power: reset: ltc2952: fix float conversion error

PATCH In review Arnd Bergmann 3 participants 6 messages Dec 4, 2021 – Dec 7, 2021 System reset/shutdown drivers

lore

Share & tools

lore

Arnd Bergmann

author

2021-12-04 22:02 UTC #0 lore raw

PATCH 1/1 power: reset: ltc2952: fix float conversion error

From: Arnd Bergmann <[email protected]>

clang-14 does not like the way this driver converts a 'long double' to an integer when the target architecture disables floating point support:

drivers/power/reset/ltc2952-poweroff.c:162:28: error: expression requires 'long double' type support, but target 'x86_64-unknown-linux' does not support it

        data->wde_interval = 300L * 1E6L;
                                  ^

Turn this into pure integer math and make it more readable at the same time using the NSEC_PER_MSEC macro instead.

Fixes 6647156c00cc ("power: reset: add LTC2952 poweroff driver")

Signed-off-by Arnd Bergmann <[email protected]>

---
 drivers/power/reset/ltc2952-poweroff.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

drivers/power/reset/ltc2952-poweroff.c +2 -2

index fbb344353fe4..9fc88a9f244c 100644
@@ -159,8 +159,8 @@ static void ltc2952_poweroff_kill(void)
 
 static void ltc2952_poweroff_default(struct ltc2952_poweroff *data)
 {
-	data->wde_interval = 300L * 1E6L;
-	data->trigger_delay = ktime_set(2, 500L*1E6L);
+	data->wde_interval = 300 * NSEC_PER_MSEC;
+	data->trigger_delay = ktime_set(2, 500 * NSEC_PER_MSEC);
 
 	hrtimer_init(&data->timer_trigger, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
 	data->timer_trigger.function = ltc2952_poweroff_timer_trigger;

Signature · 1 line

2.29.2

Nick Desaulniers

Reviewed-by on v1 1/1

2021-12-07 18:41 UTC #1 lore raw

Quoted 8 lines · Arnd Bergmann

From: Arnd Bergmann <[email protected]>
clang-14 does not like the way this driver converts a 'long double' to an integer when the target architecture disables floating point support:
drivers/power/reset/ltc2952-poweroff.c:162:28: error: expression requires 'long double' type support, but target 'x86_64-unknown-linux' does not support it

Quoted 8 lines

        data->wde_interval = 300L * 1E6L;
                                  ^
Turn this into pure integer math and make it more readable at the same time using the NSEC_PER_MSEC macro instead.
Fixes 6647156c00cc ("power: reset: add LTC2952 poweroff driver")
Signed-off-by Arnd Bergmann <[email protected]>

Thanks for the patch! The change in clang-14 (ToT) that triggered this has had another patch on top reverting these diagnostics.
https://reviews.llvm.org/D114162

That said, this change is still worthwhile for the improved semantics, IMO.

Link https://github.com/ClangBuiltLinux/linux/issues/1497

Reviewed-by Nick Desaulniers <[email protected]>

Quoted 22 lines

---
 drivers/power/reset/ltc2952-poweroff.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

drivers/power/reset/ltc2952-poweroff.c +2 -3

index fbb344353fe4..9fc88a9f244c 100644
@@ -159,8 +159,8 @@ static void ltc2952_poweroff_kill(void)
 
 static void ltc2952_poweroff_default(struct ltc2952_poweroff *data)
 {
-       data->wde_interval = 300L * 1E6L;
-       data->trigger_delay = ktime_set(2, 500L*1E6L);
+       data->wde_interval = 300 * NSEC_PER_MSEC;
+       data->trigger_delay = ktime_set(2, 500 * NSEC_PER_MSEC);
 
        hrtimer_init(&data->timer_trigger, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
        data->timer_trigger.function = ltc2952_poweroff_timer_trigger;
--

2.29.2

Signature · 2 lines

Thanks,
~Nick Desaulniers

Nick Desaulniers

2021-12-07 19:01 UTC #2 lore raw

Quoted 19 lines · Nick Desaulniers

Quoted 8 lines · Arnd Bergmann
From: Arnd Bergmann <[email protected]>
clang-14 does not like the way this driver converts a 'long double' to an integer when the target architecture disables floating point support:
drivers/power/reset/ltc2952-poweroff.c:162:28: error: expression requires 'long double' type support, but target 'x86_64-unknown-linux' does not support it
Quoted 8 lines
        data->wde_interval = 300L * 1E6L;
                                  ^
Turn this into pure integer math and make it more readable at the same time using the NSEC_PER_MSEC macro instead.
Fixes 6647156c00cc ("power: reset: add LTC2952 poweroff driver")
Signed-off-by Arnd Bergmann <[email protected]>

