On 2021/11/16 13:22, Bart Van Assche wrote:
> On 11/15/21 19:29, Damien Le Moal wrote:
>> On 2021/11/16 11:18, Bart Van Assche wrote:
>>> On 11/15/21 1:29 AM, Damien Le Moal wrote:
>>>> diff --git a/drivers/scsi/hosts.c b/drivers/scsi/hosts.c
>>>> index 8049b00b6766..c3b6812aac5b 100644
>>>> --- a/drivers/scsi/hosts.c
>>>> +++ b/drivers/scsi/hosts.c
>>>> @@ -359,6 +359,7 @@ static void scsi_host_dev_release(struct device *dev)
>>>> static struct device_type scsi_host_type = {
>>>> .name = "scsi_host",
>>>> .release = scsi_host_dev_release,
>>>> + .groups = scsi_sysfs_shost_attr_groups,
>>>> };
>>>
>>> Many SCSI LLDs use class_to_shost() to convert a device pointer into a SCSI host
>>> pointer. This patch makes the use of that macro very confusing since the SCSI
>>> host class is no longer involved in attribute registration.
>>
>> OK. But at least I think we should fix this:
>>
>> WARN_ON_ONCE(j >= ARRAY_SIZE(shost->shost_dev_attr_groups));
>>
>> to change it into:
>>
>> if (WARN_ON_ONCE(j >= ARRAY_SIZE(shost->shost_dev_attr_groups)))
>> shost->shost_dev_attr_groups[j] = NULL;
>>
>> to guarantee that the attribute groups array is NULL terminated, as it should
>> be. This will ensure that we do not end up with a kernel crash when a buggy
>> driver is loaded. No ?
>
> Hi Damien,
>
> Please take a look at
> https://lore.kernel.org/linux-scsi/ab1a9bfd-c1d2-e101-a9f3-f969ed3d1cad@xxxxxxx/.
> That patch removes the WARN_ON_ONCE() statement entirely.
Looks good !
>
> Thanks,
>
> Bart.
>
--
Damien Le Moal
Western Digital Research
