[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <85muikuh42.fsf@mojatatu.com>
Date: Fri, 14 Jun 2019 15:00:45 -0400
From: Roman Mashak <mrv@...atatu.com>
To: David Ahern <dsahern@...il.com>
Cc: Hangbin Liu <liuhangbin@...il.com>, netdev@...r.kernel.org,
Stephen Hemminger <stephen@...workplumber.org>,
Phil Sutter <phil@....cc>
Subject: Re: [iproute2 net-next PATCH] ip: add a new parameter -Numeric
David Ahern <dsahern@...il.com> writes:
> On 6/12/19 10:01 AM, Roman Mashak wrote:
>> Hangbin Liu <liuhangbin@...il.com> writes:
>>
>>> Add a new parameter '-Numeric' to show the number of protocol, scope,
>>> dsfield, etc directly instead of converting it to human readable name.
>>> Do the same on tc and ss.
>>>
>>> This patch is based on David Ahern's previous patch.
>>>
>>
>> [...]
>>
>> It would be good idea to specify the numerical format, e.g. hex or dec,
>> very often hex is more conveninet representation (for example, when
>> skbmark is encoded of IP address or such).
>>
>> Could you think of extending it '-Numeric' to have an optional argument
>> hex, and use decimal as default.
>>
>
> I do not see how such an option could work.
'numeric' is extern object, so can be accessible from other modules. But
yes, it would require to add some wrappers around print_*int() APIs to
take into account this new parameter.
Yes, it is all-or-nothing approach, one can't have filters dumping their
integers values in hex but actions attached to the filters in decimal
(or the other way around).
> It would be best for iproute2 commands to output fields as hex
> or decimal based on what makes sense for each.
For example, m_police.c historically dumps its index value in hex,
although it's not very sensible IMHO :) Now it's risky to change the code,
because many scripts will likely break.
> That said, I do not see how this patch affects fields
> such as skbmark. Can you give an example? The patch looks fine to me.
The patch only exposes 'numeric' to tc, so yes -- it does not affect
anything. I only suggested to think of more generic approach.
On the 2nd thought: there already exists argument "-raw" for tc which
currently instructs printing handles in hex representation. Why not to
adopt this for ip and ss as well rather then adding new key?
Powered by blists - more mailing lists