Re: [PATCH bpf-next] bpftool: Add BPF_F_QUERY_EFFECTIVE support in bpftool cgroup [show|tree]

[Alexei Starovoitov <ast@fb.com>]

On 6/24/19 5:47 PM, Jakub Kicinski wrote:
> On Tue, 25 Jun 2019 00:40:09 +0000, Alexei Starovoitov wrote:
>> On 6/24/19 5:30 PM, Jakub Kicinski wrote:
>>> On Tue, 25 Jun 2019 00:21:57 +0000, Alexei Starovoitov wrote:
>>>> On 6/24/19 5:16 PM, Jakub Kicinski wrote:
>>>>> On Mon, 24 Jun 2019 23:38:11 +0000, Alexei Starovoitov wrote:
>>>>>> I don't think this patch should be penalized.
>>>>>> I'd rather see we fix them all.
>>>>>
>>>>> So we are going to add this broken option just to remove it?
>>>>> I don't understand.
>>>>> I'm happy to spend the 15 minutes rewriting this if you don't
>>>>> want to penalize Takshak.
>>>>
>>>> hmm. I don't understand the 'broken' part.
>>>> The only issue I see that it could have been local vs global,
>>>> but they all should have been local.
>>>
>>> I don't think all of them.  Only --mapcompat and --bpffs.  bpffs could
>>> be argued.  On mapcompat I must have not read the patch fully, I was
>>> under the impression its a global libbpf flag :(
>>>
>>> --json, --pretty, --nomount, --debug are global because they affect
>>> global behaviour of bpftool.  The difference here is that we basically
>>> add a syscall parameter as a global option.
>>
>> sure. I only disagreed about not touching older flags.
>> --effective should be local.
>> If follow up patch means 90% rewrite then revert is better.
>> If it's 10% fixup then it's different story.
> 
> I see.  The local flag would not an option in getopt_long() sense, what
> I was thinking was about adding an "effective" keyword:
> 
> # bpftool -e cgroup show /sys/fs/cgroup/cgroup-test-work-dir/cg1/
> 
> becomes:
> 
> # bpftool cgroup show /sys/fs/cgroup/cgroup-test-work-dir/cg1/ effective
> 
> That's how we handle flags for update calls for instance..
> 
> So I think a revert :(

fair enough.
removed it and force pushed bpf-next.