aops.c

linux 内核源代码

	if (unlikely(page->index >= (i_size + PAGE_CACHE_SIZE - 1) >>
			PAGE_CACHE_SHIFT)) {
		zero_user_page(page, 0, PAGE_CACHE_SIZE, KM_USER0);
		ntfs_debug("Read outside i_size - truncated?");
		goto done;
	}
	/*
	 * This can potentially happen because we clear PageUptodate() during
	 * ntfs_writepage() of MstProtected() attributes.
	 */
	if (PageUptodate(page)) {
		unlock_page(page);
		return 0;
	}
	ni = NTFS_I(vi);
	/*
	 * Only $DATA attributes can be encrypted and only unnamed $DATA
	 * attributes can be compressed.  Index root can have the flags set but
	 * this means to create compressed/encrypted files, not that the
	 * attribute is compressed/encrypted.  Note we need to check for
	 * AT_INDEX_ALLOCATION since this is the type of both directory and
	 * index inodes.
	 */
	if (ni->type != AT_INDEX_ALLOCATION) {
		/* If attribute is encrypted, deny access, just like NT4. */
		if (NInoEncrypted(ni)) {
			BUG_ON(ni->type != AT_DATA);
			err = -EACCES;
			goto err_out;
		}
		/* Compressed data streams are handled in compress.c. */
		if (NInoNonResident(ni) && NInoCompressed(ni)) {
			BUG_ON(ni->type != AT_DATA);
			BUG_ON(ni->name_len);
			return ntfs_read_compressed_block(page);
		}
	}
	/* NInoNonResident() == NInoIndexAllocPresent() */
	if (NInoNonResident(ni)) {
		/* Normal, non-resident data stream. */
		return ntfs_read_block(page);
	}
	/*
	 * Attribute is resident, implying it is not compressed or encrypted.
	 * This also means the attribute is smaller than an mft record and
	 * hence smaller than a page, so can simply zero out any pages with
	 * index above 0.  Note the attribute can actually be marked compressed
	 * but if it is resident the actual data is not compressed so we are
	 * ok to ignore the compressed flag here.
	 */
	if (unlikely(page->index > 0)) {
		zero_user_page(page, 0, PAGE_CACHE_SIZE, KM_USER0);
		goto done;
	}
	if (!NInoAttr(ni))
		base_ni = ni;
	else
		base_ni = ni->ext.base_ntfs_ino;
	/* Map, pin, and lock the mft record. */
	mrec = map_mft_record(base_ni);
	if (IS_ERR(mrec)) {
		err = PTR_ERR(mrec);
		goto err_out;
	}
	/*
	 * If a parallel write made the attribute non-resident, drop the mft
	 * record and retry the readpage.
	 */
	if (unlikely(NInoNonResident(ni))) {
		unmap_mft_record(base_ni);
		goto retry_readpage;
	}
	ctx = ntfs_attr_get_search_ctx(base_ni, mrec);
	if (unlikely(!ctx)) {
		err = -ENOMEM;
		goto unm_err_out;
	}
	err = ntfs_attr_lookup(ni->type, ni->name, ni->name_len,
			CASE_SENSITIVE, 0, NULL, 0, ctx);
	if (unlikely(err))
		goto put_unm_err_out;
	attr_len = le32_to_cpu(ctx->attr->data.resident.value_length);
	read_lock_irqsave(&ni->size_lock, flags);
	if (unlikely(attr_len > ni->initialized_size))
		attr_len = ni->initialized_size;
	i_size = i_size_read(vi);
	read_unlock_irqrestore(&ni->size_lock, flags);
	if (unlikely(attr_len > i_size)) {
		/* Race with shrinking truncate. */
		attr_len = i_size;
	}
	addr = kmap_atomic(page, KM_USER0);
	/* Copy the data to the page. */
	memcpy(addr, (u8*)ctx->attr +
			le16_to_cpu(ctx->attr->data.resident.value_offset),
			attr_len);
	/* Zero the remainder of the page. */
	memset(addr + attr_len, 0, PAGE_CACHE_SIZE - attr_len);
	flush_dcache_page(page);
	kunmap_atomic(addr, KM_USER0);
put_unm_err_out:
	ntfs_attr_put_search_ctx(ctx);
unm_err_out:
	unmap_mft_record(base_ni);
done:
	SetPageUptodate(page);
err_out:
	unlock_page(page);
	return err;
}

