On 6/27/24 7:18 PM, Kyoungrul Kim wrote:
Yes, scsi_add_host has been moved the async function (ufshcd_device_init)as MCQ patch applied. So the ufshcd driver attaches the device without knowing whether the probe fails or not. and if host tries to remove ufshcd driver, it makes kernel panic. So it became necessary to check whether to add a host or not.
There are two scsi_add_host() calls in the UFS driver and this patch only adds one "hba->scsi_host_added = true" assignment. Shouldn't this patch add two such assignments? Thanks, Bart.
