. */
		error = msync_interval(vma, start, vma->vm_end, flags);
		if (error)
			goto out;
		start = vma->vm_end;
		vma = vma->vm_next;
	}
out:
	up_read(&current->mm->mmap_sem);
	current->flags &= ~PF_SYNCWRITE;
	return error;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* linux/mm/msync.c
	3	*
	4	* Copyright (C) 1994-1999 Linus Torvalds
	5	*/
	6
	7	/*
	8	* The msync() system call.
	9	*/
	10	#include <linux/slab.h>
	11	#include <linux/pagemap.h>
	12	#include <linux/mm.h>
	13	#include <linux/mman.h>
	14	#include <linux/hugetlb.h>
	15	#include <linux/syscalls.h>
	16
	17	#include <asm/pgtable.h>
	18	#include <asm/tlbflush.h>
	19
	20	/*
	21	* Called with mm->page_table_lock held to protect against other
	22	* threads/the swapper from ripping pte's out from under us.
	23	*/
	24
	25	static void sync_pte_range(struct vm_area_struct vma, pmd_t pmd,
	26	unsigned long addr, unsigned long end)
	27	{
	28	pte_t *pte;
	29
	30	pte = pte_offset_map(pmd, addr);
	31	do {
	32	unsigned long pfn;
	33	struct page *page;
	34
	35	if (!pte_present(*pte))
	36	continue;
b4955ce3 AK	37	if (!pte_maybe_dirty(*pte))
b4955ce3 AK	38	continue;
1da177e4 LT	39	pfn = pte_pfn(*pte);
	40	if (!pfn_valid(pfn))
	41	continue;
	42	page = pfn_to_page(pfn);
	43	if (PageReserved(page))
	44	continue;
	45
	46	if (ptep_clear_flush_dirty(vma, addr, pte) \|\|
	47	page_test_and_clear_dirty(page))
	48	set_page_dirty(page);
	49	} while (pte++, addr += PAGE_SIZE, addr != end);
	50	pte_unmap(pte - 1);
	51	}
	52
	53	static inline void sync_pmd_range(struct vm_area_struct vma, pud_t pud,
	54	unsigned long addr, unsigned long end)
	55	{
	56	pmd_t *pmd;
	57	unsigned long next;
	58
	59	pmd = pmd_offset(pud, addr);
	60	do {
	61	next = pmd_addr_end(addr, end);
	62	if (pmd_none_or_clear_bad(pmd))
	63	continue;
	64	sync_pte_range(vma, pmd, addr, next);
	65	} while (pmd++, addr = next, addr != end);
	66	}
	67
	68	static inline void sync_pud_range(struct vm_area_struct vma, pgd_t pgd,
	69	unsigned long addr, unsigned long end)
	70	{
	71	pud_t *pud;
	72	unsigned long next;
	73
	74	pud = pud_offset(pgd, addr);
	75	do {
	76	next = pud_addr_end(addr, end);
	77	if (pud_none_or_clear_bad(pud))
	78	continue;
	79	sync_pmd_range(vma, pud, addr, next);
	80	} while (pud++, addr = next, addr != end);
	81	}
	82
	83	static void sync_page_range(struct vm_area_struct *vma,
	84	unsigned long addr, unsigned long end)
	85	{
	86	struct mm_struct *mm = vma->vm_mm;
	87	pgd_t *pgd;
	88	unsigned long next;
	89
	90	/* For hugepages we can't go walking the page table normally,
	91	* but that's ok, hugetlbfs is memory based, so we don't need
	92	* to do anything more on an msync() */
	93	if (is_vm_hugetlb_page(vma))
	94	return;
	95
	96	BUG_ON(addr >= end);
	97	pgd = pgd_offset(mm, addr);
	98	flush_cache_range(vma, addr, end);
	99	spin_lock(&mm->page_table_lock);
	100	do {
	101	next = pgd_addr_end(addr, end);
	102	if (pgd_none_or_clear_bad(pgd))
103	continue;
104	sync_pud_range(vma, pgd, addr, next);
105	} while (pgd++, addr = next, addr != end);
106	spin_unlock(&mm->page_table_lock);
107	}
108
109	#ifdef CONFIG_PREEMPT
110	static inline void filemap_sync(struct vm_area_struct *vma,
111	unsigned long addr, unsigned long end)
112	{
113	const size_t chunk = 64 * 1024; /* bytes */
114	unsigned long next;
115
116	do {
117	next = addr + chunk;
118	if (next > end \|\| next < addr)
119	next = end;
120	sync_page_range(vma, addr, next);
121	cond_resched();
122	} while (addr = next, addr != end);
123	}
124	#else
125	static inline void filemap_sync(struct vm_area_struct *vma,
126	unsigned long addr, unsigned long end)
127	{
128	sync_page_range(vma, addr, end);
129	}
130	#endif
131
132	/*
133	* MS_SYNC syncs the entire file - including mappings.
