file.c

linux 内核源代码

	initialized_size = ni->initialized_size;
	BUG_ON(end > ni->allocated_size);
	read_unlock_irqrestore(&ni->size_lock, flags);
	BUG_ON(initialized_size != i_size);
	if (end > initialized_size) {
		write_lock_irqsave(&ni->size_lock, flags);
		ni->initialized_size = end;
		i_size_write(vi, end);
		write_unlock_irqrestore(&ni->size_lock, flags);
	}
	/* Mark the mft record dirty, so it gets written back. */
	flush_dcache_mft_record_page(ctx->ntfs_ino);
	mark_mft_record_dirty(ctx->ntfs_ino);
	ntfs_attr_put_search_ctx(ctx);
	unmap_mft_record(base_ni);
	ntfs_debug("Done.");
	return 0;
err_out:
	if (err == -ENOMEM) {
		ntfs_warning(vi->i_sb, "Error allocating memory required to "
				"commit the write.");
		if (PageUptodate(page)) {
			ntfs_warning(vi->i_sb, "Page is uptodate, setting "
					"dirty so the write will be retried "
					"later on by the VM.");
			/*
			 * Put the page on mapping->dirty_pages, but leave its
			 * buffers' dirty state as-is.
			 */
			__set_page_dirty_nobuffers(page);
			err = 0;
		} else
			ntfs_error(vi->i_sb, "Page is not uptodate.  Written "
					"data has been lost.");
	} else {
		ntfs_error(vi->i_sb, "Resident attribute commit write failed "
				"with error %i.", err);
		NVolSetErrors(ni->vol);
	}
	if (ctx)
		ntfs_attr_put_search_ctx(ctx);
	if (m)
		unmap_mft_record(base_ni);
	return err;
}

/**
 * ntfs_file_buffered_write -
 *
 * Locking: The vfs is holding ->i_mutex on the inode.
 */
static ssize_t ntfs_file_buffered_write(struct kiocb *iocb,
		const struct iovec *iov, unsigned long nr_segs,
		loff_t pos, loff_t *ppos, size_t count)
{
	struct file *file = iocb->ki_filp;
	struct address_space *mapping = file->f_mapping;
	struct inode *vi = mapping->host;
	ntfs_inode *ni = NTFS_I(vi);
	ntfs_volume *vol = ni->vol;
	struct page *pages[NTFS_MAX_PAGES_PER_CLUSTER];
	struct page *cached_page = NULL;
	char __user *buf = NULL;
	s64 end, ll;
	VCN last_vcn;
	LCN lcn;
	unsigned long flags;
	size_t bytes, iov_ofs = 0;	/* Offset in the current iovec. */
	ssize_t status, written;
	unsigned nr_pages;
	int err;
	struct pagevec lru_pvec;

