aops.c

linux 内核源代码

/**
 * aops.c - NTFS kernel address space operations and page cache handling.
 *	    Part of the Linux-NTFS project.
 *
 * Copyright (c) 2001-2007 Anton Altaparmakov
 * Copyright (c) 2002 Richard Russon
 *
 * This program/include file 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.
 *
 * This program/include file 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.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program (in the main directory of the Linux-NTFS
 * distribution in the file COPYING); if not, write to the Free Software
 * Foundation,Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 */

#include <linux/errno.h>
#include <linux/fs.h>
#include <linux/mm.h>
#include <linux/pagemap.h>
#include <linux/swap.h>
#include <linux/buffer_head.h>
#include <linux/writeback.h>
#include <linux/bit_spinlock.h>

#include "aops.h"
#include "attrib.h"
#include "debug.h"
#include "inode.h"
#include "mft.h"
#include "runlist.h"
#include "types.h"
#include "ntfs.h"

/**
 * ntfs_end_buffer_async_read - async io completion for reading attributes
 * @bh:		buffer head on which io is completed
 * @uptodate:	whether @bh is now uptodate or not
 *
 * Asynchronous I/O completion handler for reading pages belonging to the
 * attribute address space of an inode.  The inodes can either be files or
 * directories or they can be fake inodes describing some attribute.
 *
 * If NInoMstProtected(), perform the post read mst fixups when all IO on the
 * page has been completed and mark the page uptodate or set the error bit on
 * the page.  To determine the size of the records that need fixing up, we
 * cheat a little bit by setting the index_block_size in ntfs_inode to the ntfs
 * record size, and index_block_size_bits, to the log(base 2) of the ntfs
 * record size.
 */
static void ntfs_end_buffer_async_read(struct buffer_head *bh, int uptodate)
{
	unsigned long flags;
	struct buffer_head *first, *tmp;
	struct page *page;
	struct inode *vi;
	ntfs_inode *ni;
	int page_uptodate = 1;

	page = bh->b_page;
	vi = page->mapping->host;
	ni = NTFS_I(vi);

	if (likely(uptodate)) {
		loff_t i_size;
		s64 file_ofs, init_size;

		set_buffer_uptodate(bh);

		file_ofs = ((s64)page->index << PAGE_CACHE_SHIFT) +
				bh_offset(bh);
		read_lock_irqsave(&ni->size_lock, flags);
		init_size = ni->initialized_size;
		i_size = i_size_read(vi);
		read_unlock_irqrestore(&ni->size_lock, flags);
		if (unlikely(init_size > i_size)) {
			/* Race with shrinking truncate. */
			init_size = i_size;
		}
		/* Check for the current buffer head overflowing. */
		if (unlikely(file_ofs + bh->b_size > init_size)) {
			int ofs;

			ofs = 0;
			if (file_ofs < init_size)
				ofs = init_size - file_ofs;
			local_irq_save(flags);
			zero_user_page(page, bh_offset(bh) + ofs,
					 bh->b_size - ofs, KM_BIO_SRC_IRQ);
			local_irq_restore(flags);
		}
	} else {
		clear_buffer_uptodate(bh);
		SetPageError(page);
		ntfs_error(ni->vol->sb, "Buffer I/O error, logical block "
				"0x%llx.", (unsigned long long)bh->b_blocknr);
	}
	first = page_buffers(page);
	local_irq_save(flags);
	bit_spin_lock(BH_Uptodate_Lock, &first->b_state);
	clear_buffer_async_read(bh);
	unlock_buffer(bh);
	tmp = bh;
	do {
		if (!buffer_uptodate(tmp))
			page_uptodate = 0;
		if (buffer_async_read(tmp)) {
			if (likely(buffer_locked(tmp)))
				goto still_busy;
			/* Async buffers must be locked. */
			BUG();
		}
		tmp = tmp->b_this_page;
	} while (tmp != bh);
	bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
	local_irq_restore(flags);
	/*
	 * If none of the buffers had errors then we can set the page uptodate,
	 * but we first have to perform the post read mst fixups, if the
	 * attribute is mst protected, i.e. if NInoMstProteced(ni) is true.
	 * Note we ignore fixup errors as those are detected when
	 * map_mft_record() is called which gives us per record granularity
	 * rather than per page granularity.
	 */
	if (!NInoMstProtected(ni)) {
		if (likely(page_uptodate && !PageError(page)))
			SetPageUptodate(page);
	} else {
		u8 *kaddr;
		unsigned int i, recs;
		u32 rec_size;

		rec_size = ni->itype.index.block_size;
		recs = PAGE_CACHE_SIZE / rec_size;
		/* Should have been verified before we got here... */
		BUG_ON(!recs);
		local_irq_save(flags);
		kaddr = kmap_atomic(page, KM_BIO_SRC_IRQ);
		for (i = 0; i < recs; i++)
			post_read_mst_fixup((NTFS_RECORD*)(kaddr +
					i * rec_size), rec_size);
		kunmap_atomic(kaddr, KM_BIO_SRC_IRQ);
		local_irq_restore(flags);
		flush_dcache_page(page);
		if (likely(page_uptodate && !PageError(page)))
			SetPageUptodate(page);
	}
	unlock_page(page);
	return;
still_busy:
	bit_spin_unlock(BH_Uptodate_Lock, &first->b_state);
	local_irq_restore(flags);
	return;
}

