file.c

linux 内核源代码

	}

	if (file && should_remove_suid(file->f_path.dentry)) {
		ret = __ocfs2_write_remove_suid(inode, di_bh);
		if (ret) {
			mlog_errno(ret);
			goto out_meta_unlock;
		}
	}

	down_write(&OCFS2_I(inode)->ip_alloc_sem);
	switch (cmd) {
	case OCFS2_IOC_RESVSP:
	case OCFS2_IOC_RESVSP64:
		/*
		 * This takes unsigned offsets, but the signed ones we
		 * pass have been checked against overflow above.
		 */
		ret = ocfs2_allocate_unwritten_extents(inode, sr->l_start,
						       sr->l_len);
		break;
	case OCFS2_IOC_UNRESVSP:
	case OCFS2_IOC_UNRESVSP64:
		ret = ocfs2_remove_inode_range(inode, di_bh, sr->l_start,
					       sr->l_len);
		break;
	default:
		ret = -EINVAL;
	}
	up_write(&OCFS2_I(inode)->ip_alloc_sem);
	if (ret) {
		mlog_errno(ret);
		goto out_meta_unlock;
	}

	/*
	 * We update c/mtime for these changes
	 */
	handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS);
	if (IS_ERR(handle)) {
		ret = PTR_ERR(handle);
		mlog_errno(ret);
		goto out_meta_unlock;
	}

	if (change_size && i_size_read(inode) < size)
		i_size_write(inode, size);

	inode->i_ctime = inode->i_mtime = CURRENT_TIME;
	ret = ocfs2_mark_inode_dirty(handle, inode, di_bh);
	if (ret < 0)
		mlog_errno(ret);

	ocfs2_commit_trans(osb, handle);

out_meta_unlock:
	brelse(di_bh);
	ocfs2_meta_unlock(inode, 1);
out_rw_unlock:
	ocfs2_rw_unlock(inode, 1);

	mutex_unlock(&inode->i_mutex);
out:
	return ret;
}

int ocfs2_change_file_space(struct file *file, unsigned int cmd,
			    struct ocfs2_space_resv *sr)
{
	struct inode *inode = file->f_path.dentry->d_inode;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);;

	if ((cmd == OCFS2_IOC_RESVSP || cmd == OCFS2_IOC_RESVSP64) &&
	    !ocfs2_writes_unwritten_extents(osb))
		return -ENOTTY;
	else if ((cmd == OCFS2_IOC_UNRESVSP || cmd == OCFS2_IOC_UNRESVSP64) &&
		 !ocfs2_sparse_alloc(osb))
		return -ENOTTY;

	if (!S_ISREG(inode->i_mode))
		return -EINVAL;

	if (!(file->f_mode & FMODE_WRITE))
		return -EBADF;

	return __ocfs2_change_file_space(file, inode, file->f_pos, cmd, sr, 0);
}

static long ocfs2_fallocate(struct inode *inode, int mode, loff_t offset,
			    loff_t len)
{
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
	struct ocfs2_space_resv sr;
	int change_size = 1;

	if (!ocfs2_writes_unwritten_extents(osb))
		return -EOPNOTSUPP;

	if (S_ISDIR(inode->i_mode))
		return -ENODEV;

	if (mode & FALLOC_FL_KEEP_SIZE)
		change_size = 0;

	sr.l_whence = 0;
	sr.l_start = (s64)offset;
	sr.l_len = (s64)len;

	return __ocfs2_change_file_space(NULL, inode, offset,
					 OCFS2_IOC_RESVSP64, &sr, change_size);
}

static int ocfs2_prepare_inode_for_write(struct dentry *dentry,
					 loff_t *ppos,
					 size_t count,
					 int appending,
					 int *direct_io)
{
	int ret = 0, meta_level = 0;
	struct inode *inode = dentry->d_inode;
	loff_t saved_pos, end;

	/* 
	 * We start with a read level meta lock and only jump to an ex
	 * if we need to make modifications here.
	 */
	for(;;) {
		ret = ocfs2_meta_lock(inode, NULL, meta_level);
		if (ret < 0) {
			meta_level = -1;
			mlog_errno(ret);
			goto out;
		}