	ntfs_debug("Entering for i_ino 0x%lx, attribute type 0x%x, "
			"pos 0x%llx, count 0x%lx.",
			vi->i_ino, (unsigned)le32_to_cpu(ni->type),
			(unsigned long long)pos, (unsigned long)count);
	if (unlikely(!count))
		return 0;
	BUG_ON(NInoMstProtected(ni));
	/*
	 * If the attribute is not an index root and it is encrypted or
	 * compressed, we cannot write to it yet.  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 file is encrypted, deny access, just like NT4. */
		if (NInoEncrypted(ni)) {
			/*
			 * Reminder for later: Encrypted files are _always_
			 * non-resident so that the content can always be
			 * encrypted.
			 */
			ntfs_debug("Denying write access to encrypted file.");
			return -EACCES;
		}
		if (NInoCompressed(ni)) {
			/* Only unnamed $DATA attribute can be compressed. */
			BUG_ON(ni->type != AT_DATA);
			BUG_ON(ni->name_len);
			/*
			 * Reminder for later: If resident, the data is not
			 * actually compressed.  Only on the switch to non-
			 * resident does compression kick in.  This is in
			 * contrast to encrypted files (see above).
			 */
			ntfs_error(vi->i_sb, "Writing to compressed files is "
					"not implemented yet.  Sorry.");
			return -EOPNOTSUPP;
		}
	}
	/*
	 * If a previous ntfs_truncate() failed, repeat it and abort if it
	 * fails again.
	 */
	if (unlikely(NInoTruncateFailed(ni))) {
		down_write(&vi->i_alloc_sem);
		err = ntfs_truncate(vi);
		up_write(&vi->i_alloc_sem);
		if (err || NInoTruncateFailed(ni)) {
			if (!err)
				err = -EIO;
			ntfs_error(vol->sb, "Cannot perform write to inode "
					"0x%lx, attribute type 0x%x, because "
					"ntfs_truncate() failed (error code "
					"%i).", vi->i_ino,
					(unsigned)le32_to_cpu(ni->type), err);
			return err;
		}
	}
	/* The first byte after the write. */
	end = pos + count;
	/*
	 * If the write goes beyond the allocated size, extend the allocation
	 * to cover the whole of the write, rounded up to the nearest cluster.
	 */
	read_lock_irqsave(&ni->size_lock, flags);
	ll = ni->allocated_size;
	read_unlock_irqrestore(&ni->size_lock, flags);
	if (end > ll) {
		/* Extend the allocation without changing the data size. */
		ll = ntfs_attr_extend_allocation(ni, end, -1, pos);
		if (likely(ll >= 0)) {
			BUG_ON(pos >= ll);
			/* If the extension was partial truncate the write. */
			if (end > ll) {
				ntfs_debug("Truncating write to inode 0x%lx, "
						"attribute type 0x%x, because "
						"the allocation was only "
						"partially extended.",
						vi->i_ino, (unsigned)
						le32_to_cpu(ni->type));
				end = ll;
				count = ll - pos;
			}
		} else {
			err = ll;
			read_lock_irqsave(&ni->size_lock, flags);
			ll = ni->allocated_size;
			read_unlock_irqrestore(&ni->size_lock, flags);
			/* Perform a partial write if possible or fail. */
			if (pos < ll) {
				ntfs_debug("Truncating write to inode 0x%lx, "
						"attribute type 0x%x, because "
						"extending the allocation "
						"failed (error code %i).",
						vi->i_ino, (unsigned)
						le32_to_cpu(ni->type), err);
				end = ll;
				count = ll - pos;
			} else {
				ntfs_error(vol->sb, "Cannot perform write to "
						"inode 0x%lx, attribute type "
						"0x%x, because extending the "
						"allocation failed (error "
						"code %i).", vi->i_ino,
						(unsigned)
						le32_to_cpu(ni->type), err);
				return err;
			}
		}
	}
	pagevec_init(&lru_pvec, 0);
	written = 0;
	/*
	 * If the write starts beyond the initialized size, extend it up to the
	 * beginning of the write and initialize all non-sparse space between
	 * the old initialized size and the new one.  This automatically also
	 * increments the vfs inode->i_size to keep it above or equal to the
	 * initialized_size.
	 */
	read_lock_irqsave(&ni->size_lock, flags);
	ll = ni->initialized_size;
	read_unlock_irqrestore(&ni->size_lock, flags);
	if (pos > ll) {
		err = ntfs_attr_extend_initialized(ni, pos, &cached_page,
				&lru_pvec);
		if (err < 0) {
			ntfs_error(vol->sb, "Cannot perform write to inode "
					"0x%lx, attribute type 0x%x, because "
					"extending the initialized size "
					"failed (error code %i).", vi->i_ino,
					(unsigned)le32_to_cpu(ni->type), err);
			status = err;
			goto err_out;
		}
	}
	/*
	 * Determine the number of pages per cluster for non-resident
	 * attributes.
	 */
	nr_pages = 1;
	if (vol->cluster_size > PAGE_CACHE_SIZE && NInoNonResident(ni))
		nr_pages = vol->cluster_size >> PAGE_CACHE_SHIFT;
	/* Finally, perform the actual write. */
	last_vcn = -1;
	if (likely(nr_segs == 1))
		buf = iov->iov_base;
	do {
		VCN vcn;
		pgoff_t idx, start_idx;
		unsigned ofs, do_pages, u;
		size_t copied;

		start_idx = idx = pos >> PAGE_CACHE_SHIFT;
		ofs = pos & ~PAGE_CACHE_MASK;
		bytes = PAGE_CACHE_SIZE - ofs;
		do_pages = 1;
		if (nr_pages > 1) {
			vcn = pos >> vol->cluster_size_bits;
			if (vcn != last_vcn) {
				last_vcn = vcn;
				/*
				 * Get the lcn of the vcn the write is in.  If
				 * it is a hole, need to lock down all pages in
				 * the cluster.
				 */
				down_read(&ni->runlist.lock);
				lcn = ntfs_attr_vcn_to_lcn_nolock(ni, pos >>
						vol->cluster_size_bits, false);
				up_read(&ni->runlist.lock);
				if (unlikely(lcn < LCN_HOLE)) {
					status = -EIO;
					if (lcn == LCN_ENOMEM)
						status = -ENOMEM;
					else
						ntfs_error(vol->sb, "Cannot "
							"perform write to "
							"inode 0x%lx, "
							"attribute type 0x%x, "
							"because the attribute "
							"is corrupt.",
							vi->i_ino, (unsigned)
							le32_to_cpu(ni->type));
					break;
				}
				if (lcn == LCN_HOLE) {
					start_idx = (pos & ~(s64)
							vol->cluster_size_mask)
							>> PAGE_CACHE_SHIFT;
					bytes = vol->cluster_size - (pos &
							vol->cluster_size_mask);
					do_pages = nr_pages;
				}
			}
		}
		if (bytes > count)
			bytes = count;
		/*
		 * Bring in the user page(s) that we will copy from _first_.
		 * Otherwise there is a nasty deadlock on copying from the same
		 * page(s) as we are writing to, without it/them being marked
		 * up-to-date.  Note, at present there is nothing to stop the
		 * pages being swapped out between us bringing them into memory
		 * and doing the actual copying.
		 */
		if (likely(nr_segs == 1))
			ntfs_fault_in_pages_readable(buf, bytes);
		else
			ntfs_fault_in_pages_readable_iovec(iov, iov_ofs, bytes);
		/* Get and lock @do_pages starting at index @start_idx. */
		status = __ntfs_grab_cache_pages(mapping, start_idx, do_pages,
				pages, &cached_page, &lru_pvec);
		if (unlikely(status))
			break;
		/*
		 * For non-resident attributes, we need to fill any holes with
		 * actual clusters and ensure all bufferes are mapped.  We also
		 * need to bring uptodate any buffers that are only partially
		 * being written to.
		 */
		if (NInoNonResident(ni)) {
			status = ntfs_prepare_pages_for_non_resident_write(
					pages, do_pages, pos, bytes);
			if (unlikely(status)) {
				loff_t i_size;

