Re: [PATCH] memblock: fix a section mismatch when a function doesn't get inlined

Re: [PATCH] memblock: fix a section mismatch when a function doesn't get inlined

[Mike Rapoport]

Hi Adam,

On Sun, Sep 12, 2021 at 10:41:11PM +0200, Adam Borowski wrote:
> WARNING: modpost: vmlinux.o(.text.unlikely+0xab6): Section mismatch in
>     reference from the function memblock_phys_alloc() to the function
>     .init.text:memblock_phys_alloc_range()
> The function memblock_phys_alloc() references
> the function __init memblock_phys_alloc_range().
> This is often because memblock_phys_alloc lacks a __init
> annotation or the annotation of memblock_phys_alloc_range is wrong.

There is a bunch of other wrappers in include/linux/memblock.h that
most probably would cause the same warning.

Can add the patch below and build the kernel with the config that prevents
inlining of memblock_phys_alloc to see what the compiler thinks about other
wrappers? I believe they all would produce the same warning.

diff --git a/mm/memblock.c b/mm/memblock.c
index 28a813d9e955..6351c5754bb0 100644
--- a/mm/memblock.c
+++ b/mm/memblock.c
@@ -2089,6 +2089,32 @@ void __init memblock_free_all(void)
 	totalram_pages_add(pages);
 }
 
+void __init memblock_test_inlining(void)
+{
+	void *addr;
+
+	addr = memblock_alloc_raw(PAGE_SIZE, PAGE_SIZE);
+	if (!addr)
+		return;
+	memblock_free(__pa(addr), PAGE_SIZE);
+
+	addr = memblock_alloc_from(PAGE_SIZE, PAGE_SIZE,
+				   MEMBLOCK_ALLOC_ACCESSIBLE);
+	if (!addr)
+		return;
+	memblock_free(__pa(addr), PAGE_SIZE);
+
+	addr = memblock_alloc_low(PAGE_SIZE, PAGE_SIZE);
+	if (!addr)
+		return;
+	memblock_free(__pa(addr), PAGE_SIZE);
+
+	addr = memblock_alloc_node(PAGE_SIZE, PAGE_SIZE, NUMA_NO_NODE);
+	if (!addr)
+		return;
+	memblock_free(__pa(addr), PAGE_SIZE);
+}
+
 #if defined(CONFIG_DEBUG_FS) && defined(CONFIG_ARCH_KEEP_MEMBLOCK)
 
 static int memblock_debug_show(struct seq_file *m, void *private)

> Simplest way is to just ensure it is inlined.
> 
> Signed-off-by: Adam Borowski <kilobyte@angband.pl>
> ---
>  Found when running riscv64 randconfigs, but doesn't appear to be arch
>  specific.
> 
>  include/linux/memblock.h | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/include/linux/memblock.h b/include/linux/memblock.h
> index b066024c62e3..dfdd7c50c27d 100644
> --- a/include/linux/memblock.h
> +++ b/include/linux/memblock.h
> @@ -387,8 +387,8 @@ phys_addr_t memblock_alloc_range_nid(phys_addr_t size,
>  				      phys_addr_t end, int nid, bool exact_nid);
>  phys_addr_t memblock_phys_alloc_try_nid(phys_addr_t size, phys_addr_t align, int nid);
>  
> -static inline phys_addr_t memblock_phys_alloc(phys_addr_t size,
> -					      phys_addr_t align)
> +static __always_inline phys_addr_t memblock_phys_alloc(phys_addr_t size,
> +						       phys_addr_t align)
>  {
>  	return memblock_phys_alloc_range(size, align, 0,
>  					 MEMBLOCK_ALLOC_ACCESSIBLE);
> -- 
> 2.33.0
> 

-- 
Sincerely yours,
Mike.