Quoted 8 lines

Thanks for the patch! The change in clang-14 (ToT) that triggered this has had another patch on top reverting these diagnostics.
https://reviews.llvm.org/D114162
That said, this change is still worthwhile for the improved semantics, IMO.
Link https://github.com/ClangBuiltLinux/linux/issues/1497
Reviewed-by Nick Desaulniers <[email protected]>

Ah, a fix has already been picked up:
https://git.kernel.org/pub/scm/linux/kernel/git/sre/linux-power-supply.git/commit/?id=644106cdb89844be2496b21175b7c0c2e0fab381

Quoted 28 lines

Quoted 21 lines

---
 drivers/power/reset/ltc2952-poweroff.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

drivers/power/reset/ltc2952-poweroff.c +2 -3

index fbb344353fe4..9fc88a9f244c 100644
@@ -159,8 +159,8 @@ static void ltc2952_poweroff_kill(void)
 
 static void ltc2952_poweroff_default(struct ltc2952_poweroff *data)
 {
-       data->wde_interval = 300L * 1E6L;
-       data->trigger_delay = ktime_set(2, 500L*1E6L);
+       data->wde_interval = 300 * NSEC_PER_MSEC;
+       data->trigger_delay = ktime_set(2, 500 * NSEC_PER_MSEC);
 
        hrtimer_init(&data->timer_trigger, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
        data->timer_trigger.function = ltc2952_poweroff_timer_trigger;
--

Quoted 2 lines

2.29.2

--
Thanks,
~Nick Desaulniers

Signature · 2 lines

Thanks,
~Nick Desaulniers

Nathan Chancellor

2021-12-07 19:06 UTC #3 lore raw

In reply to Nick Desaulniers

Quoted 15 lines · Nick Desaulniers

Quoted 14 lines · Arnd Bergmann
From: Arnd Bergmann <[email protected]>
clang-14 does not like the way this driver converts a 'long double' to an integer when the target architecture disables floating point support:
drivers/power/reset/ltc2952-poweroff.c:162:28: error: expression requires 'long double' type support, but target 'x86_64-unknown-linux' does not support it
        data->wde_interval = 300L * 1E6L;
                                  ^
Turn this into pure integer math and make it more readable at the same time using the NSEC_PER_MSEC macro instead.

Quoted 6 lines

Quoted 2 lines
Fixes 6647156c00cc ("power: reset: add LTC2952 poweroff driver")
Signed-off-by Arnd Bergmann <[email protected]>
Thanks for the patch! The change in clang-14 (ToT) that triggered this has had another patch on top reverting these diagnostics.
https://reviews.llvm.org/D114162

I can still reproduce this failure on current ToT (33e3554ea33d) on x86_64 allmodconfig; it seems like that patch only fixes it for 32-bit x86.

Quoted 4 lines

That said, this change is still worthwhile for the improved semantics, IMO.
Link https://github.com/ClangBuiltLinux/linux/issues/1497
Reviewed-by Nick Desaulniers <[email protected]>

My patch appears to be accepted in Sebastian's branch already, it seems like it just needs to be sent to Linus (it also looks like that branch doesn't flow into -next, it probably should):

https://git.kernel.org/sre/linux-power-supply/c/644106cdb89844be2496b21175b7c0c2e0fab381

Cheers,
Nathan

Quoted 27 lines

Quoted 22 lines

---
 drivers/power/reset/ltc2952-poweroff.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

drivers/power/reset/ltc2952-poweroff.c +2 -3

index fbb344353fe4..9fc88a9f244c 100644
@@ -159,8 +159,8 @@ static void ltc2952_poweroff_kill(void)
 
 static void ltc2952_poweroff_default(struct ltc2952_poweroff *data)
 {
-       data->wde_interval = 300L * 1E6L;
-       data->trigger_delay = ktime_set(2, 500L*1E6L);
+       data->wde_interval = 300 * NSEC_PER_MSEC;
+       data->trigger_delay = ktime_set(2, 500 * NSEC_PER_MSEC);
 
        hrtimer_init(&data->timer_trigger, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
        data->timer_trigger.function = ltc2952_poweroff_timer_trigger;
--

2.29.2

Signature · 2 lines

Thanks,
~Nick Desaulniers

Nick Desaulniers

2021-12-07 19:16 UTC #4 lore raw

(Minus most folks, plus Phoebe)

