[PATCH] mm/memremap: Mark folio_span_valid() as __maybe_unused

[PATCH] mm/memremap: Mark folio_span_valid() as __maybe_unused

[Nathan Chancellor]

When building without CONFIG_DEBUG_VM, clang warns:

  mm/memremap.c:495:13: error: function 'folio_span_valid' is not needed and will not be emitted [-Werror,-Wunneeded-internal-declaration]
  static bool folio_span_valid(struct dev_pagemap *pgmap, struct folio *folio,
              ^
  1 error generated.

folio_span_valid() is only used within a instance of VM_WARN_ON_ONCE(),
which evaluates to BUILD_BUG_ON_INVALID() with CONFIG_DEBUG_VM=n, which
ultimately resolves to sizeof(), which is fully resolved at compile
time. Basically, the warning is flagging that folio_span_valid() is
only used in a compile time context and will not be called at run time.

Since this is expected given the configuration, mark folio_span_valid()
as __maybe_unused so that there is no warning.

Fixes: 07108d5bfeeb ("fsdax: introduce pgmap_request_folios()")
Link: https://github.com/ClangBuiltLinux/linux/issues/1739
Signed-off-by: Nathan Chancellor <nathan@kernel.org>
---

I am aware the Fixes SHA is probably not stable but I figured I would
include it anyways.

 mm/memremap.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/mm/memremap.c b/mm/memremap.c
index 53fe30bb79bb..02b796749b72 100644
--- a/mm/memremap.c
+++ b/mm/memremap.c
@@ -492,8 +492,9 @@ void free_zone_device_page(struct page *page)
 	put_dev_pagemap(page->pgmap);
 }
 
-static bool folio_span_valid(struct dev_pagemap *pgmap, struct folio *folio,
-			     int nr_folios)
+static __maybe_unused bool folio_span_valid(struct dev_pagemap *pgmap,
+					    struct folio *folio,
+					    int nr_folios)
 {
 	unsigned long pfn_start, pfn_end;
 

base-commit: f5938bbbb79d48ca167e305c228dccbecea0309e
-- 
2.38.0

RE: [PATCH] mm/memremap: Mark folio_span_valid() as __maybe_unused

[Dan Williams]

Nathan Chancellor wrote:
> When building without CONFIG_DEBUG_VM, clang warns:
> 
>   mm/memremap.c:495:13: error: function 'folio_span_valid' is not needed and will not be emitted [-Werror,-Wunneeded-internal-declaration]
>   static bool folio_span_valid(struct dev_pagemap *pgmap, struct folio *folio,
>               ^
>   1 error generated.
> 
> folio_span_valid() is only used within a instance of VM_WARN_ON_ONCE(),
> which evaluates to BUILD_BUG_ON_INVALID() with CONFIG_DEBUG_VM=n, which
> ultimately resolves to sizeof(), which is fully resolved at compile
> time. Basically, the warning is flagging that folio_span_valid() is
> only used in a compile time context and will not be called at run time.
> 
> Since this is expected given the configuration, mark folio_span_valid()
> as __maybe_unused so that there is no warning.
> 
> Fixes: 07108d5bfeeb ("fsdax: introduce pgmap_request_folios()")
> Link: https://github.com/ClangBuiltLinux/linux/issues/1739
> Signed-off-by: Nathan Chancellor <nathan@kernel.org>
> ---
> 
> I am aware the Fixes SHA is probably not stable but I figured I would
> include it anyways.

This fix looks good to me, but I assume commit-ids are not stable until
the patch moves from mm-unstable to mm-stable. Andrew, do I have that
right?

Re: [PATCH] mm/memremap: Mark folio_span_valid() as __maybe_unused

[Andrew Morton]

On Tue, 18 Oct 2022 10:54:58 -0700 Dan Williams <dan.j.williams@intel.com> wrote:

> > ---
> > 
> > I am aware the Fixes SHA is probably not stable but I figured I would
> > include it anyways.
> 
> This fix looks good to me, but I assume commit-ids are not stable until
> the patch moves from mm-unstable to mm-stable. Andrew, do I have that
> right?

Yes, mm-unstable hashes change daily.  But knowing the title of the
patch which needs the fix is quite helpful.