On 2/24/21 3:31 PM, John Garry wrote:
> On 24/02/2021 14:24, Hannes Reinecke wrote:
>> On 2/24/21 2:12 PM, John Garry wrote:
>>> On 22/02/2021 13:23, Hannes Reinecke wrote:
>>>> void scsi_forget_host(struct Scsi_Host *shost)
>>>> {
>>>> - struct scsi_device *sdev;
>>>> + struct scsi_device *sdev, *host_sdev = NULL;
>>>> unsigned long flags;
>>>> restart:
>>>> spin_lock_irqsave(shost->host_lock, flags);
>>>> list_for_each_entry(sdev, &shost->__devices, siblings) {
>>>> + if (scsi_device_is_host_dev(sdev)) {
>>>> + host_sdev = sdev;
>>> Is there actually a limit of 1x host_sdev always?
>>>
>> I would have thought so, as the whole point of having a host device is
>> that you have a (virtual) device which simulates access to the host
>> itself.
>> And as such has a 1:1 relationship to the HBA.
>
> Sure, but I think that each call to scsi_get_host_dev() for the same
> host will give a new sdev each time, right?
>
> We should protect against what is sensible and what is possible - not
> always the same :)
>
The original implementation, yes.
With my patch series, no.
Cheers,
Hannes
--
Dr. Hannes Reinecke Kernel Storage Architect
hare@xxxxxxx +49 911 74053 688
SUSE Software Solutions Germany GmbH, Maxfeldstr. 5, 90409 Nürnberg
HRB 36809 (AG Nürnberg), GF: Felix Imendörffer