		/* Clear suid / sgid if necessary. We do this here
		 * instead of later in the write path because
		 * remove_suid() calls ->setattr without any hint that
		 * we may have already done our cluster locking. Since
		 * ocfs2_setattr() *must* take cluster locks to
		 * proceeed, this will lead us to recursively lock the
		 * inode. There's also the dinode i_size state which
		 * can be lost via setattr during extending writes (we
		 * set inode->i_size at the end of a write. */
		if (should_remove_suid(dentry)) {
			if (meta_level == 0) {
				ocfs2_meta_unlock(inode, meta_level);
				meta_level = 1;
				continue;
			}

			ret = ocfs2_write_remove_suid(inode);
			if (ret < 0) {
				mlog_errno(ret);
				goto out_unlock;
			}
		}

		/* work on a copy of ppos until we're sure that we won't have
		 * to recalculate it due to relocking. */
		if (appending) {
			saved_pos = i_size_read(inode);
			mlog(0, "O_APPEND: inode->i_size=%llu\n", saved_pos);
		} else {
			saved_pos = *ppos;
		}

		end = saved_pos + count;

		/*
		 * Skip the O_DIRECT checks if we don't need
		 * them.
		 */
		if (!direct_io || !(*direct_io))
			break;

		/*
		 * There's no sane way to do direct writes to an inode
		 * with inline data.
		 */
		if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
			*direct_io = 0;
			break;
		}

		/*
		 * Allowing concurrent direct writes means
		 * i_size changes wouldn't be synchronized, so
		 * one node could wind up truncating another
		 * nodes writes.
		 */
		if (end > i_size_read(inode)) {
			*direct_io = 0;
			break;
		}

		/*
		 * We don't fill holes during direct io, so
		 * check for them here. If any are found, the
		 * caller will have to retake some cluster
		 * locks and initiate the io as buffered.
		 */
		ret = ocfs2_check_range_for_holes(inode, saved_pos, count);
		if (ret == 1) {
			*direct_io = 0;
			ret = 0;
		} else if (ret < 0)
			mlog_errno(ret);
		break;
	}

	if (appending)
		*ppos = saved_pos;

out_unlock:
	ocfs2_meta_unlock(inode, meta_level);

out:
	return ret;
}

static ssize_t ocfs2_file_aio_write(struct kiocb *iocb,
				    const struct iovec *iov,
				    unsigned long nr_segs,
				    loff_t pos)
{
	int ret, direct_io, appending, rw_level, have_alloc_sem  = 0;
	int can_do_direct;
	ssize_t written = 0;
	size_t ocount;		/* original count */
	size_t count;		/* after file limit checks */
	loff_t old_size, *ppos = &iocb->ki_pos;
	u32 old_clusters;
	struct file *file = iocb->ki_filp;
	struct inode *inode = file->f_path.dentry->d_inode;
	struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);

	mlog_entry("(0x%p, %u, '%.*s')\n", file,
		   (unsigned int)nr_segs,
		   file->f_path.dentry->d_name.len,
		   file->f_path.dentry->d_name.name);

	if (iocb->ki_left == 0)
		return 0;

	vfs_check_frozen(inode->i_sb, SB_FREEZE_WRITE);

	appending = file->f_flags & O_APPEND ? 1 : 0;
	direct_io = file->f_flags & O_DIRECT ? 1 : 0;

	mutex_lock(&inode->i_mutex);

relock:
	/* to match setattr's i_mutex -> i_alloc_sem -> rw_lock ordering */
	if (direct_io) {
		down_read(&inode->i_alloc_sem);
		have_alloc_sem = 1;
	}

	/* concurrent O_DIRECT writes are allowed */
	rw_level = !direct_io;
	ret = ocfs2_rw_lock(inode, rw_level);
	if (ret < 0) {
		mlog_errno(ret);
		goto out_sems;
	}

	can_do_direct = direct_io;
	ret = ocfs2_prepare_inode_for_write(file->f_path.dentry, ppos,
					    iocb->ki_left, appending,
					    &can_do_direct);
	if (ret < 0) {
		mlog_errno(ret);
		goto out;
	}

	/*
	 * We can't complete the direct I/O as requested, fall back to
	 * buffered I/O.
	 */
	if (direct_io && !can_do_direct) {
		ocfs2_rw_unlock(inode, rw_level);
		up_read(&inode->i_alloc_sem);

		have_alloc_sem = 0;
		rw_level = -1;

		direct_io = 0;
		goto relock;
	}

	/*
	 * To later detect whether a journal commit for sync writes is
	 * necessary, we sample i_size, and cluster count here.
	 */
	old_size = i_size_read(inode);
	old_clusters = OCFS2_I(inode)->ip_clusters;

