[U-Boot] [PATCH] fdt: fix address cell count checking in fdt_translate_address()

[Simon Glass]

Hi Przemyslaw,

On 12 January 2016 at 03:49, Przemyslaw Marczak <p.marczak at samsung.com> wrote:
> Hello Simon,
>
>
> On 01/11/2016 05:59 PM, Simon Glass wrote:
>>
>> Hi Przemyslaw,
>>
>> On 8 January 2016 at 05:01, Przemyslaw Marczak <p.marczak at samsung.com>
>> wrote:
>>>
>>> Hello Simon,
>>>
>>>
>>> On 01/07/2016 08:24 PM, Simon Glass wrote:
>>>>
>>>>
>>>> +Stephen
>>>>
>>>> On 4 January 2016 at 17:59, Simon Glass <sjg at chromium.org> wrote:
>>>>>
>>>>>
>>>>> Hi Przemyslaw,
>>>>>
>>>>> On 5 November 2015 at 23:47, Stefan Roese <sr at denx.de> wrote:
>>>>>>
>>>>>>
>>>>>> On 06.11.2015 04:16, Simon Glass wrote:
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> Hi,
>>>>>>>
>>>>>>> On 3 November 2015 at 02:57, Przemyslaw Marczak
>>>>>>> <p.marczak at samsung.com>
>>>>>>> wrote:
>>>>>>>>
>>>>>>>>
>>>>>>>>
>>>>>>>> Hello All,
>>>>>>>>
>>>>>>>>
>>>>>>>> On 10/29/2015 06:15 PM, Simon Glass wrote:
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> Hi Stefan,
>>>>>>>>>
>>>>>>>>> On 28 October 2015 at 08:37, Przemyslaw Marczak
>>>>>>>>> <p.marczak at samsung.com>
>>>>>>>>> wrote:
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>>
>>>>>>>>>> Commit: dm: core: Enable optional use of fdt_translate_address()
>>>>>>>>>>
>>>>>>>>>> Enables use of this function as default, but after this it's not
>>>>>>>>>> possible to get dev address for the case in which: '#size-cells ==
>>>>>>>>>> 0'
>>>>>>>>>>
>>>>>>>>>> This causes errors when getting address for some GPIOs, for which
>>>>>>>>>> the '#size-cells' is set to 0.
>>>>>>>>>>
>>>>>>>>>> Example error:
>>>>>>>>>> '__of_translate_address: Bad cell count for gpx0'
>>>>>>>>>>
>>>>>>>>>> Allowing for that case by modifying the macro 'OF_CHECK_COUNTS',
>>>>>>>>>> (called from )__of_translate_address(), fixes the issue.
>>>>>>>>>>
>>>>>>>>>> Now, this macro doesn't check, that '#size-cells' is greater than
>>>>>>>>>> 0.
>>>>>>>>>>
>>>>>>>>>> This is possible from the specification point of view, but I'm not
>>>>>>>>>> sure
>>>>>>>>>> that it doesn't introduce a regression for other configs.
>>>>>>>>>>
>>>>>>>>>> Please test and share the results.
>>>>>>>>>>
>>>>>>>>>> Tested-on: Odroid U3, Odroid X2, Odroid XU3, Sandbox.
>>>>>>>>>>
>>>>>>>>>> Signed-off-by: Przemyslaw Marczak <p.marczak at samsung.com>
>>>>>>>>>> Cc: Masahiro Yamada <yamada.masahiro at socionext.com>
>>>>>>>>>> Cc: Lukasz Majewski <l.majewski at samsung.com>
>>>>>>>>>> Cc: Jaehoon Chung <jh80.chung at samsung.com>
>>>>>>>>>> Cc: Stefan Roese <sr at denx.de>
>>>>>>>>>> Cc: Simon Glass <sjg at chromium.org>
>>>>>>>>>> Cc: Bin Meng <bmeng.cn at gmail.com>
>>>>>>>>>> Cc: Marek Vasut <marex at denx.de>
>>>>>>>>>> ---
>>>>>>>>>>      common/fdt_support.c | 7 +++----
>>>>>>>>>>      1 file changed, 3 insertions(+), 4 deletions(-)
>>>>>>>>>>
>>>>>>>>>> diff --git a/common/fdt_support.c b/common/fdt_support.c
>>>>>>>>>> index f86365e..5f808cc 100644
>>>>>>>>>> --- a/common/fdt_support.c
>>>>>>>>>> +++ b/common/fdt_support.c
>>>>>>>>>> @@ -946,8 +946,7 @@ void fdt_del_node_and_alias(void *blob, const
>>>>>>>>>> char
>>>>>>>>>> *alias)
>>>>>>>>>>      /* Max address size we deal with */
>>>>>>>>>>      #define OF_MAX_ADDR_CELLS      4
>>>>>>>>>>      #define OF_BAD_ADDR    ((u64)-1)
>>>>>>>>>> -#define OF_CHECK_COUNTS(na, ns)        ((na) > 0 && (na) <=
