WARN_ON() is buggy for 32 bit systems
Christophe Leroy
christophe.leroy at csgroup.eu
Wed Jan 26 23:21:49 AEDT 2022
Hi Dan,
Le 26/01/2022 à 12:56, Dan Carpenter a écrit :
>
> Hi Michael,
>
> Commit e432fe97f3e5 ("powerpc/bug: Cast to unsigned long before passing
> to inline asm") breaks WARN_ON() for 32 bit systems.
I think you missed commit db87a7199229 ("powerpc/bug: Remove specific
powerpc BUG_ON() and WARN_ON() on PPC32")
>
> arch/powerpc/include/asm/bug.h
> 109 #define WARN_ON(x) ({ \
> 110 bool __ret_warn_on = false; \
> 111 do { \
> 112 if (__builtin_constant_p((x))) { \
> 113 if (!(x)) \
> 114 break; \
> 115 __WARN(); \
> 116 __ret_warn_on = true; \
> 117 } else { \
> 118 __label__ __label_warn_on; \
> 119 \
> 120 WARN_ENTRY(PPC_TLNEI " %4, 0", \
> 121 BUGFLAG_WARNING | BUGFLAG_TAINT(TAINT_WARN), \
> 122 __label_warn_on, \
> 123 "r" ((__force long)(x))); \
> ^^^^
> If the code is "if (WARN_ON(some_u64)) {" then the cast to long will
> truncate away the high bits so it's wrong. (Or at least that's how it
> works on x86, I'm working on a work around for Smatch to be able to
> parse this WARN_ON(). I don't know anything about PowerPC.)
The code is enclosed in a #ifdef CONFIG_PPC64, it is not used for PPC32:
/arch/powerpc/include/asm/bug.h
99 #ifdef CONFIG_PPC64
100 #define BUG_ON(x) do { \
...
109 #define WARN_ON(x) ({ \
...
132 #define HAVE_ARCH_BUG_ON
133 #define HAVE_ARCH_WARN_ON
134 #endif
Christophe
>
> 124 break; \
> 125 __label_warn_on: \
> 126 __ret_warn_on = true; \
> 127 } \
> 128 } while (0); \
> 129 unlikely(__ret_warn_on); \
> 130 })
>
> regards,
> dan carpenter
>
More information about the Linuxppc-dev
mailing list