134	*
135	* MS_ASYNC does not start I/O (it used to, up to 2.5.67). Instead, it just
136	* marks the relevant pages dirty. The application may now run fsync() to
137	* write out the dirty pages and wait on the writeout and check the result.
138	* Or the application may run fadvise(FADV_DONTNEED) against the fd to start
139	* async writeout immediately.
140	* So my _not_ starting I/O in MS_ASYNC we provide complete flexibility to
141	* applications.
142	*/
143	static int msync_interval(struct vm_area_struct *vma,
144	unsigned long addr, unsigned long end, int flags)
145	{
146	int ret = 0;
147	struct file *file = vma->vm_file;
148
149	if ((flags & MS_INVALIDATE) && (vma->vm_flags & VM_LOCKED))
150	return -EBUSY;
151
152	if (file && (vma->vm_flags & VM_SHARED)) {
153	filemap_sync(vma, addr, end);
154
155	if (flags & MS_SYNC) {
156	struct address_space *mapping = file->f_mapping;
157	int err;
158
159	ret = filemap_fdatawrite(mapping);
160	if (file->f_op && file->f_op->fsync) {
161	/*
162	* We don't take i_sem here because mmap_sem
163	* is already held.
164	*/
165	err = file->f_op->fsync(file,file->f_dentry,1);
166	if (err && !ret)
167	ret = err;
168	}
169	err = filemap_fdatawait(mapping);
170	if (!ret)
171	ret = err;
172	}
173	}
174	return ret;
175	}
176
177	asmlinkage long sys_msync(unsigned long start, size_t len, int flags)
178	{
179	unsigned long end;
180	struct vm_area_struct *vma;
181	int unmapped_error, error = -EINVAL;
182
183	if (flags & MS_SYNC)
184	current->flags \|= PF_SYNCWRITE;
185
186	down_read(&current->mm->mmap_sem);
187	if (flags & ~(MS_ASYNC \| MS_INVALIDATE \| MS_SYNC))
188	goto out;
189	if (start & ~PAGE_MASK)
190	goto out;
191	if ((flags & MS_ASYNC) && (flags & MS_SYNC))
192	goto out;
193	error = -ENOMEM;
194	len = (len + ~PAGE_MASK) & PAGE_MASK;
195	end = start + len;
196	if (end < start)
197	goto out;
198	error = 0;
199	if (end == start)
200	goto out;
201	/*
202	* If the interval [start,end) covers some unmapped address ranges,
203	* just ignore them, but return -ENOMEM at the end.
204	*/
205	vma = find_vma(current->mm, start);
206	unmapped_error = 0;
207	for (;;) {
208	/* Still start < end. */
209	error = -ENOMEM;
210	if (!vma)
211	goto out;
212	/* Here start < vma->vm_end. */
213	if (start < vma->vm_start) {
214	unmapped_error = -ENOMEM;
215	start = vma->vm_start;
216	}
217	/* Here vma->vm_start <= start < vma->vm_end. */
218	if (end <= vma->vm_end) {
219	if (start < end) {
220	error = msync_interval(vma, start, end, flags);
221	if (error)
222	goto out;
223	}
224	error = unmapped_error;
225	goto out;
226	}
227	/* Here vma->vm_start <= start < vma->vm_end < end. */
228	error = msync_interval(vma, start, vma->vm_end, flags);
229	if (error)
230	goto out;
231	start = vma->vm_end;
232	vma = vma->vm_next;
233	}
234	out:
235	up_read(&current->mm->mmap_sem);
236	current->flags &= ~PF_SYNCWRITE;
237	return error;
238	}