filemap_xip.c 11 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
/*
 *	linux/mm/filemap_xip.c
 *
 * Copyright (C) 2005 IBM Corporation
 * Author: Carsten Otte <cotte@de.ibm.com>
 *
 * derived from linux/mm/filemap.c - Copyright (C) Linus Torvalds
 *
 */

#include <linux/fs.h>
#include <linux/pagemap.h>
#include <linux/module.h>
#include <linux/uio.h>
#include <linux/rmap.h>
Andrea Arcangeli's avatar
Andrea Arcangeli committed
16
#include <linux/mmu_notifier.h>
Alexey Dobriyan's avatar
Alexey Dobriyan committed
17
#include <linux/sched.h>
18 19
#include <linux/seqlock.h>
#include <linux/mutex.h>
20
#include <asm/tlbflush.h>
21
#include <asm/io.h>
22

23 24 25 26
/*
 * We do use our own empty page to avoid interference with other users
 * of ZERO_PAGE(), such as /dev/zero
 */
27 28
static DEFINE_MUTEX(xip_sparse_mutex);
static seqcount_t xip_sparse_seq = SEQCNT_ZERO;
29 30
static struct page *__xip_sparse_page;

31
/* called under xip_sparse_mutex */
32 33 34
static struct page *xip_sparse_page(void)
{
	if (!__xip_sparse_page) {
35 36
		struct page *page = alloc_page(GFP_HIGHUSER | __GFP_ZERO);

37 38
		if (page)
			__xip_sparse_page = page;
39 40 41 42
	}
	return __xip_sparse_page;
}

43 44
/*
 * This is a file read routine for execute in place files, and uses
45
 * the mapping->a_ops->get_xip_mem() function for the actual low-level
46 47 48 49
 * stuff.
 *
 * Note the struct file* is not used at all.  It may be NULL.
 */
50
static ssize_t
51 52 53
do_xip_mapping_read(struct address_space *mapping,
		    struct file_ra_state *_ra,
		    struct file *filp,
54 55 56
		    char __user *buf,
		    size_t len,
		    loff_t *ppos)
57 58
{
	struct inode *inode = mapping->host;
59 60
	pgoff_t index, end_index;
	unsigned long offset;
61 62
	loff_t isize, pos;
	size_t copied = 0, error = 0;
63

64
	BUG_ON(!mapping->a_ops->get_xip_mem);
65

66 67 68
	pos = *ppos;
	index = pos >> PAGE_CACHE_SHIFT;
	offset = pos & ~PAGE_CACHE_MASK;
69 70 71 72 73 74

	isize = i_size_read(inode);
	if (!isize)
		goto out;

	end_index = (isize - 1) >> PAGE_CACHE_SHIFT;
75 76 77 78 79
	do {
		unsigned long nr, left;
		void *xip_mem;
		unsigned long xip_pfn;
		int zero = 0;
80 81 82 83 84 85 86 87 88 89 90 91

		/* nr is the maximum number of bytes to copy from this page */
		nr = PAGE_CACHE_SIZE;
		if (index >= end_index) {
			if (index > end_index)
				goto out;
			nr = ((isize - 1) & ~PAGE_CACHE_MASK) + 1;
			if (nr <= offset) {
				goto out;
			}
		}
		nr = nr - offset;
92 93
		if (nr > len - copied)
			nr = len - copied;
94

95 96 97 98
		error = mapping->a_ops->get_xip_mem(mapping, index, 0,
							&xip_mem, &xip_pfn);
		if (unlikely(error)) {
			if (error == -ENODATA) {
99
				/* sparse */
100 101
				zero = 1;
			} else
102
				goto out;
Carsten Otte's avatar
Carsten Otte committed
103
		}
104 105 106 107 108 109

		/* If users can be writing to this page using arbitrary
		 * virtual addresses, take care about potential aliasing
		 * before reading the page on the kernel side.
		 */
		if (mapping_writably_mapped(mapping))
110
			/* address based flush */ ;
111 112

		/*
113
		 * Ok, we have the mem, so now we can copy it to user space...
114 115 116 117 118 119 120
		 *
		 * The actor routine returns how many bytes were actually used..
		 * NOTE! This may not be the same as how much of a user buffer
		 * we filled up (we may be padding etc), so we can only update
		 * "pos" here (the actor routine has to update the user buffer
		 * pointers and the remaining count).
		 */
121 122 123 124
		if (!zero)
			left = __copy_to_user(buf+copied, xip_mem+offset, nr);
		else
			left = __clear_user(buf + copied, nr);
125

