mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
* [PATCH] delayacct: fix uapi timespec64 definition
@ 2026-02-02  9:59 Arnd Bergmann
  2026-02-02 13:08 ` wang.yaxin
                   ` (2 more replies)
  0 siblings, 3 replies; 5+ messages in thread
From: Arnd Bergmann @ 2026-02-02  9:59 UTC (permalink / raw)
  To: Balbir Singh, Yang Yang, Wang Yaxin, Andrew Morton
  Cc: Arnd Bergmann, xu xin, Kun Jiang, linux-kernel

From: Arnd Bergmann <arnd@arndb.de>

The custom definition of 'struct timespec64' is incompatible with
both the kernel's internal definition and the glibc type, at least
on big-endian targets that have the tv_nsec field in a different
place, and the definition clashes with any userspace that also
defines a timespec64 structure.

Running the header check with -Wpadding enabled produces this output
that warns about the incorrect padding:

usr/include/linux/taskstats.h:25:1: error: padding struct size to alignment boundary with 4 bytes [-Werror=padded]

Remove the hack and instead use the regular __kernel_timespec
type that is meant to be used in uapi definitions.

Fixes: 29b63f6eff0e ("delayacct: add timestamp of delay max")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 include/uapi/linux/taskstats.h | 27 +++++++++------------------
 kernel/delayacct.c             |  6 ++++--
 2 files changed, 13 insertions(+), 20 deletions(-)

diff --git a/include/uapi/linux/taskstats.h b/include/uapi/linux/taskstats.h
index a66b2d84a310..cc7fccd57979 100644
--- a/include/uapi/linux/taskstats.h
+++ b/include/uapi/linux/taskstats.h
@@ -18,16 +18,7 @@
 #define _LINUX_TASKSTATS_H
 
 #include <linux/types.h>
-#ifdef __KERNEL__
-#include <linux/time64.h>
-#else
-#ifndef _LINUX_TIME64_H
-struct timespec64 {
-	__s64   tv_sec;         /* seconds */
-	long    tv_nsec;        /* nanoseconds */
-};
-#endif
-#endif
+#include <linux/time_types.h>
 
 /* Format for per-task data returned to userland when
  *	- a task exits
@@ -251,14 +242,14 @@ struct taskstats {
 	__u64	irq_delay_min;
 
 	/*v17: delay max timestamp record*/
-	struct timespec64 cpu_delay_max_ts;
-	struct timespec64 blkio_delay_max_ts;
-	struct timespec64 swapin_delay_max_ts;
-	struct timespec64 freepages_delay_max_ts;
-	struct timespec64 thrashing_delay_max_ts;
-	struct timespec64 compact_delay_max_ts;
-	struct timespec64 wpcopy_delay_max_ts;
-	struct timespec64 irq_delay_max_ts;
+	struct __kernel_timespec cpu_delay_max_ts;
+	struct __kernel_timespec blkio_delay_max_ts;
+	struct __kernel_timespec swapin_delay_max_ts;
+	struct __kernel_timespec freepages_delay_max_ts;
+	struct __kernel_timespec thrashing_delay_max_ts;
+	struct __kernel_timespec compact_delay_max_ts;
+	struct __kernel_timespec wpcopy_delay_max_ts;
+	struct __kernel_timespec irq_delay_max_ts;
 } __uapi_arch_align;
 
 
diff --git a/kernel/delayacct.c b/kernel/delayacct.c
index d58ffc63bcba..2e55c493c98b 100644
--- a/kernel/delayacct.c
+++ b/kernel/delayacct.c
@@ -18,7 +18,8 @@
 do { \
 	d->type##_delay_max = tsk->delays->type##_delay_max; \
 	d->type##_delay_min = tsk->delays->type##_delay_min; \
-	d->type##_delay_max_ts = tsk->delays->type##_delay_max_ts; \
+	d->type##_delay_max_ts.tv_sec = tsk->delays->type##_delay_max_ts.tv_sec; \
+	d->type##_delay_max_ts.tv_nsec = tsk->delays->type##_delay_max_ts.tv_nsec; \
 	tmp = d->type##_delay_total + tsk->delays->type##_delay; \
 	d->type##_delay_total = (tmp < d->type##_delay_total) ? 0 : tmp; \
 	d->type##_count += tsk->delays->type##_count; \
@@ -175,7 +176,8 @@ int delayacct_add_tsk(struct taskstats *d, struct task_struct *tsk)
 
 	d->cpu_delay_max = tsk->sched_info.max_run_delay;
 	d->cpu_delay_min = tsk->sched_info.min_run_delay;
-	d->cpu_delay_max_ts = tsk->sched_info.max_run_delay_ts;
+	d->cpu_delay_max_ts.tv_sec = tsk->sched_info.max_run_delay_ts.tv_sec;
+	d->cpu_delay_max_ts.tv_nsec = tsk->sched_info.max_run_delay_ts.tv_nsec;
 	tmp = (s64)d->cpu_delay_total + t2;
 	d->cpu_delay_total = (tmp < (s64)d->cpu_delay_total) ? 0 : tmp;
 	tmp = (s64)d->cpu_run_virtual_total + t3;
-- 
2.39.5


^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] delayacct: fix uapi timespec64 definition
  2026-02-02  9:59 [PATCH] delayacct: fix uapi timespec64 definition Arnd Bergmann
@ 2026-02-02 13:08 ` wang.yaxin
  2026-02-02 20:37 ` Andrew Morton
  2026-02-13  8:44 ` Geert Uytterhoeven
  2 siblings, 0 replies; 5+ messages in thread
From: wang.yaxin @ 2026-02-02 13:08 UTC (permalink / raw)
  To: arnd
  Cc: bsingharora, yang.yang29, akpm, arnd, xu.xin16, jiang.kun2, linux-kernel


[-- Attachment #1.1.1: Type: text/plain, Size: 987 bytes --]

>From: Arnd Bergmann <arnd@arndb.de>
>
>The custom definition of 'struct timespec64' is incompatible with
>both the kernel's internal definition and the glibc type, at least
>on big-endian targets that have the tv_nsec field in a different
>place, and the definition clashes with any userspace that also
>defines a timespec64 structure.
>
>Running the header check with -Wpadding enabled produces this output
>that warns about the incorrect padding:
>
>usr/include/linux/taskstats.h:25:1: error: padding struct size to alignment boundary with 4 bytes >[-Werror=padded]
>
>Remove the hack and instead use the regular __kernel_timespec
>type that is meant to be used in uapi definitions.
>
>Fixes: 29b63f6eff0e ("delayacct: add timestamp of delay max")
>Signed-off-by: Arnd Bergmann <arnd@arndb.de>

In uapi definition, __kernel_timespec is indeed a better solution compared to timespec64, as the former provides superior binary compatibility, nice job!

Thanks
Yaxin

[-- Attachment #1.1.2: Type: text/html , Size: 1223 bytes --]

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] delayacct: fix uapi timespec64 definition
  2026-02-02  9:59 [PATCH] delayacct: fix uapi timespec64 definition Arnd Bergmann
  2026-02-02 13:08 ` wang.yaxin
@ 2026-02-02 20:37 ` Andrew Morton
  2026-02-02 20:47   ` Arnd Bergmann
  2026-02-13  8:44 ` Geert Uytterhoeven
  2 siblings, 1 reply; 5+ messages in thread
From: Andrew Morton @ 2026-02-02 20:37 UTC (permalink / raw)
  To: Arnd Bergmann
  Cc: Balbir Singh, Yang Yang, Wang Yaxin, Arnd Bergmann, xu xin,
	Kun Jiang, linux-kernel

On Mon,  2 Feb 2026 10:59:00 +0100 Arnd Bergmann <arnd@kernel.org> wrote:

> From: Arnd Bergmann <arnd@arndb.de>
> 
> The custom definition of 'struct timespec64' is incompatible with
> both the kernel's internal definition and the glibc type, at least
> on big-endian targets that have the tv_nsec field in a different
> place, and the definition clashes with any userspace that also
> defines a timespec64 structure.
> 
> Running the header check with -Wpadding enabled produces this output
> that warns about the incorrect padding:
> 
> usr/include/linux/taskstats.h:25:1: error: padding struct size to alignment boundary with 4 bytes [-Werror=padded]
> 
> Remove the hack and instead use the regular __kernel_timespec
> type that is meant to be used in uapi definitions.
> 

Thanks.

> @@ -251,14 +242,14 @@ struct taskstats {
>  	__u64	irq_delay_min;
>  
>  	/*v17: delay max timestamp record*/
> -	struct timespec64 cpu_delay_max_ts;
> -	struct timespec64 blkio_delay_max_ts;
> -	struct timespec64 swapin_delay_max_ts;
> -	struct timespec64 freepages_delay_max_ts;
> -	struct timespec64 thrashing_delay_max_ts;
> -	struct timespec64 compact_delay_max_ts;
> -	struct timespec64 wpcopy_delay_max_ts;
> -	struct timespec64 irq_delay_max_ts;
> +	struct __kernel_timespec cpu_delay_max_ts;
> +	struct __kernel_timespec blkio_delay_max_ts;
> +	struct __kernel_timespec swapin_delay_max_ts;
> +	struct __kernel_timespec freepages_delay_max_ts;
> +	struct __kernel_timespec thrashing_delay_max_ts;
> +	struct __kernel_timespec compact_delay_max_ts;
> +	struct __kernel_timespec wpcopy_delay_max_ts;
> +	struct __kernel_timespec irq_delay_max_ts;
>  } __uapi_arch_align;

There's no __uapi_arch_align here in any tree I looked at?



