Re: [PATCH net-next v8 2/3] net sched actions: dump more than TCA_ACT_MAX_PRIO actions per batch

[Simon Horman <simon.horman@netronome.com>]

On Wed, Apr 26, 2017 at 10:46:58AM -0400, David Miller wrote:
> From: Jiri Pirko <jiri@resnulli.us>
> Date: Wed, 26 Apr 2017 15:05:06 +0200
> 
> > Wed, Apr 26, 2017 at 02:46:22PM CEST, jhs@mojatatu.com wrote:
> >>On 17-04-26 07:02 AM, Simon Horman wrote:
> >>> On Tue, Apr 25, 2017 at 06:04:45PM +0200, Jiri Pirko wrote:
> >>[..]
> >>
> >>> > So fix iproute2. It is always first kernel, then iproute2.
> >>> 
> >>> Perhaps I am missing the point or somehow misguided but I would expect that
> >>> if the UAPI uses BIT() it also provides BIT().
> >>
> >>There is a user of BIT() already in iproute2 (devlink). We can move
> >>the code to be more generally available for other iproute2 users.
> >>Then this UAPI change makes use of it.
> > 
> > Should be part of UAPI as well
> > I see that include/uapi/rdma/vmw_pvrdma-abi.h is using BIT macro.
> > I don't see BIT macro defined in UAPI (I thought it is). So either
> > define it there (not sure where) or just use "<<"
> 
> "BIT" is a pretty crazy small simple name to pollute into the global
> namespace, IMHO.

It sounds to me that it would be best to just use "<<" rather than
spending cycles posturing on how to add it to the UAPI. Existing users
of BIT in the UAPI could also be updated to use "<<" to avoid having
a misleading precedence in-tree.