126 127 128 129
		if (left) {
			error = -EFAULT;
			goto out;
		}
130

131 132 133 134 135
		copied += (nr - left);
		offset += (nr - left);
		index += offset >> PAGE_CACHE_SHIFT;
		offset &= ~PAGE_CACHE_MASK;
	} while (copied < len);
136 137

out:
138
	*ppos = pos + copied;
139 140
	if (filp)
		file_accessed(filp);
141 142

	return (copied ? copied : error);
143 144 145
}

ssize_t
146
xip_file_read(struct file *filp, char __user *buf, size_t len, loff_t *ppos)
147
{
148 149
	if (!access_ok(VERIFY_WRITE, buf, len))
		return -EFAULT;
150

151 152
	return do_xip_mapping_read(filp->f_mapping, &filp->f_ra, filp,
			    buf, len, ppos);
153
}
154
EXPORT_SYMBOL_GPL(xip_file_read);
155 156 157 158 159 160

/*
 * __xip_unmap is invoked from xip_unmap and
 * xip_write
 *
 * This function walks all vmas of the address_space and unmaps the
161
 * __xip_sparse_page when found at pgoff.
162 163 164 165 166 167 168 169 170 171 172
 */
static void
__xip_unmap (struct address_space * mapping,
		     unsigned long pgoff)
{
	struct vm_area_struct *vma;
	struct mm_struct *mm;
	struct prio_tree_iter iter;
	unsigned long address;
	pte_t *pte;
	pte_t pteval;
173
	spinlock_t *ptl;
174
	struct page *page;
175 176 177 178
	unsigned count;
	int locked = 0;

	count = read_seqcount_begin(&xip_sparse_seq);
179

180 181 182 183
	page = __xip_sparse_page;
	if (!page)
		return;

184
retry:
185 186 187 188 189 190
	spin_lock(&mapping->i_mmap_lock);
	vma_prio_tree_foreach(vma, &iter, &mapping->i_mmap, pgoff, pgoff) {
		mm = vma->vm_mm;
		address = vma->vm_start +
			((pgoff - vma->vm_pgoff) << PAGE_SHIFT);
		BUG_ON(address < vma->vm_start || address >= vma->vm_end);
Nick Piggin's avatar
Nick Piggin committed
191
		pte = page_check_address(page, mm, address, &ptl, 1);
192
		if (pte) {
193
			/* Nuke the page table entry. */
194
			flush_cache_page(vma, address, pte_pfn(*pte));
Andrea Arcangeli's avatar
Andrea Arcangeli committed
195
			pteval = ptep_clear_flush_notify(vma, address, pte);
196
			page_remove_rmap(page);
Nick Piggin's avatar
Nick Piggin committed
197
			dec_mm_counter(mm, file_rss);
198
			BUG_ON(pte_dirty(pteval));
199
			pte_unmap_unlock(pte, ptl);
Nick Piggin's avatar
Nick Piggin committed
200
			page_cache_release(page);
201 202 203
		}
	}
	spin_unlock(&mapping->i_mmap_lock);
204 205 206 207 208 209 210 211

	if (locked) {
		mutex_unlock(&xip_sparse_mutex);
	} else if (read_seqcount_retry(&xip_sparse_seq, count)) {
		mutex_lock(&xip_sparse_mutex);
		locked = 1;
		goto retry;
	}
212 213 214
}

/*
215
 * xip_fault() is invoked via the vma operations vector for a
216 217
 * mapped memory region to read in file data during a page fault.
 *
218
 * This function is derived from filemap_fault, but used for execute in place
219
 */
220
static int xip_file_fault(struct vm_area_struct *vma, struct vm_fault *vmf)
221
{
222
	struct file *file = vma->vm_file;
223 224
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
225
	pgoff_t size;
226 227 228 229
	void *xip_mem;
	unsigned long xip_pfn;
	struct page *page;
	int error;
230

231
	/* XXX: are VM_FAULT_ codes OK? */
232
again:
233
	size = (i_size_read(inode) + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT;
Nick Piggin's avatar
Nick Piggin committed
234 235
	if (vmf->pgoff >= size)
		return VM_FAULT_SIGBUS;
236

