Message-ID: <5380A931.3000103@linux-pingi.de>
Date:	Sat, 24 May 2014 16:14:09 +0200
From:	Karsten Keil <kkeil@...ux-pingi.de>
To:	Paul Bolle <pebolle@...cali.nl>
CC:	Tilman Schmidt <tilman@...p.cc>, netdev@...r.kernel.org,
	isdn4linux@...tserv.isdn4linux.de,
	"Keil, Karsten" <isdn@...ux-pingi.de>
Subject: Re: [PATCH 1/4] isdn/capi: move capi_info2str to capidrv.c

Hi Paul,

Am 24.05.2014 13:43, schrieb Paul Bolle:
> Hi Karsten,
> 
> On Sat, 2014-05-24 at 13:01 +0200, Karsten Keil wrote:
>> Am 22.05.2014 23:38, schrieb Paul Bolle:
>>> On Thu, 2014-05-22 at 08:32 +0200, Karsten Keil wrote:
>>>> I disagree here, since this is a general helper function and should be
>>>> not in a special driver, but stay in capiutils.c which is the place for
>>>> the driver independent stuff. I used this function from time to time to
>>>> instrument other places for debugging as well.
>>>
>>> Thanks for commenting. It would be nice if you could also comment on
>>> patch 4/4. You might be able to tell whether the things I say there
>>> about, well, the history of CAPI (middleware) device nodes are correct,
>>> as you were already maintainer during that period, weren't you? 
>>
>> Yes I fully agree here, I have the same opinion as Tilman already
>> stated. And thanks a lot for this work.
> 
> Good to hear, thanks!
> 
>>> This patch, 1/4, and patch 2/4, simplify the Kconfig file and the code a
>>> bit. It makes it  bit easier to understand how the CAPI code fits
>>> together. Same thing with my commit d1958f8c2f0d ("isdn/capi: Make
>>> Middleware depend on CAPI2.0"). It is also nice to drop the
>>> ISDN_DRV_AVMB1_VERBOSE_REASON name, which only makes sense to people
>>> that know the ancient history of this code.
>>
>> I'm not against dropping ISDN_DRV_AVMB1_VERBOSE_REASON completely, it
>> was introduced in a time in which some KByte of memory did count a lot,
>> since the PCs had often less than 4 MByte.
> 
> (I wasn't actually proposing to drop it. The patches rename it and move
> it, and the code it enables, around a bit.)
> 
>> Back to capi_info2str():
>> My issue is that this change moves a part of the CAPI20 specification
>> out of the capi modules. Everything from the CAPI specification (which
>> is also defines these strings), is implemented in the 2 capi modules.
> 
> capi_info2str() is currently - speaking from memory - mostly a list of
> magic constant and some strings. I know that (most of?) the magic
> constants can be mapped to defines elsewhere in the tree. I seem to
> remember that these constants came from the spec. I don't care about
> capidrv myself, so I resisted the urge to clean it all up.
> 
> But, anyhow, do you mean to say that these errors strings themselves are
> to be found in a spec?

Yes it's from the official CAPI 2.0 specification, which is available
from capi.org.

> 
>> The capidrv is not part of the CAPI20 specification, it is only the
>> interface between CAPI20 and the old isdn4linux code, you can completely
>> run CAPI20 applications without capidrv. If you disable I4L, nobody
>> could use the CAPI reason translation. Yes, only the i4l interface
>> driver did use it up to now, but this doesn't mean that it is the
>> correct place. I think that this do not make it clearer how the code
>> fits together.
>>
>>> Anyhow, this patch might complicate your local debugging practices. That
>>> might be inconvenient for you. But in mainline we see a function that's
>>> used in one place only. And I think cleaning up mainline code is what
>>> counts.
>>
>> I'm not against cleaning up.
>> If you still think, that we should move the code I do not compain again,
>> but I want make sure that you understand why it was in that place and
>> that it makes sense from the design of the CAPI20.
> 
> So it seems you're saying we could move capi_info2str() into capidrv as
> long as we store the magic constants (but, in my opinion, as proper
> defines) and the error strings in some central place. Say one of the
> capi headers. Is that what you're saying?
> 

No. I tend to see the translation of the CAPI Info codes as a helper
function of the CAPI (Common ISDN Application Interface), so the central
kernelcapi module should provide it. I see the kernelcapi a little bit
analog to the libcapi in userspace. So application get also some
supporting functions like this from the "library".
But this is only my personal taste or view.

But on the other side, as Tilman stated, capidrv is the only CAPI
application in the kernel (and will be stay the only I think) so your
move can be acceppted by me as well, since this function is not part of
the documented official Linux interface (which was approved by the CAPI
Association in 1999) and became part of the CAPI2.0 specification.


> That would be quite a bit of tedious work, but it might be worth it.
> I'll have to look into that, which might take a few days. Tilman, any
> thoughts already?
> 

No need for doing this I think, here are more important issues. The main
INFO codes are already defined in include/linux/kernelcapi.h.

Best Regards
Karsten

--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Powered by blists - more mailing lists