mm/mapping_dirty_helpers: guard hugepage pud's usage
authorZack Rusin <zackr@vmware.com>
Fri, 16 Apr 2021 22:46:18 +0000 (15:46 -0700)
committerLinus Torvalds <torvalds@linux-foundation.org>
Fri, 16 Apr 2021 23:10:37 +0000 (16:10 -0700)
Mapping dirty helpers have, so far, been only used on X86, but a port of
vmwgfx to ARM64 exposed a problem which results in a compilation error
on ARM64 systems:

  mm/mapping_dirty_helpers.c: In function `wp_clean_pud_entry':
  mm/mapping_dirty_helpers.c:172:32: error: implicit declaration of function `pud_dirty'; did you mean `pmd_dirty'? [-Werror=implicit-function-declaration]

This is due to the fact that mapping_dirty_helpers code assumes that
pud_dirty is always defined, which is not the case for architectures
that don't define CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD.

ARM64 arch is a little inconsistent when it comes to PUD hugepage
helpers, e.g. it defines pud_young but not pud_dirty but regardless of
that the core kernel code shouldn't assume that any of the PUD hugepage
helpers are available unless CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
is defined.  This prevents compilation errors whenever one of the
drivers is ported to new architectures.

Link: https://lkml.kernel.org/r/20210409165151.694574-1-zackr@vmware.com
Signed-off-by: Zack Rusin <zackr@vmware.com>
Reviewed-by: Thomas Hellstrm (Intel) <thomas_os@shipmail.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
mm/mapping_dirty_helpers.c

index b59054e..b890854 100644 (file)
@@ -165,10 +165,12 @@ static int wp_clean_pud_entry(pud_t *pud, unsigned long addr, unsigned long end,
                return 0;
        }
 
+#ifdef CONFIG_HAVE_ARCH_TRANSPARENT_HUGEPAGE_PUD
        /* Huge pud */
        walk->action = ACTION_CONTINUE;
        if (pud_trans_huge(pudval) || pud_devmap(pudval))
                WARN_ON(pud_write(pudval) || pud_dirty(pudval));
+#endif
 
        return 0;
 }