On Sat, 2025-05-10 at 05:31 +0000, 李扬韬 wrote: > Hi Slava, > > > If I understood correctly "ERROR: access time has changed for file1 > > after remount" means atime has been changed. > > In fact, it seems that it is not the atime that has been changed, but > the disk atime that has been not changed. > The inode in memory has a newer atime, but the atime is not updated > to the disk when write_inode is executed(hfs has no atime in disk > format). > > For ERROR: access time has changed for file1 after remount > > Before: > Access: 2025-05-09 14:05:40 > Modify: 2025-05-09 14:05:38 > Change: 2025-05-09 14:05:38 > > After umount&mount: > Access: 2025-05-09 14:05:38 <-- back to mtime > Modify: 2025-05-09 14:05:38 > Change: 2025-05-09 14:05:38 > > So we get inconsistent results for atime. > > Am I missing something? > This is was my point. We need to make the access time (atime) always the same as the modification time (mtime) because we cannot save the atime on disk. This is what I meant by atime == mtime. And this atime change is happening in HFS driver logic that needs to be corrected. Thanks, Slava.