/**
 * ntfs_read_block - fill a @page of an address space with data
 * @page:	page cache page to fill with data
 *
 * Fill the page @page of the address space belonging to the @page->host inode.
 * We read each buffer asynchronously and when all buffers are read in, our io
 * completion handler ntfs_end_buffer_read_async(), if required, automatically
 * applies the mst fixups to the page before finally marking it uptodate and
 * unlocking it.
 *
 * We only enforce allocated_size limit because i_size is checked for in
 * generic_file_read().
 *
 * Return 0 on success and -errno on error.
 *
 * Contains an adapted version of fs/buffer.c::block_read_full_page().
 */
static int ntfs_read_block(struct page *page)
{
	loff_t i_size;
	VCN vcn;
	LCN lcn;
	s64 init_size;
	struct inode *vi;
	ntfs_inode *ni;
	ntfs_volume *vol;
	runlist_element *rl;
	struct buffer_head *bh, *head, *arr[MAX_BUF_PER_PAGE];
	sector_t iblock, lblock, zblock;
	unsigned long flags;
	unsigned int blocksize, vcn_ofs;
	int i, nr;
	unsigned char blocksize_bits;

	vi = page->mapping->host;
	ni = NTFS_I(vi);
	vol = ni->vol;

	/* $MFT/$DATA must have its complete runlist in memory at all times. */
	BUG_ON(!ni->runlist.rl && !ni->mft_no && !NInoAttr(ni));

	blocksize = vol->sb->s_blocksize;
	blocksize_bits = vol->sb->s_blocksize_bits;

	if (!page_has_buffers(page)) {
		create_empty_buffers(page, blocksize, 0);
		if (unlikely(!page_has_buffers(page))) {
			unlock_page(page);
			return -ENOMEM;
		}
	}
	bh = head = page_buffers(page);
	BUG_ON(!bh);

	/*
	 * We may be racing with truncate.  To avoid some of the problems we
	 * now take a snapshot of the various sizes and use those for the whole
	 * of the function.  In case of an extending truncate it just means we
	 * may leave some buffers unmapped which are now allocated.  This is
	 * not a problem since these buffers will just get mapped when a write
	 * occurs.  In case of a shrinking truncate, we will detect this later
	 * on due to the runlist being incomplete and if the page is being
	 * fully truncated, truncate will throw it away as soon as we unlock
	 * it so no need to worry what we do with it.
	 */
	iblock = (s64)page->index << (PAGE_CACHE_SHIFT - blocksize_bits);
	read_lock_irqsave(&ni->size_lock, flags);
	lblock = (ni->allocated_size + blocksize - 1) >> blocksize_bits;
	init_size = ni->initialized_size;
	i_size = i_size_read(vi);
	read_unlock_irqrestore(&ni->size_lock, flags);
	if (unlikely(init_size > i_size)) {
		/* Race with shrinking truncate. */
		init_size = i_size;
	}
	zblock = (init_size + blocksize - 1) >> blocksize_bits;

	/* Loop through all the buffers in the page. */
	rl = NULL;
	nr = i = 0;
	do {
		int err = 0;

		if (unlikely(buffer_uptodate(bh)))
			continue;
		if (unlikely(buffer_mapped(bh))) {
			arr[nr++] = bh;
			continue;
		}
		bh->b_bdev = vol->sb->s_bdev;
		/* Is the block within the allowed limits? */
		if (iblock < lblock) {
			bool is_retry = false;