>>>>>>>>>> OF_MAX_ADDR_CELLS && \
>>>>>>>>>> -                       (ns) > 0)
>>>>>>>>>> +#define OF_CHECK_COUNTS(na)    ((na) > 0 && (na) <=
>>>>>>>>>> OF_MAX_ADDR_CELLS)
>>>>>>>>>>
>>>>>>>>>>      /* Debug utility */
>>>>>>>>>>      #ifdef DEBUG
>>>>>>>>>> @@ -1115,7 +1114,7 @@ static u64 __of_translate_address(void
>>>>>>>>>> *blob,
>>>>>>>>>> int
>>>>>>>>>> node_offset, const fdt32_t *in
>>>>>>>>>>
>>>>>>>>>>             /* Cound address cells & copy address locally */
>>>>>>>>>>             bus->count_cells(blob, parent, &na, &ns);
>>>>>>>>>> -       if (!OF_CHECK_COUNTS(na, ns)) {
>>>>>>>>>> +       if (!OF_CHECK_COUNTS(na)) {
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>>
>>>>>>>>> This seems to conflict with the comment at the top of this
>>>>>>>>> function:
>>>>>>>>>
>>>>>>>>>      * Note: We consider that crossing any level with #size-cells
>>>>>>>>> == 0
>>>>>>>>> to
>>>>>>>>> mean
>>>>>>>>>      * that translation is impossible (that is we are not dealing
>>>>>>>>> with
>>>>>>>>> a
>>>>>>>>> value
>>>>>>>>>      * that can be mapped to a cpu physical address). This is not
>>>>>>>>> really
>>>>>>>>> specified
>>>>>>>>>      * that way, but this is traditionally the way IBM at least do
>>>>>>>>> things
>>>>>>>>>
>>>>>>>>> What should we do here?
>>>>>>>>>
>>>>>>>>
>>>>>>>> Is that commit acceptable? I would like send V2 with removing the
>>>>>>>> above
>>>>>>>> comment.
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> That's what I am worried about. Presumably the comment is accurate
>>>>>>> today and this check has some value. I was hoping Stefan might know.
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> Unfortunately no. I just stumbled over this problem with the
>>>>>> translation of the "complex" ranges on the MVEBU platform. And
>>>>>> noticed that we already have this functionality to translate
>>>>>> the addresses the "right way".
>>>>>>
>>>>>> I'm wondering how this problem with those GPIOs is handled in
>>>>>> the kernel? I assume that it is working correctly there, right?
>>>>>> Przemyslaw, could you perhaps check this and see, why its
>>>>>> working there? And change / fix it in U-Boot accordingly?
>>>>>
>>>>>
>>>>>
>>>>> Let's pick up this patch for now as a bug-fix. We can deal with this
>>>>> problem after the release.
>>>>
>>>>
>>>>
>>>> Applied to u-boot-dm/master.
>>>>
>>>> I'll post a revert after the release. It seems like you and Stephen
>>>> are making good progress.
>>>>
>>>> - Simon
>>>>
>>>>
>>>
>>> Why so fast with this one?
>>>
>>> I think, that more proper for a temporary fix is my latest patch with
>>> #size-cells count checking only if ranges found in the parent node.
>>>
>>> I will continue the discussion with Stephen.
>>
>>
>> The release is scheduled for today, so we had to do something to fix
>> the breakage.
>>
>> Once you have a full solution figured out we can revert this patch and
>> apply what you come up with.
>>
>> Regards,
>> Simon
>>
>>
>
> Ok. It's hard to convince Stephen to accept such change, so I will send a
> patch with another solution - just bring back fdtdec_get_addr() for Exynos
> GPIO driver. And will revert this one within the patchset.

Please hold off on that. I'll accept your other patch but let's see if
Stephen wants to write something first. Using fdtdec_get_addr()
doesn't make sense although I fully understand your frustration. Let's
give it a week.

Regards,
Simon