Commit bddff633 authored by Luis Henriques's avatar Luis Henriques Committed by Ilya Dryomov
Browse files

ceph: only allow punch hole mode in fallocate 

Current implementation of cephfs fallocate isn't correct as it doesn't
really reserve the space in the cluster, which means that a subsequent
call to a write may actually fail due to lack of space.  In fact, it is
currently possible to fallocate an amount space that is larger than the
free space in the cluster.  It has behaved this way since the initial
commit ad7a60de ("ceph: punch hole support").

Since there's no easy solution to fix this at the moment, this patch
simply removes support for all fallocate operations but
FALLOC_FL_PUNCH_HOLE (which implies FALLOC_FL_KEEP_SIZE).

Link: https://tracker.ceph.com/issues/36317


Signed-off-by: default avatarLuis Henriques <lhenriques@suse.com>
Reviewed-by: default avatar"Yan, Zheng" <zyan@redhat.com>
Signed-off-by: default avatarIlya Dryomov <idryomov@gmail.com>
parent fce7a974

fs/ceph/file.c

+9 −36
Original line number Diff line number Diff line
@@ -1728,7 +1728,6 @@ static long ceph_fallocate(struct file *file, int mode, 
	struct ceph_file_info *fi = file->private_data;

	struct inode *inode = file_inode(file);

	struct ceph_inode_info *ci = ceph_inode(inode);

	struct ceph_fs_client *fsc = ceph_inode_to_client(inode);

	struct ceph_cap_flush *prealloc_cf;

	int want, got = 0;

	int dirty;
@@ -1736,10 +1735,7 @@ static long ceph_fallocate(struct file *file, int mode, 
	loff_t endoff = 0;

	loff_t size;



	if ((offset + length) > max(i_size_read(inode), fsc->max_file_size))

		return -EFBIG;



	if (mode & ~(FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))

	if (mode != (FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE))

		return -EOPNOTSUPP;



	if (!S_ISREG(inode->i_mode))
@@ -1756,18 +1752,6 @@ static long ceph_fallocate(struct file *file, int mode, 
		goto unlock;

	}



	if (!(mode & (FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE)) &&

	    ceph_quota_is_max_bytes_exceeded(inode, offset + length)) {

		ret = -EDQUOT;

		goto unlock;

	}



	if (ceph_osdmap_flag(&fsc->client->osdc, CEPH_OSDMAP_FULL) &&

	    !(mode & FALLOC_FL_PUNCH_HOLE)) {

		ret = -ENOSPC;

		goto unlock;

	}



	if (ci->i_inline_version != CEPH_INLINE_NONE) {

		ret = ceph_uninline_data(file, NULL);

		if (ret < 0)
@@ -1775,12 +1759,12 @@ static long ceph_fallocate(struct file *file, int mode, 
	}



	size = i_size_read(inode);

	if (!(mode & FALLOC_FL_KEEP_SIZE)) {

		endoff = offset + length;

		ret = inode_newsize_ok(inode, endoff);

		if (ret)



	/* Are we punching a hole beyond EOF? */

	if (offset >= size)

		goto unlock;

	}

	if ((offset + length) > size)

		length = size - offset;



	if (fi->fmode & CEPH_FILE_MODE_LAZY)

		want = CEPH_CAP_FILE_BUFFER | CEPH_CAP_FILE_LAZYIO;
@@ -1791,16 +1775,8 @@ static long ceph_fallocate(struct file *file, int mode, 
	if (ret < 0)

		goto unlock;



	if (mode & FALLOC_FL_PUNCH_HOLE) {

		if (offset < size)

	ceph_zero_pagecache_range(inode, offset, length);

	ret = ceph_zero_objects(inode, offset, length);

	} else if (endoff > size) {

		truncate_pagecache_range(inode, size, -1);

		if (ceph_inode_set_size(inode, endoff))

			ceph_check_caps(ceph_inode(inode),

				CHECK_CAPS_AUTHONLY, NULL);

	}



	if (!ret) {

		spin_lock(&ci->i_ceph_lock);
@@ -1810,9 +1786,6 @@ static long ceph_fallocate(struct file *file, int mode, 
		spin_unlock(&ci->i_ceph_lock);

		if (dirty)

			__mark_inode_dirty(inode, dirty);

		if ((endoff > size) &&

		    ceph_quota_is_max_bytes_approaching(inode, endoff))

			ceph_check_caps(ci, CHECK_CAPS_NODELAY, NULL);

	}



	ceph_put_cap_refs(ci, got);

CRA Git | Maintained and supported by SUSTech CRA and CCSE