RE: [linux-next:master 1753/1936] mm/swapfile.c:1517:20: warning: function 'swap_is_last_ref' is not needed and will not be emitted

["Li, Philip" <philip.li@intel.com>]

> Subject: Re: [linux-next:master 1753/1936] mm/swapfile.c:1517:20: warning: function 'swap_is_last_ref' is not
> needed and will not be emitted
> 
> On 04/09/25 at 09:54am, SeongJae Park wrote:
> > Hello,
> >
> > On Tue, 8 Apr 2025 17:39:59 -0700 Andrew Morton <akpm@linux-foundation.org> wrote:
> >
> > > On Tue, 8 Apr 2025 18:06:11 +0800 kernel test robot <lkp@intel.com> wrote:
> > >
> > > > tree:   https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
> > > > head:   7702d0130dc002bab2c3571ddb6ff68f82d99aea
> > > > commit: 6bb001b6b64ec20d900d0d81294a0f031af59c3e [1753/1936] mm: swap: enable swap_entry_range_free() to
> drop any kind of last ref
> > > > config: hexagon-randconfig-001-20250408 (https://download.01.org/0day-
> ci/archive/20250408/202504081841.jbL7yL0r-lkp@intel.com/config)
> > > > compiler: clang version 21.0.0git (https://github.com/llvm/llvm-project
> 92c93f5286b9ff33f27ff694d2dc33da1c07afdd)
> > > > reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250408/202504081841.jbL7yL0r-
> lkp@intel.com/reproduce)
> > > >
> 
> Hi lkp team,

Hi Baoquan, sorry for late reply. This is an issue in the bot that when clang-21 is not yet released. Early
on we tried to add a reminder like below (suggested by another similar case).

	# clang-21 might not be officially released. Try a lower version if installation fails.

We will add this reminder asap. Thus when clang-21 fails to install, clang-20 could be an alternative
to reproduce.

Sorry for inconvenience.

