On 2020-05-15 03:30, Avri Altman wrote: > @@ -4628,6 +4628,9 @@ static int ufshcd_slave_alloc(struct scsi_device *sdev) > > ufshcd_get_lu_power_on_wp_status(hba, sdev); > > + if (ufshcd_is_hpb_supported(hba)) > + ufshpb_attach_sdev(hba, sdev); > + > return 0; > } This approach is unusual because: - Direct calls from the ufshcd module into the ufshpb module make it impossible to unload the latter kernel module. - No other SCSI LLD calls directly into a device handler. Thanks, Bart.