^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] delayacct: fix uapi timespec64 definition
  2026-02-02 20:37 ` Andrew Morton
@ 2026-02-02 20:47   ` Arnd Bergmann
  0 siblings, 0 replies; 5+ messages in thread
From: Arnd Bergmann @ 2026-02-02 20:47 UTC (permalink / raw)
  To: Andrew Morton, Arnd Bergmann
  Cc: Balbir Singh, Yang Yang, Wang Yaxin, xu xin, Kun Jiang, linux-kernel

On Mon, Feb 2, 2026, at 21:37, Andrew Morton wrote:
>>  	/*v17: delay max timestamp record*/
>> -	struct timespec64 cpu_delay_max_ts;
>> -	struct timespec64 blkio_delay_max_ts;
>> -	struct timespec64 swapin_delay_max_ts;
>> -	struct timespec64 freepages_delay_max_ts;
>> -	struct timespec64 thrashing_delay_max_ts;
>> -	struct timespec64 compact_delay_max_ts;
>> -	struct timespec64 wpcopy_delay_max_ts;
>> -	struct timespec64 irq_delay_max_ts;
>> +	struct __kernel_timespec cpu_delay_max_ts;
>> +	struct __kernel_timespec blkio_delay_max_ts;
>> +	struct __kernel_timespec swapin_delay_max_ts;
>> +	struct __kernel_timespec freepages_delay_max_ts;
>> +	struct __kernel_timespec thrashing_delay_max_ts;
>> +	struct __kernel_timespec compact_delay_max_ts;
>> +	struct __kernel_timespec wpcopy_delay_max_ts;
>> +	struct __kernel_timespec irq_delay_max_ts;
>>  } __uapi_arch_align;
>
> There's no __uapi_arch_align here in any tree I looked at?

My mistake, that was a remnant of a work-in-progress
series that I used to detect unintended padding in uapi
structures.

Thanks for fixing up the patch and applying the correct
version.

      Arnd

^ permalink raw reply	[flat|nested] 5+ messages in thread

* Re: [PATCH] delayacct: fix uapi timespec64 definition
  2026-02-02  9:59 [PATCH] delayacct: fix uapi timespec64 definition Arnd Bergmann
  2026-02-02 13:08 ` wang.yaxin
  2026-02-02 20:37 ` Andrew Morton
@ 2026-02-13  8:44 ` Geert Uytterhoeven
  2 siblings, 0 replies; 5+ messages in thread
From: Geert Uytterhoeven @ 2026-02-13  8:44 UTC (permalink / raw)
  To: Arnd Bergmann
  Cc: Balbir Singh, Yang Yang, Wang Yaxin, Andrew Morton,
	Arnd Bergmann, xu xin, Kun Jiang, linux-kernel

On Mon, 2 Feb 2026 at 11:05, Arnd Bergmann <arnd@kernel.org> wrote:
> From: Arnd Bergmann <arnd@arndb.de>
>
> The custom definition of 'struct timespec64' is incompatible with
> both the kernel's internal definition and the glibc type, at least
> on big-endian targets that have the tv_nsec field in a different
> place, and the definition clashes with any userspace that also
> defines a timespec64 structure.
>
> Running the header check with -Wpadding enabled produces this output
> that warns about the incorrect padding:
>
> usr/include/linux/taskstats.h:25:1: error: padding struct size to alignment boundary with 4 bytes [-Werror=padded]
>
> Remove the hack and instead use the regular __kernel_timespec
> type that is meant to be used in uapi definitions.
>
> Fixes: 29b63f6eff0e ("delayacct: add timestamp of delay max")

JFTR, this commit was rebased, and is now commit 503efe850c7463a1
("delayacct: add timestamp of delay max") upstream.
Fixes: 503efe850c7463a1 ("delayacct: add timestamp of delay max")

> Signed-off-by: Arnd Bergmann <arnd@arndb.de>

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2026-02-13  8:44 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-02-02  9:59 [PATCH] delayacct: fix uapi timespec64 definition Arnd Bergmann
2026-02-02 13:08 ` wang.yaxin
2026-02-02 20:37 ` Andrew Morton
2026-02-02 20:47   ` Arnd Bergmann
2026-02-13  8:44 ` Geert Uytterhoeven

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®