[PATCH] clocksource/drivers/arm_arch_timer: Force inlining of erratum_set_next_event_generic()

[PATCH] clocksource/drivers/arm_arch_timer: Force inlining of erratum_set_next_event_generic()

[Marc Zyngier]

With some specific kernel configuration and Clang, the kernel fails
to like with something like:

ld.lld: error: undefined symbol: __compiletime_assert_200
>>> referenced by arch_timer.h:156 (./arch/arm64/include/asm/arch_timer.h:156)
>>>               clocksource/arm_arch_timer.o:(erratum_set_next_event_generic) in archive drivers/built-in.a

ld.lld: error: undefined symbol: __compiletime_assert_197
>>> referenced by arch_timer.h:133 (./arch/arm64/include/asm/arch_timer.h:133)
>>>               clocksource/arm_arch_timer.o:(erratum_set_next_event_generic) in archive drivers/built-in.a
make: *** [Makefile:1161: vmlinux] Error 1

These are due to the BUILD_BUG() macros contained in the low-level
accessors (arch_timer_reg_{write,read}_cp15) being emitted, as the
access type wasn't known at compile time.

Fix this by making erratum_set_next_event_generic() __force_inline,
resulting in the 'access' parameter to be resolved at compile time,
similarly to what is already done for set_next_event().

Fixes: 4775bc63f880 ("Add build-time guards for unhandled register accesses")
Reported-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
Cc: Sami Tolvanen <samitolvanen@google.com>
Cc: Nick Desaulniers <ndesaulniers@google.com>
---
 drivers/clocksource/arm_arch_timer.c | 9 +++++++--
 1 file changed, 7 insertions(+), 2 deletions(-)

diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
index 9a04eacc4412..1ecd52f903b8 100644
--- a/drivers/clocksource/arm_arch_timer.c
+++ b/drivers/clocksource/arm_arch_timer.c
@@ -394,8 +394,13 @@ EXPORT_SYMBOL_GPL(timer_unstable_counter_workaround);
 
 static atomic_t timer_unstable_counter_workaround_in_use = ATOMIC_INIT(0);
 
-static void erratum_set_next_event_generic(const int access, unsigned long evt,
-						struct clock_event_device *clk)
+/*
+ * Force the inlining of this function so that the register accesses
+ * can be themselves correctly inlined.
+ */
+static __always_inline
+void erratum_set_next_event_generic(const int access, unsigned long evt,
+				    struct clock_event_device *clk)
 {
 	unsigned long ctrl;
 	u64 cval;
-- 
2.30.2


_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] clocksource/drivers/arm_arch_timer: Force inlining of erratum_set_next_event_generic()

[Sami Tolvanen]

On Wed, Nov 17, 2021 at 3:35 AM Marc Zyngier <maz@kernel.org> wrote:
>
> With some specific kernel configuration and Clang, the kernel fails
> to like with something like:
>
> ld.lld: error: undefined symbol: __compiletime_assert_200
> >>> referenced by arch_timer.h:156 (./arch/arm64/include/asm/arch_timer.h:156)
> >>>               clocksource/arm_arch_timer.o:(erratum_set_next_event_generic) in archive drivers/built-in.a
>
> ld.lld: error: undefined symbol: __compiletime_assert_197
> >>> referenced by arch_timer.h:133 (./arch/arm64/include/asm/arch_timer.h:133)
> >>>               clocksource/arm_arch_timer.o:(erratum_set_next_event_generic) in archive drivers/built-in.a
> make: *** [Makefile:1161: vmlinux] Error 1
>
> These are due to the BUILD_BUG() macros contained in the low-level
> accessors (arch_timer_reg_{write,read}_cp15) being emitted, as the
> access type wasn't known at compile time.
>
> Fix this by making erratum_set_next_event_generic() __force_inline,
> resulting in the 'access' parameter to be resolved at compile time,
> similarly to what is already done for set_next_event().
>
> Fixes: 4775bc63f880 ("Add build-time guards for unhandled register accesses")
> Reported-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Signed-off-by: Marc Zyngier <maz@kernel.org>
> Cc: Mark Rutland <mark.rutland@arm.com>
> Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
> Cc: Sami Tolvanen <samitolvanen@google.com>
> Cc: Nick Desaulniers <ndesaulniers@google.com>
> ---
>  drivers/clocksource/arm_arch_timer.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)

Thanks, Marc. I can confirm that this fixes the build issue for me.