Quoted 20 lines · Nathan Chancellor

Quoted 18 lines · Nick Desaulniers
Quoted 16 lines · Arnd Bergmann
From: Arnd Bergmann <[email protected]>
clang-14 does not like the way this driver converts a 'long double' to an integer when the target architecture disables floating point support:
drivers/power/reset/ltc2952-poweroff.c:162:28: error: expression requires 'long double' type support, but target 'x86_64-unknown-linux' does not support it
        data->wde_interval = 300L * 1E6L;
                                  ^
Turn this into pure integer math and make it more readable at the same time using the NSEC_PER_MSEC macro instead.
Fixes 6647156c00cc ("power: reset: add LTC2952 poweroff driver")
Signed-off-by Arnd Bergmann <[email protected]>

Quoted 7 lines

Quoted 3 lines
Thanks for the patch! The change in clang-14 (ToT) that triggered this has had another patch on top reverting these diagnostics.
https://reviews.llvm.org/D114162
I can still reproduce this failure on current ToT (33e3554ea33d) on x86_64 allmodconfig; it seems like that patch only fixes it for 32-bit x86.

As a heads up Phoebe, in case there's any interest in doing something similar to https://reviews.llvm.org/D114162 but for 64b.

Quoted 43 lines

Quoted 5 lines
That said, this change is still worthwhile for the improved semantics, IMO.
Link https://github.com/ClangBuiltLinux/linux/issues/1497
Reviewed-by Nick Desaulniers <[email protected]>
My patch appears to be accepted in Sebastian's branch already, it seems like it just needs to be sent to Linus (it also looks like that branch doesn't flow into -next, it probably should):
https://git.kernel.org/sre/linux-power-supply/c/644106cdb89844be2496b21175b7c0c2e0fab381
Cheers,
Nathan
Quoted 27 lines
Quoted 20 lines
---
 drivers/power/reset/ltc2952-poweroff.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/power/reset/ltc2952-poweroff.c +2 -3
index fbb344353fe4..9fc88a9f244c 100644
@@ -159,8 +159,8 @@ static void ltc2952_poweroff_kill(void)
 
 static void ltc2952_poweroff_default(struct ltc2952_poweroff *data)
 {
-       data->wde_interval = 300L * 1E6L;
-       data->trigger_delay = ktime_set(2, 500L*1E6L);
+       data->wde_interval = 300 * NSEC_PER_MSEC;
+       data->trigger_delay = ktime_set(2, 500 * NSEC_PER_MSEC);
 
        hrtimer_init(&data->timer_trigger, CLOCK_MONOTONIC, HRTIMER_MODE_REL);
        data->timer_trigger.function = ltc2952_poweroff_timer_trigger;
--
Quoted 2 lines
2.29.2
--
Thanks,
~Nick Desaulniers

Signature · 2 lines

Thanks,
~Nick Desaulniers

Nathan Chancellor

2021-12-07 19:21 UTC #5 lore raw

Quoted 27 lines · Nick Desaulniers

(Minus most folks, plus Phoebe)
Quoted 24 lines · Nathan Chancellor
Quoted 21 lines · Nick Desaulniers
Quoted 8 lines · Arnd Bergmann
From: Arnd Bergmann <[email protected]>
clang-14 does not like the way this driver converts a 'long double' to an integer when the target architecture disables floating point support:
drivers/power/reset/ltc2952-poweroff.c:162:28: error: expression requires 'long double' type support, but target 'x86_64-unknown-linux' does not support it
Quoted 8 lines
        data->wde_interval = 300L * 1E6L;
                                  ^
Turn this into pure integer math and make it more readable at the same time using the NSEC_PER_MSEC macro instead.
Fixes 6647156c00cc ("power: reset: add LTC2952 poweroff driver")
Signed-off-by Arnd Bergmann <[email protected]>
Thanks for the patch! The change in clang-14 (ToT) that triggered this has had another patch on top reverting these diagnostics.
https://reviews.llvm.org/D114162

Quoted 6 lines

Quoted 3 lines
I can still reproduce this failure on current ToT (33e3554ea33d) on x86_64 allmodconfig; it seems like that patch only fixes it for 32-bit x86.
As a heads up Phoebe, in case there's any interest in doing something similar to https://reviews.llvm.org/D114162 but for 64b.

Phoebe pointed out on our issue tracker that GCC will error when long double is used with -mno-80837 as well:

https://github.com/ClangBuiltLinux/linux/issues/1497#issuecomment-983245202

It just appears that GCC is able to avoid these errors in the particular cases that the kernel triggers in clang:

https://github.com/ClangBuiltLinux/linux/issues/1497#issuecomment-983739365

Cheers,
Nathan