Hi! I have a doubt related to the behaviour of fsync(). I am tinkering with Ext4 on a slightly older linux kernel (5.5.10) and I observed that dirty flags (I_DIRTY_INODE and I_DIRTY_PAGES) maintained in vfs inode state (inode->i_state) do not get cleared after fsync(). I observed this behaviour using the following simple program: open(file); write(file); fsync(file); close(file); //I checked status of inode->i_state inside close() Since, both dirty data and metadata get synced to disk (ext4_sync_file()), shouldn't inode state be updated to not-dirty after fsync? Could someone please point out why this isn't done or whether my observation is incorrect? PS: I looked around the source code of fsync() in latest kernel (v6.13-rc3) and I think this behaviour still exists in latest kernel also.
