Re: PROBLEM：there exists a wrong return value of function ieee802154_add_iface()

Re: PROBLEM：there exists a wrong return value of function ieee802154_add_iface()

[Alexander Aring]

Hi,

I think this is a good time to remove the deprecated netlink interface
for 802.15.4.

Stefan please say your opinion here. There should be really nobody which
use this interface anymore.

- Alex

Re: PROBLEM：there exists a wrong return value of function ieee802154_add_iface()

[Stefan Schmidt]

Hello.

Missing a bit context here as the original message never made it to the 
list.

On 10/03/16 10:10, Alexander Aring wrote:
> Hi,
>
> I think this is a good time to remove the deprecated netlink interface
> for 802.15.4.
>
> Stefan please say your opinion here. There should be really nobody which
> use this interface anymore.

My mind haven't really changed on this topic. I understand why you want 
it removed but given that the general rule of the kernel is to have 
backward compatibility we can't just remove it. Dropping the old netlink 
interface breaks userspace compatibility, even if it only is for 
deprecated tools like lowpan-tools, etc.

My take on it is that we live with it. Keeping it around but not 
extending it. Maybe add a warning to dmesg iwhen some application uses 
the old interface.

Is it more code cleanliness/maintenance or bug reports that urge you to 
get rid of it?

regards
Stefan Schmidt

Re: PROBLEM：there exists a wrong return value of function ieee802154_add_iface()

[Alexander Aring]

On Thu, Mar 10, 2016 at 05:39:12PM +0100, Stefan Schmidt wrote:
> Hello.
> 
> Missing a bit context here as the original message never made it to the
> list.
> 

linux-wpan was cc'ed inside the mail, but there was some html stuff
there... I don't know why it isn't arrived at linux-wpan.

Anyway...

There exists a bugzilla entry for that and John Linville said that I
should care about that, see [0].

> On 10/03/16 10:10, Alexander Aring wrote:
> >Hi,
> >
> >I think this is a good time to remove the deprecated netlink interface
> >for 802.15.4.
> >
> >Stefan please say your opinion here. There should be really nobody which
> >use this interface anymore.
> 
> My mind haven't really changed on this topic. I understand why you want it
> removed but given that the general rule of the kernel is to have backward
> compatibility we can't just remove it. Dropping the old netlink interface
> breaks userspace compatibility, even if it only is for deprecated tools like
> lowpan-tools, etc.
> 
> My take on it is that we live with it. Keeping it around but not extending
> it. Maybe add a warning to dmesg iwhen some application uses the old
> interface.
> 
> Is it more code cleanliness/maintenance or bug reports that urge you to get
> rid of it?

This is one of the smallest problem, I can live with "live with it". But
when I need to touch it because "nl802154" and "ieee802154" use the same
INTERNAL interface, e.g. llsec. I don't want to handle backwards
compatibility for that.

This complete handling with the damnd backwards compatibility for a
interface which is totally broken. (Yes I think [0] is one of the
smallest issues). We getting stucked in developing of 802.15.4 to
develop a better netlink interface (nl802154).

The old interface I would describe as:

- deprecated
- broken
- there exists no official userspace tool to handle everything, e.g.
  llsec, txpower, csma, etc...

I want to remove it otherwise the branch will die again, because stupid
handling a broken interface which nobody should use anymore.

My argument is:

Don't forget the 802.15.4 branch is still experimental, no maintainer
acked this commit [1]. For EXPERIMENTAL things you can be sure that the
UAPI interface will not be changed later and I would still say this
branch is EXPERIMENTAL.


I also have 3-4 things which we need to change inside the new interface
(nl802154), because the deprecated already handle it wrong and I simple
copy&pasted it... but it was wrong. Then handling of two UAPI's which
need to access the INTERNAL api (cfg802154) will be mostly impossible.

- Alex

[0] https://bugzilla.kernel.org/show_bug.cgi?id=114231
[1] https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/net/ieee802154?id=f4671a90c418b5aae14b61a9fc9d79c629403ca0

Re: PROBLEM：there exists a wrong return value of function ieee802154_add_iface()

[Stefan Schmidt]

Hello.

On 14/03/16 10:30, Alexander Aring wrote:
> On Thu, Mar 10, 2016 at 05:39:12PM +0100, Stefan Schmidt wrote:
>> Hello.
>>
>> Missing a bit context here as the original message never made it to the
>> list.
>>
> linux-wpan was cc'ed inside the mail, but there was some html stuff
> there... I don't know why it isn't arrived at linux-wpan.

If it had html in it vger would reject it.

> Anyway...
>
> There exists a bugzilla entry for that and John Linville said that I
> should care about that, see [0].

OK

>> On 10/03/16 10:10, Alexander Aring wrote:
>>> Hi,
>>>
>>> I think this is a good time to remove the deprecated netlink interface
>>> for 802.15.4.
>>>
>>> Stefan please say your opinion here. There should be really nobody which
>>> use this interface anymore.
>> My mind haven't really changed on this topic. I understand why you want it
>> removed but given that the general rule of the kernel is to have backward
>> compatibility we can't just remove it. Dropping the old netlink interface
>> breaks userspace compatibility, even if it only is for deprecated tools like
>> lowpan-tools, etc.
>>
>> My take on it is that we live with it. Keeping it around but not extending
>> it. Maybe add a warning to dmesg iwhen some application uses the old
>> interface.
>>
>> Is it more code cleanliness/maintenance or bug reports that urge you to get
>> rid of it?
> This is one of the smallest problem, I can live with "live with it". But
> when I need to touch it because "nl802154" and "ieee802154" use the same
> INTERNAL interface, e.g. llsec. I don't want to handle backwards
> compatibility for that.
>
> This complete handling with the damnd backwards compatibility for a
> interface which is totally broken. (Yes I think [0] is one of the
> smallest issues). We getting stucked in developing of 802.15.4 to
> develop a better netlink interface (nl802154).
>
> The old interface I would describe as:
>
> - deprecated
> - broken
> - there exists no official userspace tool to handle everything, e.g.
>    llsec, txpower, csma, etc...
>
> I want to remove it otherwise the branch will die again, because stupid
> handling a broken interface which nobody should use anymore.
>
> My argument is:
>
> Don't forget the 802.15.4 branch is still experimental, no maintainer
> acked this commit [1]. For EXPERIMENTAL things you can be sure that the
> UAPI interface will not be changed later and I would still say this
> branch is EXPERIMENTAL.
>
>
> I also have 3-4 things which we need to change inside the new interface
> (nl802154), because the deprecated already handle it wrong and I simple
> copy&pasted it... but it was wrong. Then handling of two UAPI's which
> need to access the INTERNAL api (cfg802154) will be mostly impossible.

You are sold on it and you are doing a lot of work there. Give it a try.

Maybe you are lucky and it will just go through and we only have to deal 
with a few complaints about lowpan-tools no longer working with 4.6/4.7.
Be prepared though that if everything goes south you might find yourself 
being yelled at by Linus. :) Just saying.

regards
Stefan Schmidt