Re: [PATCH v3 2/3] ntfs: sync the volume dirty bit with the recorded error state

[Hongling Zeng <zhongling0719@126.com>]

在 2026年09月08日 12:10, Namjae Jeon 写道:
>> @@ -1621,6 +1666,11 @@ static bool load_system_files(struct ntfs_volume *vol)
>>                          ntfs_error(sb, "%s.  Mounting read-only%s", es1, es2);
>>                  }
>>                  NVolSetErrors(vol);
>> +               /*
>> +                * Remember it for the lifetime of the mount: see
>> +                * ntfs_sync_volume_dirty_state().
>> +                */
>> +               NVolSetHibernated(vol);
> Could you explain why NVolSetHibernated(vol) is called for non-zero
> return from check_windows_hibernation_status() ?
Hi:
   A non-zero return does not only mean "definitely hibernated": > 0
   means hibernation was established or assumed (the documented
   "treated as if Windows is suspended" default), while < 0 means the
   check itself failed, so hibernation could not be ruled out.

   NV_Hibernated is a conservative write gate, not a positive claim: it
   keeps ntfs_sync_volume_dirty_state() from writing the $Volume record
   on this mount. It has to cover the whole branch because this series
   changes what NVolErrors() means for writes: it used to suppress the
   flags write, and now the on-disk dirty bit is derived from it on
   read-write mounts. Without the separate gate, an rw mount of a
   hibernated volume (errors=continue) would gain a filesystem-internal
   $Volume write on the first sync_fs() or at unmount - exactly what
   the hibernation check exists to prevent.

   err < 0 is included for the same reason, only more so: when the
   check cannot complete, we cannot prove the volume is safe to write.
   Gating wrongly costs an untouched dirty bit; not gating wrongly
   costs Windows resuming on a modified image.

Thanks!