#ifdef NTFS_RW

/**
 * ntfs_write_block - write a @page to the backing store
 * @page:	page cache page to write out
 * @wbc:	writeback control structure
 *
 * This function is for writing pages belonging to non-resident, non-mst
 * protected attributes to their backing store.
 *
 * For a page with buffers, map and write the dirty buffers asynchronously
 * under page writeback. For a page without buffers, create buffers for the
 * page, then proceed as above.
 *
 * If a page doesn't have buffers the page dirty state is definitive. If a page
 * does have buffers, the page dirty state is just a hint, and the buffer dirty
 * state is definitive. (A hint which has rules: dirty buffers against a clean
 * page is illegal. Other combinations are legal and need to be handled. In
 * particular a dirty page containing clean buffers for example.)
 *
 * Return 0 on success and -errno on error.
 *
 * Based on ntfs_read_block() and __block_write_full_page().
 */
static int ntfs_write_block(struct page *page, struct writeback_control *wbc)
{
	VCN vcn;
	LCN lcn;
	s64 initialized_size;
	loff_t i_size;
	sector_t block, dblock, iblock;
	struct inode *vi;
	ntfs_inode *ni;
	ntfs_volume *vol;
	runlist_element *rl;
	struct buffer_head *bh, *head;
	unsigned long flags;
	unsigned int blocksize, vcn_ofs;
	int err;
	bool need_end_writeback;
	unsigned char blocksize_bits;

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

	ntfs_debug("Entering for inode 0x%lx, attribute type 0x%x, page index "
			"0x%lx.", ni->mft_no, ni->type, page->index);

	BUG_ON(!NInoNonResident(ni));
	BUG_ON(NInoMstProtected(ni));
	blocksize = vol->sb->s_blocksize;
	blocksize_bits = vol->sb->s_blocksize_bits;
	if (!page_has_buffers(page)) {
		BUG_ON(!PageUptodate(page));
		create_empty_buffers(page, blocksize,
				(1 << BH_Uptodate) | (1 << BH_Dirty));
		if (unlikely(!page_has_buffers(page))) {
			ntfs_warning(vol->sb, "Error allocating page "
					"buffers.  Redirtying page so we try "
					"again later.");
			/*
			 * Put the page back on mapping->dirty_pages, but leave
			 * its buffers' dirty state as-is.
			 */
			redirty_page_for_writepage(wbc, page);
			unlock_page(page);
			return 0;
		}
	}
	bh = head = page_buffers(page);
	BUG_ON(!bh);

	/* NOTE: Different naming scheme to ntfs_read_block()! */

	/* The first block in the page. */
	block = (s64)page->index << (PAGE_CACHE_SHIFT - blocksize_bits);

	read_lock_irqsave(&ni->size_lock, flags);
	i_size = i_size_read(vi);
	initialized_size = ni->initialized_size;
	read_unlock_irqrestore(&ni->size_lock, flags);

	/* The first out of bounds block for the data size. */
	dblock = (i_size + blocksize - 1) >> blocksize_bits;

	/* The last (fully or partially) initialized block. */
	iblock = initialized_size >> blocksize_bits;

	/*
	 * Be very careful.  We have no exclusion from __set_page_dirty_buffers
	 * here, and the (potentially unmapped) buffers may become dirty at
	 * any time.  If a buffer becomes dirty here after we've inspected it
	 * then we just miss that fact, and the page stays dirty.
	 *
	 * Buffers outside i_size may be dirtied by __set_page_dirty_buffers;
	 * handle that here by just cleaning them.
	 */

	/*
	 * Loop through all the buffers in the page, mapping all the dirty
	 * buffers to disk addresses and handling any aliases from the
	 * underlying block device's mapping.
	 */
	rl = NULL;
	err = 0;
	do {
		bool is_retry = false;

		if (unlikely(block >= dblock)) {
			/*
			 * Mapped buffers outside i_size will occur, because
			 * this page can be outside i_size when there is a
			 * truncate in progress. The contents of such buffers
			 * were zeroed by ntfs_writepage().
			 *
			 * FIXME: What about the small race window where
			 * ntfs_writepage() has not done any clearing because
			 * the page was within i_size but before we get here,
			 * vmtruncate() modifies i_size?
			 */
			clear_buffer_dirty(bh);
			set_buffer_uptodate(bh);
			continue;
		}

		/* Clean buffers are not written out, so no need to map them. */
		if (!buffer_dirty(bh))
			continue;

