Re: Fwd: [PATCH v7 1/2] x86: Add Isolated Memory Regions for Quark X1000

["Bryan O'Donoghue" <pure.logic@nexus-software.ie>]

On 30/01/15 14:04, Andy Shevchenko wrote:
>> On 30/01/15 12:03, Andy Shevchenko wrote:
>
>>> When CONFIG_DEBUGFS=n, you will get error pointer here, which is not NULL.
>>>
>>> So, the proper check is
>>> if (IS_ERR())
>>>    return PTR_ERR();
>>> if (!file)
>>>    return -ENOMEM;
>>
>> Yeah I saw that. Also saw that most other code doesn't bother trapping
>> those return values - so skipped it.
>> No issue adding.
>
> Ok.
>
>>>> +               } else {
>>>> +                       reg = i;
>>>
>>> Do we go always through all IMRs and choose the last one?
>>> If no, break is missed here.
>>
>> Yep - we always choose the last one.
>
> OK.
>
>>>> +       ret = imr_check_params(base, size);
>>>> +       if (ret == -EINVAL || (ret == -ENOMEM && reg == -1))
>>>
>>> reg base size (0 correct, 1 wrong):
>>>
>>> 0 0 0 — which should be used? what is the priority?
>>> 0 x 1 — index
>>> 0 1 x — index
>>> 1 0 0 — address
>>> 1 0 1 — an error
>>> 1 1 1 — an error
>>>
>>> Thus, could it be simpler? Like
>>> if (reg < 0 && ret) ?
>>
>> ret will be EINVAL for unaligned base or size
>> ret will be ENOMEM when reg == -1 and size == 0
>
> ENOMEM only when size == 0.
>
> I have read the function description. From my point of view there is
> no difference what is in (base, size) if we have index set.

I should be clearer :)

Will change so that -ENOMEM won't be returned. ENOMEM transmits no 
useful information...

>
>> I could probably write it like this to make it clearer
>> (ret == -EINVAL || (reg == -1 && size == 0)
>>          return -EINVAL;
>>
>> traps unaligned input - for address range tear-down
>> traps zero sized - for address range tear-down
>>
>> Allows index based teardown i.e. reg >= 0
>
> Again, seems a logic mistake here, or description of function is not correct.
> Do we care what (base, size) if index is set? If so, why is it not
> mentioned properly in the description?

No we don't if index >= 0 base and size are ignored => teardown by index.

I'll make that clear in a comment with the refactored statement above. 
-ENOMEM is redundant

--
BOD