cluster: RHEL6 - gfs2_convert: mark buffer dirty when switching dirs from meta to data

Gitweb: http://git.fedorahosted.org/git/?p=cluster.git;a=commitdiff;h=b75e3df9469... Commit: b75e3df94696cdc41a52e4c2ccd7dfd44c9a9d52 Parent: 129b016470e49e7fdd40a6116285c976c88307b3 Author: Bob Peterson <rpeterso(a)redhat.com&gt; AuthorDate: Mon Dec 17 14:47:50 2012 -0600 Committer: Bob Peterson <rpeterso(a)redhat.com&gt; CommitterDate: Thu Jan 3 07:34:24 2013 -0700 gfs2_convert: mark buffer dirty when switching dirs from meta to data This patch changes function inode_renumber so that it properly marks the rgrp bitmap dirty after switching bits from "meta" to "data". This happens when directory leaf, hash table, eattr, etc. blocks come in from GFS1 as "meta" and need to be switched to GFS2 as "data". In GFS2, only dinodes get the "meta" designation. This conversion was taking place, but the code broke out of the loop before properly marking the buffer as modified. So if no other modifications were done to that bitmap, the bitmap change would be forgotten. rhbz#888053 --- gfs2/convert/gfs2_convert.c | 2 +- 1 files changed, 1 insertions(+), 1 deletions(-) diff --git a/gfs2/convert/gfs2_convert.c b/gfs2/convert/gfs2_convert.c index 2a72504..ee41117 100644 --- a/gfs2/convert/gfs2_convert.c +++ b/gfs2/convert/gfs2_convert.c @@ -1051,10 +1051,10 @@ static int inode_renumber(struct gfs2_sbd *sbp, uint64_t root_inode_addr, osi_li ~(0x03 << (GFS2_BIT_SIZE * byte_bit)); rgd->bh[blk]->b_data[buf_offset + bitmap_byte] |= (0x01 << (GFS2_BIT_SIZE * byte_bit)); + bmodified(rgd->bh[blk]); break; } bitmap_byte -= (sbp->bsize - buf_offset); - bmodified(rgd->bh[blk]); } } brelse(bh);