Thanks

> 
> When I tried to reproduce, I got below failure. Please help check what's
> wrong with it. Thanks in advance.
> 
> [~]# COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang-21 ~/lkp-tests/kbuild/make.cross W=1 O=build_dir
> ARCH=hexagon olddefconfig
> Compiler will be installed in /root/0day
> Cannot find clang-21 under https://cdn.kernel.org/pub/tools/llvm/files
> Please set new URL env variable and rerun
>   * crosstool provided by kernel org:
>     export URL=https://cdn.kernel.org/pub/tools/llvm/files
>   * crosstool provided by 0-Day CI:
>     export URL=https://download.01.org/0day-ci/cross-package
> clang crosstool install failed
> Install clang compiler failed
> setup_crosstool failed
> 
> > > > If you fix the issue in a separate patch/commit (i.e. not just a new version of
> > > > the same patch/commit), kindly add following tags
> > > > | Reported-by: kernel test robot <lkp@intel.com>
> > > > | Closes: https://lore.kernel.org/oe-kbuild-all/202504081841.jbL7yL0r-lkp@intel.com/
> > > >
> > > > All warnings (new ones prefixed by >>):
> > > >
> > > > >> mm/swapfile.c:1517:20: warning: function 'swap_is_last_ref' is not needed and will not be emitted [-
> Wunneeded-internal-declaration]
> > >
> > > It's annoying that gcc warns about this.
> > >
> > > >     1517 | static inline bool swap_is_last_ref(unsigned char count)
> > > >          |                    ^~~~~~~~~~~~~~~~
> > > >    1 warning generated.
> > > >
> > > >
> > > > vim +/swap_is_last_ref +1517 mm/swapfile.c
> > > >
> > > >   1512
> > > >   1513	/*
> > > >   1514	 * Drop the last ref(1, SWAP_HAS_CACHE or SWAP_MAP_SHMEM) of swap entries,
> > > >   1515	 * caller have to ensure all entries belong to the same cgroup and cluster.
> > > >   1516	 */
> > > > > 1517	static inline bool swap_is_last_ref(unsigned char count)
> > > >   1518	{
> > > >   1519		return (count == SWAP_HAS_CACHE) || (count == 1) ||
> > > >   1520		       (count == SWAP_MAP_SHMEM);
> > > >   1521	}
> > > >   1522
> > >
> > > : static inline bool swap_is_last_ref(unsigned char count)
> > > : {
> > > : 	return (count == SWAP_HAS_CACHE) || (count == 1) ||
> > > : 	       (count == SWAP_MAP_SHMEM);
> > > : }
> > > :
> > > : static void swap_entries_free(struct swap_info_struct *si,
> > > : 			      struct swap_cluster_info *ci,
> > > : 			      swp_entry_t entry, unsigned int nr_pages)
> > > : {
> > > : 	unsigned long offset = swp_offset(entry);
> > > : 	unsigned char *map = si->swap_map + offset;
> > > : 	unsigned char *map_end = map + nr_pages;
> > > :
> > > : 	/* It should never free entries across different clusters */
> > > : 	VM_BUG_ON(ci != offset_to_cluster(si, offset + nr_pages - 1));
> > > : 	VM_BUG_ON(cluster_is_empty(ci));
> > > : 	VM_BUG_ON(ci->count < nr_pages);
> > > :
> > > : 	ci->count -= nr_pages;
> > > : 	do {
> > > : 		VM_BUG_ON(!swap_is_last_ref(*map));
> > >
> > >
> > > And it's annoying that VM_BUG_ON() doesn't reference its arg when
> > > CONFIG_DEBUG_VM=n.
> > >
> > > I'll add this:
> > >
> > > --- a/mm/swapfile.c~mm-swap-enable-swap_entry_range_free-to-drop-any-kind-of-last-ref-fix
> > > +++ a/mm/swapfile.c
> > > @@ -1514,11 +1514,13 @@ fallback:
> > >   * Drop the last ref(1, SWAP_HAS_CACHE or SWAP_MAP_SHMEM) of swap entries,
> > >   * caller have to ensure all entries belong to the same cgroup and cluster.
> > >   */
> > > +#ifdef CONFIG_DEBUG_VM
> > >  static inline bool swap_is_last_ref(unsigned char count)
> > >  {
> > >  	return (count == SWAP_HAS_CACHE) || (count == 1) ||
> > >  	       (count == SWAP_MAP_SHMEM);
> > >  }
> > > +#endif
> > >
> > >  static void swap_entries_free(struct swap_info_struct *si,
> > >  			      struct swap_cluster_info *ci,
> > > _
> >
> > This makes my build fails when CONFIG_DEBUG_VM is unset as below.
> >
> >     linux/mm/swapfile.c: In function ‘swap_entries_free’:
> >     linux/mm/swapfile.c:1565:28: error: implicit declaration of function ‘swap_is_last_ref’; did you mean
> ‘swap_is_last_map’? [-Werror=implicit-function-declaration]
> >      1565 |                 VM_BUG_ON(!swap_is_last_ref(*map));
> >           |                            ^~~~~~~~~~~~~~~~
> >     linux/include/linux/build_bug.h:30:63: note: in definition of macro ‘BUILD_BUG_ON_INVALID’
> >        30 | #define BUILD_BUG_ON_INVALID(e) ((void)(sizeof((__force long)(e))))
> >           |                                                               ^
> >     linux/mm/swapfile.c:1565:17: note: in expansion of macro ‘VM_BUG_ON’
> >      1565 |                 VM_BUG_ON(!swap_is_last_ref(*map));
> >           |                 ^~~~~~~~~
> >     cc1: some warnings being treated as errors
> >
> > Maybe we can open-code the condition, or add a macro implementation of
> > swap_is_last_ref() for !CONFIG_DEBUG_VM case?
> 
> I think we shold do the similar thing as this commit does:
> 6e8e04291d81 mm/zsmalloc: add __maybe_unused attribute for is_first_zpdesc()
> 
> Just add a __maybe_unused attribute to mute the warning. I haven't
> tested it because the reproducer failed in my side.
> 
> diff --git a/mm/swapfile.c b/mm/swapfile.c
> index 2887ed5d34cc..08b52346cc39 100644
> --- a/mm/swapfile.c
> +++ b/mm/swapfile.c
> @@ -1539,7 +1539,7 @@ static bool swap_entries_put_map_nr(struct swap_info_struct *si,
>   * Drop the last ref(1, SWAP_HAS_CACHE or SWAP_MAP_SHMEM) of swap entries,
>   * caller have to ensure all entries belong to the same cgroup and cluster.
>   */
> -static inline bool swap_is_last_ref(unsigned char count)
> +static inline bool __maybe_unused swap_is_last_ref(unsigned char count)
>  {
>  	return (count == SWAP_HAS_CACHE) || (count == 1) ||
>  	       (count == SWAP_MAP_SHMEM);
>