On 8/31/20 8:38 PM, Andrew Morton wrote: > (cc Vishal & Joao) > > On Mon, 31 Aug 2020 15:11:42 +0800 kernel test robot <lkp@xxxxxxxxx> wrote: > >> tree: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master >> head: b36c969764ab12faebb74711c942fa3e6eaf1e96 >> commit: f7782cc946fea99b8a4a0a83b6c3775a6beff8ce [3093/3958] ACPI: HMAT: refactor hmat_register_target_device to hmem_register_device >> config: arm64-randconfig-r016-20200831 (attached as .config) >> compiler: aarch64-linux-gcc (GCC) 9.3.0 >> 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 >> git checkout f7782cc946fea99b8a4a0a83b6c3775a6beff8ce >> # save the attached .config to linux build tree >> COMPILER_INSTALL_PATH=$HOME/0day COMPILER=gcc-9.3.0 make.cross ARCH=arm64 >> >> If you fix the issue, kindly add following tag as appropriate >> Reported-by: kernel test robot <lkp@xxxxxxxxx> >> >> All errors (new ones prefixed by >>): >> >> aarch64-linux-ld: drivers/acpi/numa/hmat.o: in function `hmat_register_target_devices': >>>> drivers/acpi/numa/hmat.c:658: undefined reference to `hmem_register_device' >> Here's a proposed fix (first patch) but I'm not sure if there's a better one, though. Also, after fixing the one reported above, there's an additional flaw with the reported kconfig. Specifically ARM64 with NUMA=y and MEMORY_HOTPLUG=n. It wasn't reported above, but we would see it right after. See the second patch for that. The two patches could be just one patch as they are introduced by the refactor above ("ACPI: HMAT: refactor hmat_register_target_device to hmem_register_device"), but thought I would split them up per-issue while we discuss the fix. I can attach them merged if you prefer. Joao ---------------------->8------------------- From: Joao Martins <joao.m.martins@xxxxxxxxxx> Date: Tue, 1 Sep 2020 07:29:44 -0400 Subject: [PATCH 1/2] device-dax: CONFIG_DEV_DAX_HMEM_DEVICES should depend on CONFIG_DAX=y HMAT requires hmem_register_device() which is now placed under drivers/dax. ACPI_HMAT is a bool and built-in, and to use hmem_register_device() CONFIG_DAX also needs to be builtin. Reported-by: kernel test robot <lkp@xxxxxxxxx> Signed-off-by: Joao Martins <joao.m.martins@xxxxxxxxxx> --- drivers/dax/Kconfig | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/dax/Kconfig b/drivers/dax/Kconfig index a229f45d34aa..a66dff78f298 100644 --- a/drivers/dax/Kconfig +++ b/drivers/dax/Kconfig @@ -49,7 +49,7 @@ config DEV_DAX_HMEM Say M if unsure. config DEV_DAX_HMEM_DEVICES - depends on DEV_DAX_HMEM + depends on DEV_DAX_HMEM && DAX=y def_bool y config DEV_DAX_KMEM -- 2.17.1 From: Joao Martins <joao.m.martins@xxxxxxxxxx> Date: Tue, 1 Sep 2020 08:06:35 -0400 Subject: [PATCH 2/2] device-dax: Require MEMORY_HOTPLUG 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, so reflect that in its header file declaration. Otherwise, ARM64 configs with CONFIG_NUMA=y and CONFIG_MEMORY_HOTPLUG=n see: drivers/dax/hmem/device.o: in function `hmem_register_one': drivers/dax/hmem/device.c:85: undefined reference to `phys_to_target_node' X86 is a special case in which both can be defined without depending on MEMORY_HOTPLUG. The other case is powerpc but like ARM64 dependencies, the arch-specific override requires CONFIG_NUMA and CONFIG_MEMORY_HOTPLUG. Signed-off-by: Joao Martins <joao.m.martins@xxxxxxxxxx> --- drivers/dax/Kconfig | 1 + 1 file changed, 1 insertion(+) diff --git a/drivers/dax/Kconfig b/drivers/dax/Kconfig index a66dff78f298..6e86ac52f41a 100644 --- a/drivers/dax/Kconfig +++ b/drivers/dax/Kconfig @@ -49,6 +49,7 @@ config DEV_DAX_HMEM Say M if unsure. config DEV_DAX_HMEM_DEVICES + depends on NUMA && (MEMORY_HOTPLUG || X86) # for phys_to_target_node() depends on DEV_DAX_HMEM && DAX=y def_bool y -- 2.17.1