		/* Make sure we have enough initialized size. */
		if (unlikely((block >= iblock) &&
				(initialized_size < i_size))) {
			/*
			 * If this page is fully outside initialized size, zero
			 * out all pages between the current initialized size
			 * and the current page. Just use ntfs_readpage() to do
			 * the zeroing transparently.
			 */
			if (block > iblock) {
				// TODO:
				// For each page do:
				// - read_cache_page()
				// Again for each page do:
				// - wait_on_page_locked()
				// - Check (PageUptodate(page) &&
				//			!PageError(page))
				// Update initialized size in the attribute and
				// in the inode.
				// Again, for each page do:
				//	__set_page_dirty_buffers();
				// page_cache_release()
				// We don't need to wait on the writes.
				// Update iblock.
			}
			/*
			 * The current page straddles initialized size. Zero
			 * all non-uptodate buffers and set them uptodate (and
			 * dirty?). Note, there aren't any non-uptodate buffers
			 * if the page is uptodate.
			 * FIXME: For an uptodate page, the buffers may need to
			 * be written out because they were not initialized on
			 * disk before.
			 */
			if (!PageUptodate(page)) {
				// TODO:
				// Zero any non-uptodate buffers up to i_size.
				// Set them uptodate and dirty.
			}
			// TODO:
			// Update initialized size in the attribute and in the
			// inode (up to i_size).
			// Update iblock.
			// FIXME: This is inefficient. Try to batch the two
			// size changes to happen in one go.
			ntfs_error(vol->sb, "Writing beyond initialized size "
					"is not supported yet. Sorry.");
			err = -EOPNOTSUPP;
			break;
			// Do NOT set_buffer_new() BUT DO clear buffer range
			// outside write request range.
			// set_buffer_uptodate() on complete buffers as well as
			// set_buffer_dirty().
		}

		/* No need to map buffers that are already mapped. */
		if (buffer_mapped(bh))
			continue;

		/* Unmapped, dirty buffer. Need to map it. */
		bh->b_bdev = vol->sb->s_bdev;

		/* Convert block into corresponding vcn and offset. */
		vcn = (VCN)block << blocksize_bits;
		vcn_ofs = vcn & vol->cluster_size_mask;
		vcn >>= vol->cluster_size_bits;
		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 point to correct block. */
			bh->b_blocknr = ((lcn << vol->cluster_size_bits) +
					vcn_ofs) >> blocksize_bits;
			set_buffer_mapped(bh);
			continue;
		}
		/* It is a hole, need to instantiate it. */
		if (lcn == LCN_HOLE) {
			u8 *kaddr;
			unsigned long *bpos, *bend;

			/* Check if the buffer is zero. */
			kaddr = kmap_atomic(page, KM_USER0);
			bpos = (unsigned long *)(kaddr + bh_offset(bh));
			bend = (unsigned long *)((u8*)bpos + blocksize);
			do {
				if (unlikely(*bpos))
					break;
			} while (likely(++bpos < bend));
			kunmap_atomic(kaddr, KM_USER0);
			if (bpos == bend) {
				/*
				 * Buffer is zero and sparse, no need to write
				 * it.
				 */
				bh->b_blocknr = -1;
				clear_buffer_dirty(bh);
				continue;
			}
			// TODO: Instantiate the hole.
			// clear_buffer_new(bh);
			// unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
			ntfs_error(vol->sb, "Writing into sparse regions is "
					"not supported yet. Sorry.");
			err = -EOPNOTSUPP;
			break;
		}
		/* 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, truncate has cut it out
		 * of the runlist.  Just clean and clear the buffer and set it
		 * uptodate so it can get discarded by the VM.
		 */
		if (err == -ENOENT || lcn == LCN_ENOENT) {
			bh->b_blocknr = -1;
			clear_buffer_dirty(bh);
			zero_user_page(page, bh_offset(bh), blocksize,
					KM_USER0);
			set_buffer_uptodate(bh);
			err = 0;
			continue;
		}
		/* Failed to map the buffer, even after retrying. */
		if (!err)
			err = -EIO;
		bh->b_blocknr = -1;
		ntfs_error(vol->sb, "Failed to write to 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);
		break;
	} while (block++, (bh = bh->b_this_page) != head);

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

	/* For the error case, need to reset bh to the beginning. */
	bh = head;

	/* Just an optimization, so ->readpage() is not called later. */
	if (unlikely(!PageUptodate(page))) {
		int uptodate = 1;