Donate to e Foundation | Murena handsets with /e/OS | Own a part of Murena! Learn more

Commit 7f4a2a97 authored by Mark Fasheh's avatar Mark Fasheh
Browse files

ocfs2: always unmap in ocfs2_data_convert_worker() 



Mmap-heavy clustered workloads were sometimes finding stale data on mmap
reads. The solution is to call unmap_mapping_range() on any down convert of
a data lock.

Signed-off-by: default avatarMark Fasheh <mark.fasheh@oracle.com>
parent 6c2aad05

fs/ocfs2/dlmglue.c

+9 −1
Original line number Original line Diff line number Diff line
@@ -2718,6 +2718,15 @@ static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres, 
       	inode = ocfs2_lock_res_inode(lockres);

       	inode = ocfs2_lock_res_inode(lockres);

	mapping = inode->i_mapping;

	mapping = inode->i_mapping;





	/*

	 * We need this before the filemap_fdatawrite() so that it can

	 * transfer the dirty bit from the PTE to the

	 * page. Unfortunately this means that even for EX->PR

	 * downconverts, we'll lose our mappings and have to build

	 * them up again.

	 */

	unmap_mapping_range(mapping, 0, 0, 0);



	if (filemap_fdatawrite(mapping)) {

	if (filemap_fdatawrite(mapping)) {

		mlog(ML_ERROR, "Could not sync inode %llu for downconvert!",

		mlog(ML_ERROR, "Could not sync inode %llu for downconvert!",

		     (unsigned long long)OCFS2_I(inode)->ip_blkno);

		     (unsigned long long)OCFS2_I(inode)->ip_blkno);
@@ -2725,7 +2734,6 @@ static int ocfs2_data_convert_worker(struct ocfs2_lock_res *lockres, 
	sync_mapping_buffers(mapping);

	sync_mapping_buffers(mapping);

	if (blocking == LKM_EXMODE) {

	if (blocking == LKM_EXMODE) {

		truncate_inode_pages(mapping, 0);

		truncate_inode_pages(mapping, 0);

		unmap_mapping_range(mapping, 0, 0, 0);

	} else {

	} else {

		/* We only need to wait on the I/O if we're not also

		/* We only need to wait on the I/O if we're not also

		 * truncating pages because truncate_inode_pages waits

		 * truncating pages because truncate_inode_pages waits