Re: [PATCH v3 2/7] can: Add can_common.h for vcan device setup

[Richard Palethorpe <rpalethorpe@suse.de>]

Hello Petr & Oliver,

Oliver Hartkopp <socketcan@hartkopp.net> writes:

> Hi Petr,
>
> On 26.01.21 22:28, Petr Vorel wrote:
>> Hi Oliver, Richie,
>> 
>>>>> --- /dev/null
>>>>> +++ b/testcases/network/can/filter-tests/can_common.h
>>>>> @@ -0,0 +1,75 @@
>>>>> +// SPDX-License-Identifier: GPL-2.0-or-later
>>>>> +/*
>>>>> + * Copyright (c) 2021 SUSE LLC
>>>>> + */
>>>>> +
>>>>> +#include <stdio.h>
>>>>> +#include <stdlib.h>
>>>>> +#include <unistd.h>
>>>>> +#include <string.h>
>>>>> +
>>>>> +#include <sys/types.h>
>>>>> +#include <sys/socket.h>
>>>>> +#include <sys/ioctl.h>
>>>>> +#include <sys/time.h>
>>>>> +
>>>>> +#include "tst_cmd.h"
>>>>> +#include "tst_safe_stdio.h"
>>>>> +#include "tst_safe_file_ops.h"
>>>>> +
>>>>> +#include <linux/if.h>
>>>>> +#include <linux/can.h>
>>>>> +#include <linux/can/raw.h>
>>>>> +
>>>>> +#ifndef IFF_ECHO
>>>>> +# define IFF_ECHO (1<<18)
>>>>> +#endif
>> Thanks a lot, Richie!
>> 
>>>> IFF_ECHO was included into Linux 2.6.25 together with the CAN
>>>> subsystem itself.
>> 
>>>> So when you run the tests on Kernels < 2.6.25 you don't have CAN
>>>> support and don't need IFF_ECHO too.
>> 
>>> Petr, what kernel version and/or distro version did compilation fail on?
>> 
>>> There is a small chance someone might be compiling with old kernel
>>> headers relative to their kernel. However it is a challenge to compile
>>> LTP with such an old user land.
>> No, we don't support 2.6.25 :). I was playing with Buildroot distro
>> in my spare time.
>> These embedded toolchains suffer compatibility problems (usually uclibc-ng and
>> sometimes musl lack the support). This problem was when using sourcery-arm-*.
>
> :-/
>
>> But this is definitely not a blocker for this patchset. That lapi is not a must,
>> I can fix it some time later. I usually fix few of these problems before each
>> LTP release.
>
> Ok. No problem. I wasn't aware that e.g. musl or other toolchains
> select such strange starting points for their include files.

I wonder Petr, is it still necessary to define IFF_ECHO now only
<linux/if.h> is included? Or do they somehow symlink linux/if.h ->
net/if.h?

Indeed it seems the current version of uclibc-ng doesn't include
IFF_ECHO in <net/if.h>. OTOH musl does define it.

-- 
Thank you,
Richard.