237 238 239 240 241
	error = mapping->a_ops->get_xip_mem(mapping, vmf->pgoff, 0,
						&xip_mem, &xip_pfn);
	if (likely(!error))
		goto found;
	if (error != -ENODATA)
Nick Piggin's avatar
Nick Piggin committed
242
		return VM_FAULT_OOM;
243 244

	/* sparse block */
245 246
	if ((vma->vm_flags & (VM_WRITE | VM_MAYWRITE)) &&
	    (vma->vm_flags & (VM_SHARED | VM_MAYSHARE)) &&
247
	    (!(mapping->host->i_sb->s_flags & MS_RDONLY))) {
248 249
		int err;

250
		/* maybe shared writable, allocate new block */
251
		mutex_lock(&xip_sparse_mutex);
252 253
		error = mapping->a_ops->get_xip_mem(mapping, vmf->pgoff, 1,
							&xip_mem, &xip_pfn);
254
		mutex_unlock(&xip_sparse_mutex);
255
		if (error)
Nick Piggin's avatar
Nick Piggin committed
256
			return VM_FAULT_SIGBUS;
257
		/* unmap sparse mappings at pgoff from all other vmas */
Nick Piggin's avatar
Nick Piggin committed
258
		__xip_unmap(mapping, vmf->pgoff);
259 260 261 262 263 264 265 266

found:
		err = vm_insert_mixed(vma, (unsigned long)vmf->virtual_address,
							xip_pfn);
		if (err == -ENOMEM)
			return VM_FAULT_OOM;
		BUG_ON(err);
		return VM_FAULT_NOPAGE;
267
	} else {
268 269 270 271 272 273 274 275 276 277 278 279 280
		int err, ret = VM_FAULT_OOM;

		mutex_lock(&xip_sparse_mutex);
		write_seqcount_begin(&xip_sparse_seq);
		error = mapping->a_ops->get_xip_mem(mapping, vmf->pgoff, 0,
							&xip_mem, &xip_pfn);
		if (unlikely(!error)) {
			write_seqcount_end(&xip_sparse_seq);
			mutex_unlock(&xip_sparse_mutex);
			goto again;
		}
		if (error != -ENODATA)
			goto out;
281 282
		/* not shared and writable, use xip_sparse_page() */
		page = xip_sparse_page();
Nick Piggin's avatar
Nick Piggin committed
283
		if (!page)
284 285 286 287 288 289 290 291 292 293
			goto out;
		err = vm_insert_page(vma, (unsigned long)vmf->virtual_address,
							page);
		if (err == -ENOMEM)
			goto out;

		ret = VM_FAULT_NOPAGE;
out:
		write_seqcount_end(&xip_sparse_seq);
		mutex_unlock(&xip_sparse_mutex);
294

295
		return ret;
296
	}
297 298
}

299
static const struct vm_operations_struct xip_file_vm_ops = {
300
	.fault	= xip_file_fault,
301 302 303 304
};

int xip_file_mmap(struct file * file, struct vm_area_struct * vma)
{
305
	BUG_ON(!file->f_mapping->a_ops->get_xip_mem);
306 307 308

	file_accessed(file);
	vma->vm_ops = &xip_file_vm_ops;
309
	vma->vm_flags |= VM_CAN_NONLINEAR | VM_MIXEDMAP;
310 311 312 313 314
	return 0;
}
EXPORT_SYMBOL_GPL(xip_file_mmap);

static ssize_t
315 316
__xip_file_write(struct file *filp, const char __user *buf,
		  size_t count, loff_t pos, loff_t *ppos)
317
{
318
	struct address_space * mapping = filp->f_mapping;
319
	const struct address_space_operations *a_ops = mapping->a_ops;
320 321 322 323 324
	struct inode 	*inode = mapping->host;
	long		status = 0;
	size_t		bytes;
	ssize_t		written = 0;

325
	BUG_ON(!mapping->a_ops->get_xip_mem);
326 327 328 329 330

	do {
		unsigned long index;
		unsigned long offset;
		size_t copied;
331 332
		void *xip_mem;
		unsigned long xip_pfn;
333 334 335 336 337 338 339

		offset = (pos & (PAGE_CACHE_SIZE -1)); /* Within page */
		index = pos >> PAGE_CACHE_SHIFT;
		bytes = PAGE_CACHE_SIZE - offset;
		if (bytes > count)
			bytes = count;

