On Thu, 2017-06-29 at 15:35 -0400, David Miller wrote:

From: Mike Galbraith <redacted>
Date: Wed, 28 Jun 2017 10:01:00 +0200

quoted

Greetings network wizards,

The latest RT explicitly disables CONFIG_NET_RX_BUSY_POLL, thus
uncovering $subject.  Below is what I did about it.

      -Mike

net: Move napi_hash_add/del() inside CONFIG_NET_RX_BUSY_POLL

Since 545cd5e5ec54 ("net: Busy polling should ignore sender CPUs"),
kernel build fails when CONFIG_NET_RX_BUSY_POLL is disabled.  Move
napi_hash_add/del() accordingly.

Banged-upon-by: Mike Galbraith [off-list ref]

First, this is in no way a standard signoff or ack tag.

It's not intended to be, it's intended to stress that this in not a
submission, it's a local fix for a problem that network folks will
likely fix up differently... but who knows, showing what I did about it
after taking a look could perhaps save a wee bit of labor, so...

Hohum, so much for good intentions.  I though I was clearly showing
that separation of the polling business was not as complete as other
changes led me to believe it is intended to be.

Second, you must provide the complete set of build failure details
so that anyone just reading your posting can understand and fully
review your patch without having to go anywhere else for the
pertinent information.

Here's the naked failure instead.

  CC      net/core/dev.o
net/core/dev.c: In function ‘napi_hash_add’:
net/core/dev.c:5315:43: error: ‘MIN_NAPI_ID’ undeclared (first use in this function)
   if (unlikely(++napi_gen_id < MIN_NAPI_ID))
                                           ^
net/core/dev.c:5315:43: note: each undeclared identifier is reported only once for each function it appears in
scripts/Makefile.build:302: recipe for target 'net/core/dev.o' failed
make[1]: *** [net/core/dev.o] Error 1
Makefile:1663: recipe for target 'net/core/dev.o' failed
make: *** [net/core/dev.o] Error 2

	-Mike