	/* communicate with ocfs2_dio_end_io */
	ocfs2_iocb_set_rw_locked(iocb, rw_level);

	if (direct_io) {
		ret = generic_segment_checks(iov, &nr_segs, &ocount,
					     VERIFY_READ);
		if (ret)
			goto out_dio;

		ret = generic_write_checks(file, ppos, &count,
					   S_ISBLK(inode->i_mode));
		if (ret)
			goto out_dio;

		written = generic_file_direct_write(iocb, iov, &nr_segs, *ppos,
						    ppos, count, ocount);
		if (written < 0) {
			ret = written;
			goto out_dio;
		}
	} else {
		written = generic_file_aio_write_nolock(iocb, iov, nr_segs,
							*ppos);
	}

out_dio:
	/* buffered aio wouldn't have proper lock coverage today */
	BUG_ON(ret == -EIOCBQUEUED && !(file->f_flags & O_DIRECT));

	if ((file->f_flags & O_SYNC && !direct_io) || IS_SYNC(inode)) {
		/*
		 * The generic write paths have handled getting data
		 * to disk, but since we don't make use of the dirty
		 * inode list, a manual journal commit is necessary
		 * here.
		 */
		if (old_size != i_size_read(inode) ||
		    old_clusters != OCFS2_I(inode)->ip_clusters) {
			ret = journal_force_commit(osb->journal->j_journal);
			if (ret < 0)
				written = ret;
		}
	}

	/* 
	 * deep in g_f_a_w_n()->ocfs2_direct_IO we pass in a ocfs2_dio_end_io
	 * function pointer which is called when o_direct io completes so that
	 * it can unlock our rw lock.  (it's the clustered equivalent of
	 * i_alloc_sem; protects truncate from racing with pending ios).
	 * Unfortunately there are error cases which call end_io and others
	 * that don't.  so we don't have to unlock the rw_lock if either an
	 * async dio is going to do it in the future or an end_io after an
	 * error has already done it.
	 */
	if (ret == -EIOCBQUEUED || !ocfs2_iocb_is_rw_locked(iocb)) {
		rw_level = -1;
		have_alloc_sem = 0;
	}

out:
	if (rw_level != -1)
		ocfs2_rw_unlock(inode, rw_level);

out_sems:
	if (have_alloc_sem)
		up_read(&inode->i_alloc_sem);

	mutex_unlock(&inode->i_mutex);

	mlog_exit(ret);
	return written ? written : ret;
}

static ssize_t ocfs2_file_splice_write(struct pipe_inode_info *pipe,
				       struct file *out,
				       loff_t *ppos,
				       size_t len,
				       unsigned int flags)
{
	int ret;
	struct inode *inode = out->f_path.dentry->d_inode;

	mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", out, pipe,
		   (unsigned int)len,
		   out->f_path.dentry->d_name.len,
		   out->f_path.dentry->d_name.name);

	inode_double_lock(inode, pipe->inode);

	ret = ocfs2_rw_lock(inode, 1);
	if (ret < 0) {
		mlog_errno(ret);
		goto out;
	}

	ret = ocfs2_prepare_inode_for_write(out->f_path.dentry, ppos, len, 0,
					    NULL);
	if (ret < 0) {
		mlog_errno(ret);
		goto out_unlock;
	}

	ret = generic_file_splice_write_nolock(pipe, out, ppos, len, flags);

out_unlock:
	ocfs2_rw_unlock(inode, 1);
out:
	inode_double_unlock(inode, pipe->inode);

	mlog_exit(ret);
	return ret;
}

static ssize_t ocfs2_file_splice_read(struct file *in,
				      loff_t *ppos,
				      struct pipe_inode_info *pipe,
				      size_t len,
				      unsigned int flags)
{
	int ret = 0;
	struct inode *inode = in->f_path.dentry->d_inode;

	mlog_entry("(0x%p, 0x%p, %u, '%.*s')\n", in, pipe,
		   (unsigned int)len,
		   in->f_path.dentry->d_name.len,
		   in->f_path.dentry->d_name.name);