Tested-by: Sami Tolvanen <samitolvanen@google.com>

Sami

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel

Re: [PATCH] clocksource/drivers/arm_arch_timer: Force inlining of erratum_set_next_event_generic()

[Nathan Chancellor]

On Wed, Nov 17, 2021 at 11:35:32AM +0000, Marc Zyngier wrote:
> With some specific kernel configuration and Clang, the kernel fails
> to like with something like:

Specifically, CONFIG_CFI_CLANG=y from what I could tell from looking
into this: https://github.com/ClangBuiltLinux/linux/issues/1503

> ld.lld: error: undefined symbol: __compiletime_assert_200
> >>> referenced by arch_timer.h:156 (./arch/arm64/include/asm/arch_timer.h:156)
> >>>               clocksource/arm_arch_timer.o:(erratum_set_next_event_generic) in archive drivers/built-in.a
> 
> ld.lld: error: undefined symbol: __compiletime_assert_197
> >>> referenced by arch_timer.h:133 (./arch/arm64/include/asm/arch_timer.h:133)
> >>>               clocksource/arm_arch_timer.o:(erratum_set_next_event_generic) in archive drivers/built-in.a
> make: *** [Makefile:1161: vmlinux] Error 1
> 
> These are due to the BUILD_BUG() macros contained in the low-level
> accessors (arch_timer_reg_{write,read}_cp15) being emitted, as the
> access type wasn't known at compile time.
> 
> Fix this by making erratum_set_next_event_generic() __force_inline,
> resulting in the 'access' parameter to be resolved at compile time,
> similarly to what is already done for set_next_event().
> 
> Fixes: 4775bc63f880 ("Add build-time guards for unhandled register accesses")

It looks like the prefix was removed? Shouldn't it be:

Fixes: 4775bc63f880 ("clocksource/arm_arch_timer: Add build-time guards for unhandled register accesses")

> Reported-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
> Signed-off-by: Marc Zyngier <maz@kernel.org>
> Cc: Mark Rutland <mark.rutland@arm.com>
> Cc: Daniel Lezcano <daniel.lezcano@linaro.org>
> Cc: Sami Tolvanen <samitolvanen@google.com>
> Cc: Nick Desaulniers <ndesaulniers@google.com>

Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Tested-by: Nathan Chancellor <nathan@kernel.org>

> ---
>  drivers/clocksource/arm_arch_timer.c | 9 +++++++--
>  1 file changed, 7 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/clocksource/arm_arch_timer.c b/drivers/clocksource/arm_arch_timer.c
> index 9a04eacc4412..1ecd52f903b8 100644
> --- a/drivers/clocksource/arm_arch_timer.c
> +++ b/drivers/clocksource/arm_arch_timer.c
> @@ -394,8 +394,13 @@ EXPORT_SYMBOL_GPL(timer_unstable_counter_workaround);
>  
>  static atomic_t timer_unstable_counter_workaround_in_use = ATOMIC_INIT(0);
>  
> -static void erratum_set_next_event_generic(const int access, unsigned long evt,
> -						struct clock_event_device *clk)
> +/*
> + * Force the inlining of this function so that the register accesses
> + * can be themselves correctly inlined.
> + */
> +static __always_inline
> +void erratum_set_next_event_generic(const int access, unsigned long evt,
> +				    struct clock_event_device *clk)
>  {
>  	unsigned long ctrl;
>  	u64 cval;
> -- 
> 2.30.2
> 

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel