[rfc, PATCH v1 1/1] hrtimers: Refactor hrtimer_clock_to_base_table initialisation

[rfc, PATCH v1 1/1] hrtimers: Refactor hrtimer_clock_to_base_table initialisation

[Andy Shevchenko]

Clang complains about overlapped initialisers in the
hrtimer_clock_to_base_table definition. With `make W=1` and
CONFIG_WERROR=y (which is default nowadays) this breaks
the build:

  CC      kernel/time/hrtimer.o
kernel/time/hrtimer.c:124:21: error: initializer overrides prior initialization of this subobject [-Werror,-Winitializer-overrides]
  124 |         [CLOCK_REALTIME]        = HRTIMER_BASE_REALTIME,

kernel/time/hrtimer.c:122:27: note: previous initialization is here
  122 |         [0 ... MAX_CLOCKS - 1]  = HRTIMER_MAX_CLOCK_BASES,

(and similar for CLOCK_MONOTONIC, CLOCK_BOOTTIME, and CLOCK_TAI).

Refactor hrtimer_clock_to_base_table initialisation to make
the compiler happy.

Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
---

TBH, I don't like much this solution as it diminishes the point of that
override to be there in the first place. I haven't found better alternatives
as they may be too intrusive. Another one might be to remove this table,
but in such case the replacement might add latency to some cases (although
I haven't checked the generated code for, for example, switch-case approach).
In any case this is an rfc, I would be glad if somebody fixes this in a better
way.

 kernel/time/hrtimer.c | 20 ++++++++++++++------
 1 file changed, 14 insertions(+), 6 deletions(-)

diff --git a/kernel/time/hrtimer.c b/kernel/time/hrtimer.c
index deb1aa32814e..8ec6ff36b9ad 100644
--- a/kernel/time/hrtimer.c
+++ b/kernel/time/hrtimer.c
@@ -118,13 +118,21 @@ DEFINE_PER_CPU(struct hrtimer_cpu_base, hrtimer_bases) =
 };
 
 static const int hrtimer_clock_to_base_table[MAX_CLOCKS] = {
-	/* Make sure we catch unsupported clockids */
-	[0 ... MAX_CLOCKS - 1]	= HRTIMER_MAX_CLOCK_BASES,
+	[CLOCK_REALTIME]			= HRTIMER_BASE_REALTIME,
+	[CLOCK_MONOTONIC]			= HRTIMER_BASE_MONOTONIC,
+	[CLOCK_PROCESS_CPUTIME_ID]		= HRTIMER_MAX_CLOCK_BASES,
+	[CLOCK_THREAD_CPUTIME_ID]		= HRTIMER_MAX_CLOCK_BASES,
+	[CLOCK_MONOTONIC_RAW]			= HRTIMER_MAX_CLOCK_BASES,
+	[CLOCK_REALTIME_COARSE]			= HRTIMER_MAX_CLOCK_BASES,
+	[CLOCK_MONOTONIC_COARSE]		= HRTIMER_MAX_CLOCK_BASES,
+	[CLOCK_BOOTTIME]			= HRTIMER_BASE_BOOTTIME,
+	[CLOCK_REALTIME_ALARM]			= HRTIMER_MAX_CLOCK_BASES,
+	[CLOCK_BOOTTIME_ALARM]			= HRTIMER_MAX_CLOCK_BASES,
+	[CLOCK_SGI_CYCLE]			= HRTIMER_MAX_CLOCK_BASES,
+	[CLOCK_TAI]				= HRTIMER_BASE_TAI,
 
-	[CLOCK_REALTIME]	= HRTIMER_BASE_REALTIME,
-	[CLOCK_MONOTONIC]	= HRTIMER_BASE_MONOTONIC,
-	[CLOCK_BOOTTIME]	= HRTIMER_BASE_BOOTTIME,
-	[CLOCK_TAI]		= HRTIMER_BASE_TAI,
+	/* Make sure we catch unsupported clockids */
+	[CLOCK_TAI + 1 ... MAX_CLOCKS - 1]	= HRTIMER_MAX_CLOCK_BASES,
 };
 
 static inline bool hrtimer_base_is_online(struct hrtimer_cpu_base *base)
-- 
2.45.1.3035.g276e886db78b

Re: [rfc, PATCH v1 1/1] hrtimers: Refactor hrtimer_clock_to_base_table initialisation

[Andy Shevchenko]

On Mon, Feb 10, 2025 at 10:26:02AM +0200, Andy Shevchenko wrote:
> Clang complains about overlapped initialisers in the
> hrtimer_clock_to_base_table definition. With `make W=1` and
> CONFIG_WERROR=y (which is default nowadays) this breaks
> the build:
> 
>   CC      kernel/time/hrtimer.o
> kernel/time/hrtimer.c:124:21: error: initializer overrides prior initialization of this subobject [-Werror,-Winitializer-overrides]
>   124 |         [CLOCK_REALTIME]        = HRTIMER_BASE_REALTIME,
> 
> kernel/time/hrtimer.c:122:27: note: previous initialization is here
>   122 |         [0 ... MAX_CLOCKS - 1]  = HRTIMER_MAX_CLOCK_BASES,
> 
> (and similar for CLOCK_MONOTONIC, CLOCK_BOOTTIME, and CLOCK_TAI).
> 
> Refactor hrtimer_clock_to_base_table initialisation to make
> the compiler happy.

For the record, GCC also fails here.

-- 
With Best Regards,
Andy Shevchenko

Re: [rfc, PATCH v1 1/1] hrtimers: Refactor hrtimer_clock_to_base_table initialisation

[Thomas Gleixner]

On Mon, Feb 10 2025 at 10:26, Andy Shevchenko wrote:
> Clang complains about overlapped initialisers in the
> hrtimer_clock_to_base_table definition. With `make W=1` and
> CONFIG_WERROR=y (which is default nowadays) this breaks
> the build:
>
>   CC      kernel/time/hrtimer.o
> kernel/time/hrtimer.c:124:21: error: initializer overrides prior initialization of this subobject [-Werror,-Winitializer-overrides]
>   124 |         [CLOCK_REALTIME]        = HRTIMER_BASE_REALTIME,
>
> kernel/time/hrtimer.c:122:27: note: previous initialization is here
>   122 |         [0 ... MAX_CLOCKS - 1]  = HRTIMER_MAX_CLOCK_BASES,
>
> (and similar for CLOCK_MONOTONIC, CLOCK_BOOTTIME, and CLOCK_TAI).
>
> Refactor hrtimer_clock_to_base_table initialisation to make
> the compiler happy.
>
> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> ---
>
> TBH, I don't like much this solution as it diminishes the point of that
> override to be there in the first place. I haven't found better alternatives
> as they may be too intrusive. Another one might be to remove this table,
> but in such case the replacement might add latency to some cases (although
> I haven't checked the generated code for, for example, switch-case approach).

The only place this table is used is in __hrtimer_init(), which is not a
hot-path. The four resulting comparisons are probably not even noticable.

Thanks,

        tglx