[PATCH v5 02/29] kconfig: Add tools support to CONFIG_IS_ENABLED()

[Alex G.]

On 10/7/21 4:04 PM, Tom Rini wrote:
> On Thu, Oct 07, 2021 at 03:33:32PM -0500, Alex G. wrote:
>>
>>
>> On 10/7/21 2:39 PM, Tom Rini wrote:
>>> On Thu, Oct 07, 2021 at 02:32:42PM -0500, Alex G. wrote:
>>>>
>>>>
>>>> On 10/7/21 1:50 PM, Simon Glass wrote:
>>>>> Hi Tom,
>>>>>
>>>>> On Thu, 7 Oct 2021 at 12:30, Tom Rini <trini at konsulko.com> wrote:
>>>>>>
>>>>>> On Thu, Oct 07, 2021 at 12:02:24PM -0600, Simon Glass wrote:
>>>>>>> Hi Tom,
>>>>>>>
>>>>>>> On Thu, 7 Oct 2021 at 07:42, Tom Rini <trini at konsulko.com> wrote:
>>>>>>>>
>>>>>>>> On Thu, Oct 07, 2021 at 07:32:04AM -0600, Simon Glass wrote:
>>>>>>>>> Hi Tom,
>>>>>>>>>
>>>>>>>>> On Wed, 6 Oct 2021 at 20:52, Tom Rini <trini at konsulko.com> wrote:
>>>>>>>>>>
>>>>>>>>>> On Wed, Oct 06, 2021 at 08:49:13PM -0600, Simon Glass wrote:
>>>>>>>>>>> Hi Tom,
>>>>>>>>>>>
>>>>>>>>>>> On Wed, 6 Oct 2021 at 18:26, Tom Rini <trini at konsulko.com> wrote:
>>>>>>>>>>>>
>>>>>>>>>>>> On Sat, Sep 25, 2021 at 07:43:15PM -0600, Simon Glass wrote:
>>>>>>>>>>>>
>>>>>>>>>>>>> At present we must separately test for the host build for many options,
>>>>>>>>>>>>> since we force them to be enabled. For example, CONFIG_FIT is always
>>>>>>>>>>>>> enabled in the host tools, even if CONFIG_FIT is not enabled by the
>>>>>>>>>>>>> board itself.
>>>>>>>>>>>>>
>>>>>>>>>>>>> It would be more convenient if we could use, for example,
>>>>>>>>>>>>> CONFIG_IS_ENABLED(FIT) and get CONFIG_HOST_FIT, when building for the
>>>>>>>>>>>>> host. Add support for this.
>>>>>>>>>>>>>
>>>>>>>>>>>>> With this and the tools_build() function, we should be able to remove all
>>>>>>>>>>>>> the #ifdefs currently needed in code that is build by tools and targets.
>>>>>>>>>>>>>
>>>>>>>>>>>>> This will be even nicer when we move to using CONFIG(xxx) everywhere,
>>>>>>>>>>>>> since all the #ifdef and IS_ENABLED/CONFIG_IS_ENABLED stuff will go away.
>>>>>>>>>>>>>
>>>>>>>>>>>>> Signed-off-by: Simon Glass <sjg at chromium.org>
>>>>>>>>>>>>> Suggested-by: Rasmus Villemoes <rasmus.villemoes at prevas.dk> # b4f73886
>>>>>>>>>>>>> Reviewed-by: Alexandru Gagniuc <mr.nuke.me at gmail.com>
>>>>>>>>>>>>
>>>>>>>>>>>> The problem here is we don't include <linux/kconfig.h> automatically
>>>>>>>>>>>> when building host stuff, I believe.  This is why doing this breaks
>>>>>>>>>>>> test_mkimage_hashes for me on am335x_evm with:
>>>>>>>>>>>> /tmp/.bm-work/am335x_evm/tools/mkimage -D -I dts -O dtb -i /tmp/.bm-work/am335x_evm -f /home/trini/work/u-boot/u-boot/test/py/tests/vboot//hash-images.its /tmp/.bm-work/am335x_evm/test.fit
>>>>>>>>>>>> *** stack smashing detected ***: <unknown> terminated
>>>>>>>>>>>
>>>>>>>>>>> Oh dear, and no CI coverage.
>>>>>>>>>>>
>>>>>>>>>>> I was reluctant to include kconfig.h everywhere but perhaps that is
>>>>>>>>>>> the best approach. Will take a look ASAP.
>>>>>>>>>>
>>>>>>>>>> Maybe we need to think a bit harder too about how we structure
>>>>>>>>>> intentionally shared code.
>>>>>>>>>>
>>>>>>>>>> Why not, for example, for these common algorithms, rely on typical
>>>>>>>>>> system headers/libraries in the tooling, which means we validated U-Boot
>>>>>>>>>> vs common reference, rather than just our implementations?
>>>>>>>>>
>>>>>>>>> Do you mean we use openssl for sha1, for example?
>>>>>>>>
>>>>>>>> I guess, yes.  Just flat out saying we require openssl for tools, and
>>>>>>>> doing our best to not make compatibility with libressl difficult, seems
>>>>>>>> likely to cause less headaches for people than what we already require
>>>>>>>> in terms of Python.
>>>>>>>
>>>>>>> I'm OK with that, although I do think the problem identified here
>>>>>>> (CONFIG_SHA256 not enabled) is somewhat sideways from that. We already
>>>>>>
>>>>>> OK, I've taken what you posted on IRC and folded that in, continuing
>>>>>> tests now.
>>>>>>
>>>>>>> use separate code paths to run hashing. Perhaps we could make it
>>>>>>> optional?
>>>>>>>
>>>>>>> What about those people that complain about crypto libraries on their systems?
>>>>>>
>>>>>> I'm not sure how big a problem that really is, currently.  I guess one
>>>>>> thing would be to make a separate thread on it, and put it in the next
>>>>>> -rc email as well, for people to explain why it would be a hardship.
>>>>>> That in turn, I think, is coming down to modern vs very old openssl
>>>>>> support, rather than having any at all.
>>>>>
>>>>> OK I'll take a look at some point.
>>>>>
>>>>> Or perhaps Alex might like to?
>>>>
>>>> We just got a complain about OpenSSL yesterday [1]
>>>>
>>>> Alex
>>>>
>>>> [1] https://lists.denx.de/pipermail/u-boot/2021-October/462728.html
>>>
>>> Oh goodness, LibreELC is a custom build system... I'll have to chime in
>>> there, thanks.
>>
>> I am in favor of keeping libcrapto separate. We still need our own code for
>> CRC32 and other weak or non-crypto hashes, a tidbit which makes me doubt the
>> wisdom of relying entirely on an external lib.
>>
>> I had to make a similar decision when writing the hashes test. Originally, I
>> was going to use pyCrypto, crcelk, to re-hash everything and compare to
>> mkimage. It turned out to be neither necessarry nor efficient.
> 
> Is there perhaps a happy medium?  Or do we just need to think harder on
> how to make the code U-Boot needs shared between target and host tools
> clean and clear and obvious enough?

I think hard that's an honorable goal irrespective of the status of 
libcrypto. libcrypto isolation is a happy side-effect.

Alex