[PATCH v2] time: Make sure jiffies_to_msecs() preserves non-zero time periods

[PATCH v2] time: Make sure jiffies_to_msecs() preserves non-zero time periods

[Geert Uytterhoeven]

For the common cases where 1000 is a multiple of HZ, or HZ is a multiple
of 1000, jiffies_to_msecs() never returns zero when passed a non-zero
time period.

However, if HZ > 1000 and not an integer multiple of 1000 (e.g. 1024 or
1200, as used on alpha and DECstation), jiffies_to_msecs() may return
zero for small non-zero time periods.  This may break code that relies
on receiving back a non-zero value.

jiffies_to_usecs() does not need such a fix, as <linux/jiffies.h> does
not support values of HZ larger than 12287, thus rejecting any
problematic huge values of HZ.

Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
---
v2:
  - Add examples of affected systems,
  - Use DIV_ROUND_UP().
---
 kernel/time/time.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/kernel/time/time.c b/kernel/time/time.c
index 6fa99213fc720e4b..2b41e8e2d31db26f 100644
--- a/kernel/time/time.c
+++ b/kernel/time/time.c
@@ -28,6 +28,7 @@
  */
 
 #include <linux/export.h>
+#include <linux/kernel.h>
 #include <linux/timex.h>
 #include <linux/capability.h>
 #include <linux/timekeeper_internal.h>
@@ -314,9 +315,10 @@ unsigned int jiffies_to_msecs(const unsigned long j)
 	return (j + (HZ / MSEC_PER_SEC) - 1)/(HZ / MSEC_PER_SEC);
 #else
 # if BITS_PER_LONG == 32
-	return (HZ_TO_MSEC_MUL32 * j) >> HZ_TO_MSEC_SHR32;
+	return (HZ_TO_MSEC_MUL32 * j + (1ULL << HZ_TO_MSEC_SHR32) - 1) >>
+	       HZ_TO_MSEC_SHR32;
 # else
-	return (j * HZ_TO_MSEC_NUM) / HZ_TO_MSEC_DEN;
+	return DIV_ROUND_UP(j * HZ_TO_MSEC_NUM, HZ_TO_MSEC_DEN);
 # endif
 #endif
 }
-- 
2.17.1

Re: [PATCH v2] time: Make sure jiffies_to_msecs() preserves non-zero time periods

[Geert Uytterhoeven]

CC alpha, mips

On Fri, Jun 22, 2018 at 9:54 AM Geert Uytterhoeven <geert@linux-m68k.org> wrote:
> For the common cases where 1000 is a multiple of HZ, or HZ is a multiple
> of 1000, jiffies_to_msecs() never returns zero when passed a non-zero
> time period.
>
> However, if HZ > 1000 and not an integer multiple of 1000 (e.g. 1024 or
> 1200, as used on alpha and DECstation), jiffies_to_msecs() may return
> zero for small non-zero time periods.  This may break code that relies
> on receiving back a non-zero value.
>
> jiffies_to_usecs() does not need such a fix, as <linux/jiffies.h> does
> not support values of HZ larger than 12287, thus rejecting any
> problematic huge values of HZ.
>
> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
> ---
> v2:
>   - Add examples of affected systems,
>   - Use DIV_ROUND_UP().
> ---
>  kernel/time/time.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/kernel/time/time.c b/kernel/time/time.c
> index 6fa99213fc720e4b..2b41e8e2d31db26f 100644
> --- a/kernel/time/time.c
> +++ b/kernel/time/time.c
> @@ -28,6 +28,7 @@
>   */
>
>  #include <linux/export.h>
> +#include <linux/kernel.h>
>  #include <linux/timex.h>
>  #include <linux/capability.h>
>  #include <linux/timekeeper_internal.h>
> @@ -314,9 +315,10 @@ unsigned int jiffies_to_msecs(const unsigned long j)
>         return (j + (HZ / MSEC_PER_SEC) - 1)/(HZ / MSEC_PER_SEC);
>  #else
>  # if BITS_PER_LONG == 32
> -       return (HZ_TO_MSEC_MUL32 * j) >> HZ_TO_MSEC_SHR32;
> +       return (HZ_TO_MSEC_MUL32 * j + (1ULL << HZ_TO_MSEC_SHR32) - 1) >>
> +              HZ_TO_MSEC_SHR32;
>  # else
> -       return (j * HZ_TO_MSEC_NUM) / HZ_TO_MSEC_DEN;
> +       return DIV_ROUND_UP(j * HZ_TO_MSEC_NUM, HZ_TO_MSEC_DEN);
>  # endif
>  #endif
>  }
> --
> 2.17.1

Re: [PATCH v2] time: Make sure jiffies_to_msecs() preserves non-zero time periods

[Arnd Bergmann]

On Fri, Jun 22, 2018 at 9:54 AM, Geert Uytterhoeven
<geert@linux-m68k.org> wrote:
> For the common cases where 1000 is a multiple of HZ, or HZ is a multiple
> of 1000, jiffies_to_msecs() never returns zero when passed a non-zero
> time period.
>
> However, if HZ > 1000 and not an integer multiple of 1000 (e.g. 1024 or
> 1200, as used on alpha and DECstation), jiffies_to_msecs() may return
> zero for small non-zero time periods.  This may break code that relies
> on receiving back a non-zero value.
>
> jiffies_to_usecs() does not need such a fix, as <linux/jiffies.h> does
> not support values of HZ larger than 12287, thus rejecting any
> problematic huge values of HZ.
>
> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
> ---

Looks good to me,

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

Re: [PATCH v2] time: Make sure jiffies_to_msecs() preserves non-zero time periods

[Thomas Gleixner]

On Fri, 22 Jun 2018, Geert Uytterhoeven wrote:
> For the common cases where 1000 is a multiple of HZ, or HZ is a multiple
> of 1000, jiffies_to_msecs() never returns zero when passed a non-zero
> time period.
> 
> However, if HZ > 1000 and not an integer multiple of 1000 (e.g. 1024 or
> 1200, as used on alpha and DECstation), jiffies_to_msecs() may return
> zero for small non-zero time periods.  This may break code that relies
> on receiving back a non-zero value.
> 
> jiffies_to_usecs() does not need such a fix, as <linux/jiffies.h> does
> not support values of HZ larger than 12287, thus rejecting any
> problematic huge values of HZ.