	/*
	 * See the comment in ocfs2_file_aio_read()
	 */
	ret = ocfs2_meta_lock(inode, NULL, 0);
	if (ret < 0) {
		mlog_errno(ret);
		goto bail;
	}
	ocfs2_meta_unlock(inode, 0);

	ret = generic_file_splice_read(in, ppos, pipe, len, flags);

bail:
	mlog_exit(ret);
	return ret;
}

static ssize_t ocfs2_file_aio_read(struct kiocb *iocb,
				   const struct iovec *iov,
				   unsigned long nr_segs,
				   loff_t pos)
{
	int ret = 0, rw_level = -1, have_alloc_sem = 0, lock_level = 0;
	struct file *filp = iocb->ki_filp;
	struct inode *inode = filp->f_path.dentry->d_inode;

	mlog_entry("(0x%p, %u, '%.*s')\n", filp,
		   (unsigned int)nr_segs,
		   filp->f_path.dentry->d_name.len,
		   filp->f_path.dentry->d_name.name);

	if (!inode) {
		ret = -EINVAL;
		mlog_errno(ret);
		goto bail;
	}

	/* 
	 * buffered reads protect themselves in ->readpage().  O_DIRECT reads
	 * need locks to protect pending reads from racing with truncate.
	 */
	if (filp->f_flags & O_DIRECT) {
		down_read(&inode->i_alloc_sem);
		have_alloc_sem = 1;

		ret = ocfs2_rw_lock(inode, 0);
		if (ret < 0) {
			mlog_errno(ret);
			goto bail;
		}
		rw_level = 0;
		/* communicate with ocfs2_dio_end_io */
		ocfs2_iocb_set_rw_locked(iocb, rw_level);
	}

	/*
	 * We're fine letting folks race truncates and extending
	 * writes with read across the cluster, just like they can
	 * locally. Hence no rw_lock during read.
	 * 
	 * Take and drop the meta data lock to update inode fields
	 * like i_size. This allows the checks down below
	 * generic_file_aio_read() a chance of actually working. 
	 */
	ret = ocfs2_meta_lock_atime(inode, filp->f_vfsmnt, &lock_level);
	if (ret < 0) {
		mlog_errno(ret);
		goto bail;
	}
	ocfs2_meta_unlock(inode, lock_level);

	ret = generic_file_aio_read(iocb, iov, nr_segs, iocb->ki_pos);
	if (ret == -EINVAL)
		mlog(ML_ERROR, "generic_file_aio_read returned -EINVAL\n");

	/* buffered aio wouldn't have proper lock coverage today */
	BUG_ON(ret == -EIOCBQUEUED && !(filp->f_flags & O_DIRECT));

	/* see ocfs2_file_aio_write */
	if (ret == -EIOCBQUEUED || !ocfs2_iocb_is_rw_locked(iocb)) {
		rw_level = -1;
		have_alloc_sem = 0;
	}

bail:
	if (have_alloc_sem)
		up_read(&inode->i_alloc_sem);
	if (rw_level != -1) 
		ocfs2_rw_unlock(inode, rw_level);
	mlog_exit(ret);

	return ret;
}

const struct inode_operations ocfs2_file_iops = {
	.setattr	= ocfs2_setattr,
	.getattr	= ocfs2_getattr,
	.permission	= ocfs2_permission,
	.fallocate	= ocfs2_fallocate,
};

const struct inode_operations ocfs2_special_file_iops = {
	.setattr	= ocfs2_setattr,
	.getattr	= ocfs2_getattr,
	.permission	= ocfs2_permission,
};

const struct file_operations ocfs2_fops = {
	.read		= do_sync_read,
	.write		= do_sync_write,
	.mmap		= ocfs2_mmap,
	.fsync		= ocfs2_sync_file,
	.release	= ocfs2_file_release,
	.open		= ocfs2_file_open,
	.aio_read	= ocfs2_file_aio_read,
	.aio_write	= ocfs2_file_aio_write,
	.ioctl		= ocfs2_ioctl,
#ifdef CONFIG_COMPAT
	.compat_ioctl   = ocfs2_compat_ioctl,
#endif
	.splice_read	= ocfs2_file_splice_read,
	.splice_write	= ocfs2_file_splice_write,
};

const struct file_operations ocfs2_dops = {
	.read		= generic_read_dir,
	.readdir	= ocfs2_readdir,
	.fsync		= ocfs2_sync_file,
	.ioctl		= ocfs2_ioctl,
#ifdef CONFIG_COMPAT
	.compat_ioctl   = ocfs2_compat_ioctl,
#endif
};