Re: hexagon: build error in -next due to 'mm: memcontrol: per-lruvec stats infrastructure'

From: Andrew Morton
Date: Fri Jun 16 2017 - 16:37:38 EST

On Fri, 16 Jun 2017 16:15:23 -0400 Johannes Weiner <hannes@xxxxxxxxxxx> wrote:

> In any case, memcontrol.h doesn't/shouldn't need hardirq.h. When that
> include is removed, the below patch compiles on: x86 allno, x86_64
> allno, and my regular x86_64 config:
> ---
> Subject: mm-memcontrol-per-lruvec-stats-infrastructure-fix-4

Did you try x86_64 allmodconfig? I'm getting a mess:

In file included from ./include/linux/mm.h:1032,
from ./include/linux/highmem.h:7,
from ./include/linux/bio.h:21,
from ./include/linux/writeback.h:205,
from ./include/linux/memcontrol.h:28,
from ./include/linux/swap.h:8,
from ./include/linux/suspend.h:4,
from arch/x86/kernel/asm-offsets.c:12:
./include/linux/vmstat.h: In function 'lruvec_page_state':
./include/linux/vmstat.h:362: error: implicit declaration of function 'mem_cgroup_disabled'
./include/linux/vmstat.h:365: error: dereferencing pointer to incomplete type
./include/linux/vmstat.h:365: error: type defaults to 'int' in declaration of 'type name'

Presumably because we have memcontrol.h indirectly including mm.h which
includes vmstat.h (from a stupid place) and with this patch we have
vmstat.h including memcontrol.h.