[dpdk-dev] [PATCH] nohz: disable traceback on NO_HZ warning

[dpdk-dev] [PATCH] nohz: disable traceback on NO_HZ warning

[Stephen Hemminger]

From: "Charles (Chas) Williams" <ciwillia@brocade.com>

A NO_HZ_FULL kernel needs to have a stable clock source like a non-stop
(or invariant) TSC.  Unfortunately, this CPU feature/flag isn't advertised
by most hypervisors because they want the ability to migrate or save
virtual machines which would affect the TSC.

This means that a kernel with NO_HZ_FULL configured would often generate
a kernel traceback on boot which causes users to generate false
alarms. NO_HZ_FULL is an optimization not a hard requirement.
Keep the message, just lose the traceback.

Signed-off-by: Charles (Chas) Williams <ciwillia@brocade.com>
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
---
 kernel/time/tick-sched.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
index 7c7ec45..2e1c90c 100644
--- a/kernel/time/tick-sched.c
+++ b/kernel/time/tick-sched.c
@@ -188,8 +188,8 @@ static bool can_stop_full_tick(void)
 		 * Don't allow the user to think they can get
 		 * full NO_HZ with this machine.
 		 */
-		WARN_ONCE(tick_nohz_full_running,
-			  "NO_HZ FULL will not work with unstable sched clock");
+		if (tick_nohz_full_running)
+			pr_notice_once("NO_HZ FULL will not work with unstable sched clock\n");
 		return false;
 	}
 #endif
-- 
2.1.4

Re: [dpdk-dev] [PATCH] nohz: disable traceback on NO_HZ warning

[Stephen Hemminger]

On Mon, 14 Dec 2015 10:26:07 -0800
Stephen Hemminger <stephen@networkplumber.org> wrote:

> From: "Charles (Chas) Williams" <ciwillia@brocade.com>
> 
> A NO_HZ_FULL kernel needs to have a stable clock source like a non-stop
> (or invariant) TSC.  Unfortunately, this CPU feature/flag isn't advertised
> by most hypervisors because they want the ability to migrate or save
> virtual machines which would affect the TSC.
> 
> This means that a kernel with NO_HZ_FULL configured would often generate
> a kernel traceback on boot which causes users to generate false
> alarms. NO_HZ_FULL is an optimization not a hard requirement.
> Keep the message, just lose the traceback.
> 
> Signed-off-by: Charles (Chas) Williams <ciwillia@brocade.com>
> Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
> ---
>  kernel/time/tick-sched.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/kernel/time/tick-sched.c b/kernel/time/tick-sched.c
> index 7c7ec45..2e1c90c 100644
> --- a/kernel/time/tick-sched.c
> +++ b/kernel/time/tick-sched.c
> @@ -188,8 +188,8 @@ static bool can_stop_full_tick(void)
>  		 * Don't allow the user to think they can get
>  		 * full NO_HZ with this machine.
>  		 */
> -		WARN_ONCE(tick_nohz_full_running,
> -			  "NO_HZ FULL will not work with unstable sched clock");
> +		if (tick_nohz_full_running)
> +			pr_notice_once("NO_HZ FULL will not work with unstable sched clock\n");
>  		return false;
>  	}
>  #endif

Never mind, wrong list