			/* Convert iblock into corresponding vcn and offset. */
			vcn = (VCN)iblock << blocksize_bits >>
					vol->cluster_size_bits;
			vcn_ofs = ((VCN)iblock << blocksize_bits) &
					vol->cluster_size_mask;
			if (!rl) {
lock_retry_remap:
				down_read(&ni->runlist.lock);
				rl = ni->runlist.rl;
			}
			if (likely(rl != NULL)) {
				/* Seek to element containing target vcn. */
				while (rl->length && rl[1].vcn <= vcn)
					rl++;
				lcn = ntfs_rl_vcn_to_lcn(rl, vcn);
			} else
				lcn = LCN_RL_NOT_MAPPED;
			/* Successful remap. */
			if (lcn >= 0) {
				/* Setup buffer head to correct block. */
				bh->b_blocknr = ((lcn << vol->cluster_size_bits)
						+ vcn_ofs) >> blocksize_bits;
				set_buffer_mapped(bh);
				/* Only read initialized data blocks. */
				if (iblock < zblock) {
					arr[nr++] = bh;
					continue;
				}
				/* Fully non-initialized data block, zero it. */
				goto handle_zblock;
			}
			/* It is a hole, need to zero it. */
			if (lcn == LCN_HOLE)
				goto handle_hole;
			/* If first try and runlist unmapped, map and retry. */
			if (!is_retry && lcn == LCN_RL_NOT_MAPPED) {
				is_retry = true;
				/*
				 * Attempt to map runlist, dropping lock for
				 * the duration.
				 */
				up_read(&ni->runlist.lock);
				err = ntfs_map_runlist(ni, vcn);
				if (likely(!err))
					goto lock_retry_remap;
				rl = NULL;
			} else if (!rl)
				up_read(&ni->runlist.lock);
			/*
			 * If buffer is outside the runlist, treat it as a
			 * hole.  This can happen due to concurrent truncate
			 * for example.
			 */
			if (err == -ENOENT || lcn == LCN_ENOENT) {
				err = 0;
				goto handle_hole;
			}
			/* Hard error, zero out region. */
			if (!err)
				err = -EIO;
			bh->b_blocknr = -1;
			SetPageError(page);
			ntfs_error(vol->sb, "Failed to read from inode 0x%lx, "
					"attribute type 0x%x, vcn 0x%llx, "
					"offset 0x%x because its location on "
					"disk could not be determined%s "
					"(error code %i).", ni->mft_no,
					ni->type, (unsigned long long)vcn,
					vcn_ofs, is_retry ? " even after "
					"retrying" : "", err);
		}
		/*
		 * Either iblock was outside lblock limits or
		 * ntfs_rl_vcn_to_lcn() returned error.  Just zero that portion
		 * of the page and set the buffer uptodate.
		 */
handle_hole:
		bh->b_blocknr = -1UL;
		clear_buffer_mapped(bh);
handle_zblock:
		zero_user_page(page, i * blocksize, blocksize, KM_USER0);
		if (likely(!err))
			set_buffer_uptodate(bh);
	} while (i++, iblock++, (bh = bh->b_this_page) != head);

	/* Release the lock if we took it. */
	if (rl)
		up_read(&ni->runlist.lock);

	/* Check we have at least one buffer ready for i/o. */
	if (nr) {
		struct buffer_head *tbh;

		/* Lock the buffers. */
		for (i = 0; i < nr; i++) {
			tbh = arr[i];
			lock_buffer(tbh);
			tbh->b_end_io = ntfs_end_buffer_async_read;
			set_buffer_async_read(tbh);
		}
		/* Finally, start i/o on the buffers. */
		for (i = 0; i < nr; i++) {
			tbh = arr[i];
			if (likely(!buffer_uptodate(tbh)))
				submit_bh(READ, tbh);
			else
				ntfs_end_buffer_async_read(tbh, 1);
		}
		return 0;
	}
	/* No i/o was scheduled on any of the buffers. */
	if (likely(!PageError(page)))
		SetPageUptodate(page);
	else /* Signal synchronous i/o error. */
		nr = -EIO;
	unlock_page(page);
	return nr;
}

/**
 * ntfs_readpage - fill a @page of a @file with data from the device
 * @file:	open file to which the page @page belongs or NULL
 * @page:	page cache page to fill with data
 *
 * For non-resident attributes, ntfs_readpage() fills the @page of the open
 * file @file by calling the ntfs version of the generic block_read_full_page()
 * function, ntfs_read_block(), which in turn creates and reads in the buffers
 * associated with the page asynchronously.
 *
 * For resident attributes, OTOH, ntfs_readpage() fills @page by copying the
 * data from the mft record (which at this stage is most likely in memory) and
 * fills the remainder with zeroes. Thus, in this case, I/O is synchronous, as
 * even if the mft record is not cached at this point in time, we need to wait
 * for it to be read in before we can do the copy.
 *
 * Return 0 on success and -errno on error.
 */
static int ntfs_readpage(struct file *file, struct page *page)
{
	loff_t i_size;
	struct inode *vi;
	ntfs_inode *ni, *base_ni;
	u8 *addr;
	ntfs_attr_search_ctx *ctx;
	MFT_RECORD *mrec;
	unsigned long flags;
	u32 attr_len;
	int err = 0;

retry_readpage:
	BUG_ON(!PageLocked(page));
	vi = page->mapping->host;
	i_size = i_size_read(vi);
	/* Is the page fully outside i_size? (truncate in progress) */