Re: [PATCH] scsi: ses: Fix out-of-bounds memory access in ses_enclosure_data_process()

[Waiman Long]

On 5/20/19 11:46 AM, James Bottomley wrote:
> On Mon, 2019-05-20 at 11:24 -0400, Waiman Long wrote:
>> On 5/20/19 10:52 AM, James Bottomley wrote:
>>> On Mon, 2019-05-20 at 10:41 -0400, Waiman Long wrote:
>>> [...]
>>>>> --- a/drivers/scsi/ses.c
>>>>> +++ b/drivers/scsi/ses.c
>>>>> @@ -605,9 +605,14 @@ static void
>>>>> ses_enclosure_data_process(struct
>>>>> enclosure_device *edev,
>>>>>  			     /* these elements are optional */
>>>>>  			     type_ptr[0] ==
>>>>> ENCLOSURE_COMPONENT_SCSI_TARGET_PORT ||
>>>>>  			     type_ptr[0] ==
>>>>> ENCLOSURE_COMPONENT_SCSI_INITIATOR_PORT ||
>>>>> -			     type_ptr[0] ==
>>>>> ENCLOSURE_COMPONENT_CONTROLLER_ELECTRONICS))
>>>>> +			     type_ptr[0] ==
>>>>> ENCLOSURE_COMPONENT_CONTROLLER_ELECTRONICS)) {
>>>>>  				addl_desc_ptr +=
>>>>> addl_desc_ptr[1]
>>>>> + 2;
>>>>>  
>>>>> +				/* Ensure no out-of-bounds
>>>>> memory
>>>>> access */
>>>>> +				if (addl_desc_ptr >= ses_dev-
>>>>>> page10 +
>>>>> +						     ses_dev-
>>>>>> page10_len)
>>>>> +					addl_desc_ptr = NULL;
>>>>> +			}
>>>>>  		}
>>>>>  	}
>>>>>  	kfree(buf);
>>>> Ping! Any comment on this patch.
>>> The update looks fine to me:
>>>
>>> Reviewed-by: James E.J. Bottomley <jejb@xxxxxxxxxxxxx>
>>>
>>> It might also be interesting to find out how the proliant is
>>> structuring this descriptor array to precipitate the out of bounds:
>>> Is it just an off by one or something more serious?
>> I didn't look into the detail the enclosure message returned by the
>> hardware, but I believe it may have more description entries (page7)
>> than extended description entries (page10).
>>
>> I can try to reserve the system and find out what exactly is wrong
>> with that system if you really want to find that out.
> Please.  What I'm interested in is whether this is simply a bug in the
> array firmware, in which case the fix is sufficient, or whether there's
> some problem with the parser, like mismatched expectations over added
> trailing nulls or something.
>
> James
>
OK, will let you know once I get hold of the system.

Cheers,
Longman