340 341 342
		status = a_ops->get_xip_mem(mapping, index, 0,
						&xip_mem, &xip_pfn);
		if (status == -ENODATA) {
343
			/* we allocate a new page unmap it */
344
			mutex_lock(&xip_sparse_mutex);
345 346
			status = a_ops->get_xip_mem(mapping, index, 1,
							&xip_mem, &xip_pfn);
347
			mutex_unlock(&xip_sparse_mutex);
348
			if (!status)
349 350
				/* unmap page at pgoff from all other vmas */
				__xip_unmap(mapping, index);
351 352
		}

353
		if (status)
354 355
			break;

Nick Piggin's avatar
Nick Piggin committed
356
		copied = bytes -
357
			__copy_from_user_nocache(xip_mem + offset, buf, bytes);
Nick Piggin's avatar
Nick Piggin committed
358

359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377
		if (likely(copied > 0)) {
			status = copied;

			if (status >= 0) {
				written += status;
				count -= status;
				pos += status;
				buf += status;
			}
		}
		if (unlikely(copied != bytes))
			if (status >= 0)
				status = -EFAULT;
		if (status < 0)
			break;
	} while (count);
	*ppos = pos;
	/*
	 * No need to use i_size_read() here, the i_size
378
	 * cannot change under us because we hold i_mutex.
379 380 381 382 383 384 385 386 387
	 */
	if (pos > inode->i_size) {
		i_size_write(inode, pos);
		mark_inode_dirty(inode);
	}

	return written ? written : status;
}

388 389 390
ssize_t
xip_file_write(struct file *filp, const char __user *buf, size_t len,
	       loff_t *ppos)
391
{
392 393 394 395 396
	struct address_space *mapping = filp->f_mapping;
	struct inode *inode = mapping->host;
	size_t count;
	loff_t pos;
	ssize_t ret;
397

398
	mutex_lock(&inode->i_mutex);
399

400 401 402
	if (!access_ok(VERIFY_READ, buf, len)) {
		ret=-EFAULT;
		goto out_up;
403 404 405
	}

	pos = *ppos;
406
	count = len;
407 408 409

	vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);

410 411
	/* We can write back this queue in page reclaim */
	current->backing_dev_info = mapping->backing_dev_info;
412

413 414 415
	ret = generic_write_checks(filp, &pos, &count, S_ISBLK(inode->i_mode));
	if (ret)
		goto out_backing;
416
	if (count == 0)
417
		goto out_backing;
418

419
	ret = file_remove_suid(filp);
420 421
	if (ret)
		goto out_backing;
422

423
	file_update_time(filp);
424

425
	ret = __xip_file_write (filp, buf, count, pos, ppos);
426

427 428 429
 out_backing:
	current->backing_dev_info = NULL;
 out_up:
430
	mutex_unlock(&inode->i_mutex);
431 432
	return ret;
}
433
EXPORT_SYMBOL_GPL(xip_file_write);
434 435 436

/*
 * truncate a page used for execute in place
437
 * functionality is analog to block_truncate_page but does use get_xip_mem
438 439 440 441 442 443 444 445 446
 * to get the page instead of page cache
 */
int
xip_truncate_page(struct address_space *mapping, loff_t from)
{
	pgoff_t index = from >> PAGE_CACHE_SHIFT;
	unsigned offset = from & (PAGE_CACHE_SIZE-1);
	unsigned blocksize;
	unsigned length;
447 448 449
	void *xip_mem;
	unsigned long xip_pfn;
	int err;
450

451
	BUG_ON(!mapping->a_ops->get_xip_mem);
452 453 454 455 456 457 458 459 460 461

	blocksize = 1 << mapping->host->i_blkbits;
	length = offset & (blocksize - 1);

	/* Block boundary? Nothing to do */
	if (!length)
		return 0;

	length = blocksize - length;

462 463 464 465
	err = mapping->a_ops->get_xip_mem(mapping, index, 0,
						&xip_mem, &xip_pfn);
	if (unlikely(err)) {
		if (err == -ENODATA)
466 467
			/* Hole? No need to truncate */
			return 0;
468
		else
469
			return err;
Carsten Otte's avatar
Carsten Otte committed
470
	}
471
	memset(xip_mem + offset, 0, length);
472
	return 0;
473 474
}
EXPORT_SYMBOL_GPL(xip_truncate_page);