Re: [PATCH] docs: panic: Disclaimer about console verbosity when using panic_print with pstore

[Petr Mladek]

On Wed 2026-08-05 17:49:04, Guilherme G. Piccoli wrote:
> Users of panic_print / panic_sys_info are usually aware of the potentially
> huge amount of information displayed with such options - their goal by
> using them is indeed to dump more information during panic!
> 
> What might come unnoticed at first is the impact that the console_verbose()
> call on panic path has, specially in cases that users rely on pstore as
> a means to collect the panic logs. Recent experience shows that dumping
> all tasks in an ARM64 system (with qcom_geni tty driver) gets a 2 minute
> delay on panic time due to writing such information to console. But in
> that case, pstore was used, so there was no need for console printing,
> and suppressing that reduced the full time to less than 1 second.
> 
> The option "printk.console_no_auto_verbose" aims exactly at this use
> case: avoid automatically increasing the loglevel during panic path.
> So, let's officially document this to avoid users wasting their time
> either on long panic printing or debugging the panic delays.

Make sense. Few nits below.

> Signed-off-by: Guilherme G. Piccoli <gpiccoli@xxxxxxxxxx>
> ---
>  Documentation/admin-guide/kernel-parameters.txt | 5 +++++
>  1 file changed, 5 insertions(+)
> 
> diff --git a/Documentation/admin-guide/kernel-parameters.txt b/Documentation/admin-guide/kernel-parameters.txt
> index b5493a7f8f22..04c0ef2cf910 100644
> --- a/Documentation/admin-guide/kernel-parameters.txt
> +++ b/Documentation/admin-guide/kernel-parameters.txt
> @@ -4862,6 +4862,9 @@ Kernel parameters
>  			so there are risks of losing older messages in the log.
>  			Use this option carefully, maybe worth to setup a
>  			bigger log buffer with "log_buf_len" along with this.
> +			Also consider using "printk.console_no_auto_verbose=Y"
> +			if using this along with pstore, to avoid extra delays
> +			due to increased console verbosity during panic.

The long blobs of text are hard to read. I would use this opportunity
to add some blank lines, for example:

<proposal>
	panic_print=	Bitmask for printing system info when panic happens.
			User can chose combination of the following bits:

			bit 0: print all tasks info
			bit 1: print system memory info
			bit 2: print timer info
			bit 3: print locks info if CONFIG_LOCKDEP is on
			bit 4: print ftrace buffer
			bit 5: replay all kernel messages on consoles at the end of panic
			bit 6: print all CPUs backtrace (if available in the arch)
			bit 7: print only tasks in uninterruptible (blocked) state

			*Be aware* that this option may print a _lot_ of lines.

			There is an increased risk of losing older messages in
			the log. Maybe worth to setup a bigger log buffer with
			"log_buf_len" along with this.

			Also consider using "printk.console_no_auto_verbose=Y"
			if using this along with pstore, to avoid extra delays
			due to increased console verbosity during panic.
</proposal>

>  
>  	panic_sys_info= A comma separated list of extra information to be dumped
>                          on panic.
> @@ -4877,6 +4880,8 @@ Kernel parameters
>                          blocked_tasks:  print only tasks in uninterruptible (blocked) state
>  
>                          This is a human readable alternative to the 'panic_print' option.
> +                        Please be aware of the caveats using this option, check
> +                        the panic_print information above regarding delays.

This is too focused on the "delays" problem. I would write it a
generic way:

<proposal>
			Please be aware of the caveats using this option, see
			the 'panic_print' option for more details.
</proposal>

>  	panic_console_replay
>  			When panic happens, replay all kernel messages on

Best Regards,
Petr