				do {
					unlock_page(pages[--do_pages]);
					page_cache_release(pages[do_pages]);
				} while (do_pages);
				/*
				 * The write preparation may have instantiated
				 * allocated space outside i_size.  Trim this
				 * off again.  We can ignore any errors in this
				 * case as we will just be waisting a bit of
				 * allocated space, which is not a disaster.
				 */
				i_size = i_size_read(vi);
				if (pos + bytes > i_size)
					vmtruncate(vi, i_size);
				break;
			}
		}
		u = (pos >> PAGE_CACHE_SHIFT) - pages[0]->index;
		if (likely(nr_segs == 1)) {
			copied = ntfs_copy_from_user(pages + u, do_pages - u,
					ofs, buf, bytes);
			buf += copied;
		} else
			copied = ntfs_copy_from_user_iovec(pages + u,
					do_pages - u, ofs, &iov, &iov_ofs,
					bytes);
		ntfs_flush_dcache_pages(pages + u, do_pages - u);
		status = ntfs_commit_pages_after_write(pages, do_pages, pos,
				bytes);
		if (likely(!status)) {
			written += copied;
			count -= copied;
			pos += copied;
			if (unlikely(copied != bytes))
				status = -EFAULT;
		}
		do {
			unlock_page(pages[--do_pages]);
			mark_page_accessed(pages[do_pages]);
			page_cache_release(pages[do_pages]);
		} while (do_pages);
		if (unlikely(status))
			break;
		balance_dirty_pages_ratelimited(mapping);
		cond_resched();
	} while (count);
err_out:
	*ppos = pos;
	if (cached_page)
		page_cache_release(cached_page);
	/* For now, when the user asks for O_SYNC, we actually give O_DSYNC. */
	if (likely(!status)) {
		if (unlikely((file->f_flags & O_SYNC) || IS_SYNC(vi))) {
			if (!mapping->a_ops->writepage || !is_sync_kiocb(iocb))
				status = generic_osync_inode(vi, mapping,
						OSYNC_METADATA|OSYNC_DATA);
		}
  	}
	pagevec_lru_add(&lru_pvec);
	ntfs_debug("Done.  Returning %s (written 0x%lx, status %li).",
			written ? "written" : "status", (unsigned long)written,
			(long)status);
	return written ? written : status;
}

/**
 * ntfs_file_aio_write_nolock -
 */
static ssize_t ntfs_file_aio_write_nolock(struct kiocb *iocb,
		const struct iovec *iov, unsigned long nr_segs, loff_t *ppos)
{
	struct file *file = iocb->ki_filp;
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	loff_t pos;
	size_t count;		/* after file limit checks */
	ssize_t written, err;

	count = 0;
	err = generic_segment_checks(iov, &nr_segs, &count, VERIFY_READ);
	if (err)
		return err;
	pos = *ppos;
	vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);
	/* We can write back this queue in page reclaim. */
	current->backing_dev_info = mapping->backing_dev_info;
	written = 0;
	err = generic_write_checks(file, &pos, &count, S_ISBLK(inode->i_mode));
	if (err)
		goto out;
	if (!count)
		goto out;
	err = remove_suid(file->f_path.dentry);
	if (err)
		goto out;
	file_update_time(file);
	written = ntfs_file_buffered_write(iocb, iov, nr_segs, pos, ppos,
			count);
out:
	current->backing_dev_info = NULL;
	return written ? written : err;
}

/**
 * ntfs_file_aio_write -
 */
static ssize_t ntfs_file_aio_write(struct kiocb *iocb, const struct iovec *iov,
		unsigned long nr_segs, loff_t pos)
{
	struct file *file = iocb->ki_filp;
	struct address_space *mapping = file->f_mapping;
	struct inode *inode = mapping->host;
	ssize_t ret;

	BUG_ON(iocb->ki_pos != pos);

	mutex_lock(&inode->i_mutex);
	ret = ntfs_file_aio_write_nolock(iocb, iov, nr_segs, &iocb->ki_pos);
	mutex_unlock(&inode->i_mutex);
	if (ret > 0 && ((file->f_flags & O_SYNC) || IS_SYNC(inode))) {
		int err = sync_page_range(inode, mapping, pos, ret);
		if (err < 0)
			ret = err;
	}
	return ret;
}

/**
 * ntfs_file_writev -
 *
 * Basically the same as generic_file_writev() ex