[PATCH] printk: fix wrong number of CPUs required to increase the ring buffer

[PATCH] printk: fix wrong number of CPUs required to increase the ring buffer

[Sang-Heon Jeon]

The default LOG_BUF_SHIFT is 17, which means 128 KB, and
LOG_CPU_MAX_BUF_SHIFT is 12, which means 4 KB per CPU.

By default the ring buffer is increased when
(num_possible_cpus() - 1) * 4 KB > 64 KB, which requires 18 or more CPUs.

Fix the wrong comment and help text.

Signed-off-by: Sang-Heon Jeon <ekffu200098@gmail.com>
---
 init/Kconfig           | 2 +-
 kernel/printk/printk.c | 2 +-
 2 files changed, 2 insertions(+), 2 deletions(-)

diff --git a/init/Kconfig b/init/Kconfig
index 60448fd4ee12..446cbb59a43b 100644
--- a/init/Kconfig
+++ b/init/Kconfig
@@ -843,7 +843,7 @@ config LOG_CPU_MAX_BUF_SHIFT
 	  with more CPUs. Therefore this value is used only when the sum of
 	  contributions is greater than the half of the default kernel ring
 	  buffer as defined by LOG_BUF_SHIFT. The default values are set
-	  so that more than 16 CPUs are needed to trigger the allocation.
+	  so that 18 or more CPUs are needed to trigger the allocation.
 
 	  Also this option is ignored when "log_buf_len" kernel parameter is
 	  used as it forces an exact (power of two) size of the ring buffer.
diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index 00f8e1291564..c7d961256c88 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -1087,7 +1087,7 @@ static void __init log_buf_add_cpu(void)
 
 	cpu_extra = (num_possible_cpus() - 1) * __LOG_CPU_MAX_BUF_LEN;
 
-	/* by default this will only continue through for large > 64 CPUs */
+	/* by default this will only continue through for 18 or more CPUs */
 	if (cpu_extra <= __LOG_BUF_LEN / 2)
 		return;
 

base-commit: b03d22ec8186f3f6b40b95a5a678100cf80116cb
-- 
2.43.0

Re: [PATCH] printk: fix wrong number of CPUs required to increase the ring buffer

[Petr Mladek]

On Tue 2026-09-22 01:38:27, Sang-Heon Jeon wrote:
> The default LOG_BUF_SHIFT is 17, which means 128 KB, and
> LOG_CPU_MAX_BUF_SHIFT is 12, which means 4 KB per CPU.
> 
> By default the ring buffer is increased when
> (num_possible_cpus() - 1) * 4 KB > 64 KB, which requires 18 or more CPUs.

Good catch!

> Fix the wrong comment and help text.
> 
> Signed-off-by: Sang-Heon Jeon <ekffu200098@gmail.com>
> ---
>  init/Kconfig           | 2 +-
>  kernel/printk/printk.c | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/init/Kconfig b/init/Kconfig
> index 60448fd4ee12..446cbb59a43b 100644
> --- a/init/Kconfig
> +++ b/init/Kconfig
> @@ -843,7 +843,7 @@ config LOG_CPU_MAX_BUF_SHIFT
>  	  with more CPUs. Therefore this value is used only when the sum of
>  	  contributions is greater than the half of the default kernel ring
>  	  buffer as defined by LOG_BUF_SHIFT. The default values are set
> -	  so that more than 16 CPUs are needed to trigger the allocation.
> +	  so that 18 or more CPUs are needed to trigger the allocation.

The number 18 is pretty ugly. It was supposed to be a power of 2.
I would prefer to fix the computation instead.
>  
>  	  Also this option is ignored when "log_buf_len" kernel parameter is
>  	  used as it forces an exact (power of two) size of the ring buffer.
> diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
> index 00f8e1291564..c7d961256c88 100644
> --- a/kernel/printk/printk.c
> +++ b/kernel/printk/printk.c
> @@ -1087,7 +1087,7 @@ static void __init log_buf_add_cpu(void)
>  
>  	cpu_extra = (num_possible_cpus() - 1) * __LOG_CPU_MAX_BUF_LEN;

If we remove the minus one:

	cpu_extra = num_possible_cpus() * __LOG_CPU_MAX_BUF_LEN;

The it should be > 16 and we could keep init/Kconfig as is.

> -	/* by default this will only continue through for large > 64 CPUs */
> +	/* by default this will only continue through for 18 or more CPUs */

Right, this should be > 16 with the current defaults.

>  	if (cpu_extra <= __LOG_BUF_LEN / 2)
>  		return;

Would you like to prepare v2?

Best Regards,
Petr