Re: [linux-next:master 2890/4758] ld.lld: error: undefined symbol: phys_to_target_node

[Joao Martins <joao.m.martins@oracle.com>]

On 9/3/20 6:49 AM, kernel test robot wrote:
> tree:   https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
> head:   4442749a203151a319a5bb8d0b983b84253a6931
> commit: bd295a352cfa24e5110a53f978edb48b7c21ff8f [2890/4758] ACPI: HMAT: attach a device for each soft-reserved range
> config: x86_64-randconfig-a002-20200902 (attached as .config)
> compiler: clang version 12.0.0 (https://github.com/llvm/llvm-project 224d8153b53b16cf535ea1a55afdfe1ec5b1374f)
> reproduce (this is a W=1 build):
>         wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
>         chmod +x ~/bin/make.cross
>         # install x86_64 cross compiling tool for clang build
>         # apt-get install binutils-x86-64-linux-gnu
>         git checkout bd295a352cfa24e5110a53f978edb48b7c21ff8f
>         # save the attached .config to linux build tree
>         COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross ARCH=x86_64 
> 
> If you fix the issue, kindly add following tag as appropriate
> Reported-by: kernel test robot <lkp@intel.com>
> 
> All errors (new ones prefixed by >>):
> 
>>> ld.lld: error: undefined symbol: phys_to_target_node
>    >>> referenced by device.c:84 (drivers/dax/hmem/device.c:84)
>    >>>               dax/hmem/device.o:(hmem_register_one) in archive drivers/built-in.a
> 
This would require the proposed fix below the scissors mark.

I had reported this failure before, but it needed an adjustment so I'm re-sending it.

--------------------------->8----------------------------

From 622c1297b7d76a319b07a29192fe9bfe5c2b6b7c Mon Sep 17 00:00:00 2001
From: Joao Martins <joao.m.martins@oracle.com>
Date: Tue, 1 Sep 2020 08:06:35 -0400
Subject: [PATCH] device-dax: Require NUMA_KEEP_MEMINFO for
 phys_to_target_node()

phys_to_target_node() and memory_add_physaddr_to_nid() are only defined
with both CONFIG_NUMA=y and CONFIG_MEMORY_HOTPLUG=y.  Oherwise we see:

drivers/dax/hmem/device.o: in function `hmem_register_one':
drivers/dax/hmem/device.c:85: undefined reference to `phys_to_target_node'

lkp also reports such build failures:

ld.lld: error: undefined symbol: phys_to_target_node

X86 is a special case in which both can be defined without depending on
MEMORY_HOTPLUG=y. The other case is powerpc (for
memory_add_physaddr_to_nid) but like ARM64 dependencies, the
arch-specific override requires CONFIG_NUMA and CONFIG_MEMORY_HOTPLUG.

In all cases this ends up depending on CONFIG_NUMA_KEEP_MEMINFO to
signal the presence of phys_to_target_node() which is what
CONFIG_DEV_DAX_HMEM_DEVICES should depend on. Additionally, the X86
arch-override is not selected unless X86_PMEM_LEGACY=y. So on
CONFIG_DEV_DAX_HMEM we select CONFIG_NUMA_KEEP_INFO ourselves if both
NUMA and X86 are set.

Reported-by: kernel test robot <lkp@intel.com>
Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
---
 drivers/dax/Kconfig | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/dax/Kconfig b/drivers/dax/Kconfig
index a66dff78f298..567428e10b7b 100644
--- a/drivers/dax/Kconfig
+++ b/drivers/dax/Kconfig
@@ -35,6 +35,7 @@ config DEV_DAX_PMEM
 config DEV_DAX_HMEM
        tristate "HMEM DAX: direct access to 'specific purpose' memory"
        depends on EFI_SOFT_RESERVE
+       select NUMA_KEEP_MEMINFO if (NUMA && X86)
        default DEV_DAX
        help
          EFI 2.8 platforms, and others, may advertise 'specific purpose'
@@ -49,6 +50,7 @@ config DEV_DAX_HMEM
          Say M if unsure.

 config DEV_DAX_HMEM_DEVICES
+       depends on NUMA_KEEP_MEMINFO # for phys_to_target_node()
        depends on DEV_DAX_HMEM && DAX=y
        def_bool y

-- 
2.17.1