Message ID | 20160424045551.GI20980@thunk.org |
---|---|
State | Awaiting Upstream, archived |
Headers | show |
On Sun 24-04-16 00:55:51, Ted Tso wrote: > I had to add a !IS_NOQUOTA check: > > diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c > index 576f64a..250c2df 100644 > --- a/fs/ext4/inode.c > +++ b/fs/ext4/inode.c > @@ -693,6 +693,7 @@ out_sem: > if (map->m_flags & EXT4_MAP_NEW && > !(map->m_flags & EXT4_MAP_UNWRITTEN) && > !(flags & EXT4_GET_BLOCKS_ZERO) && > + !IS_NOQUOTA(inode) && > ext4_should_order_data(inode)) { > ret = ext4_jbd2_file_inode(handle, inode); > if (ret) > > > In order to prevent crashes when writing to the quota file (see > below). Good catch. The fix looks correct to me, thanks for fixing this up. Since data for quota files is actually journalled, it may be actually a cleaner fix to reflect this in ext4_inode_journal_mode() and thus ext4_should_order_data() would catch this case. But this has more potential for breakage elsewhere so I will do that as a separate cleanup patch. Honza
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c index 576f64a..250c2df 100644 --- a/fs/ext4/inode.c +++ b/fs/ext4/inode.c @@ -693,6 +693,7 @@ out_sem: if (map->m_flags & EXT4_MAP_NEW && !(map->m_flags & EXT4_MAP_UNWRITTEN) && !(flags & EXT4_GET_BLOCKS_ZERO) && + !IS_NOQUOTA(inode) && ext4_should_order_data(inode)) { ret = ext4_jbd2_file_inode(handle, inode); if (ret)