Sorry, I'm not understanding that sentence at all.

> Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>

This lacks a stable tag, right?

> ---
> v2:
>   - Add examples of affected systems,
>   - Use DIV_ROUND_UP().
> ---
>  kernel/time/time.c | 6 ++++--
>  1 file changed, 4 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/time/time.c b/kernel/time/time.c
> index 6fa99213fc720e4b..2b41e8e2d31db26f 100644
> --- a/kernel/time/time.c
> +++ b/kernel/time/time.c
> @@ -28,6 +28,7 @@
>   */
>  
>  #include <linux/export.h>
> +#include <linux/kernel.h>
>  #include <linux/timex.h>
>  #include <linux/capability.h>
>  #include <linux/timekeeper_internal.h>
> @@ -314,9 +315,10 @@ unsigned int jiffies_to_msecs(const unsigned long j)
>  	return (j + (HZ / MSEC_PER_SEC) - 1)/(HZ / MSEC_PER_SEC);
>  #else
>  # if BITS_PER_LONG == 32
> -	return (HZ_TO_MSEC_MUL32 * j) >> HZ_TO_MSEC_SHR32;
> +	return (HZ_TO_MSEC_MUL32 * j + (1ULL << HZ_TO_MSEC_SHR32) - 1) >>
> +	       HZ_TO_MSEC_SHR32;
>  # else
> -	return (j * HZ_TO_MSEC_NUM) / HZ_TO_MSEC_DEN;
> +	return DIV_ROUND_UP(j * HZ_TO_MSEC_NUM, HZ_TO_MSEC_DEN);
>  # endif
>  #endif
>  }
> -- 
> 2.17.1
> 
> 

Re: [PATCH v2] time: Make sure jiffies_to_msecs() preserves non-zero time periods

[Geert Uytterhoeven]

Hi Thomas,

On Fri, Jun 22, 2018 at 2:49 PM Thomas Gleixner <tglx@linutronix.de> wrote:
> On Fri, 22 Jun 2018, Geert Uytterhoeven wrote:
> > For the common cases where 1000 is a multiple of HZ, or HZ is a multiple
> > of 1000, jiffies_to_msecs() never returns zero when passed a non-zero
> > time period.
> >
> > However, if HZ > 1000 and not an integer multiple of 1000 (e.g. 1024 or
> > 1200, as used on alpha and DECstation), jiffies_to_msecs() may return
> > zero for small non-zero time periods.  This may break code that relies
> > on receiving back a non-zero value.
> >
> > jiffies_to_usecs() does not need such a fix, as <linux/jiffies.h> does
> > not support values of HZ larger than 12287, thus rejecting any
> > problematic huge values of HZ.
>
> Sorry, I'm not understanding that sentence at all.

Sorry for being unclear.

1 jiffy can only be less than 1µs if HZ > 1000000.
But include/linux/jiffies.h checks if HZ >= 12288, and does #error otherwise.
In addition, there's a "BUILD_BUG_ON(HZ > USEC_PER_SEC)" in time.c

> > Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
>
> This lacks a stable tag, right?

Up to the maintainer to add, isn't it?

Thanks!

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

Re: [PATCH v2] time: Make sure jiffies_to_msecs() preserves non-zero time periods

[Thomas Gleixner]

[-- Attachment #1: Type: text/plain, Size: 1418 bytes --]

On Fri, 22 Jun 2018, Geert Uytterhoeven wrote:
> On Fri, Jun 22, 2018 at 2:49 PM Thomas Gleixner <tglx@linutronix.de> wrote:
> > On Fri, 22 Jun 2018, Geert Uytterhoeven wrote:
> > > For the common cases where 1000 is a multiple of HZ, or HZ is a multiple
> > > of 1000, jiffies_to_msecs() never returns zero when passed a non-zero
> > > time period.
> > >
> > > However, if HZ > 1000 and not an integer multiple of 1000 (e.g. 1024 or
> > > 1200, as used on alpha and DECstation), jiffies_to_msecs() may return
> > > zero for small non-zero time periods.  This may break code that relies
> > > on receiving back a non-zero value.
> > >
> > > jiffies_to_usecs() does not need such a fix, as <linux/jiffies.h> does
> > > not support values of HZ larger than 12287, thus rejecting any
> > > problematic huge values of HZ.
> >
> > Sorry, I'm not understanding that sentence at all.
> 
> Sorry for being unclear.
> 
> 1 jiffy can only be less than 1µs if HZ > 1000000.
> But include/linux/jiffies.h checks if HZ >= 12288, and does #error otherwise.
> In addition, there's a "BUILD_BUG_ON(HZ > USEC_PER_SEC)" in time.c

Hmm, ok. Care to reword?

> > > Signed-off-by: Geert Uytterhoeven <geert@linux-m68k.org>
> >
> > This lacks a stable tag, right?
> 
> Up to the maintainer to add, isn't it?

Yes and no. At least a hint that this has been broken by commit X or has
been broken forever would be appreciated.

Thanks,

	tglx