2013-09-20 15:34:16

by Kees Cook

Subject: [PATCH RESEND] printk: report boot console names during cut-over

This reports the names of boot consoles as they're being disabled to
help identify which is which during cut-over. Helps answer the question
"which boot console actually got activated?" once the regular console
is running, mostly when debugging boot console failures.

Signed-off-by: Kees Cook <[email protected]>
---
kernel/printk/printk.c | 16 ++++++++++------
1 file changed, 10 insertions(+), 6 deletions(-)

diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
index b4e8500..0dc54c9 100644
--- a/kernel/printk/printk.c
+++ b/kernel/printk/printk.c
@@ -2361,14 +2361,18 @@ void register_console(struct console *newcon)
if (bcon &&
((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
!keep_bootcon) {
- /* we need to iterate through twice, to make sure we print
- * everything out, before we unregister the console(s)
+ /* We need to iterate through all boot consoles, to make
+ * sure we print everything out, before we unregister them.
*/
- printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n",
- newcon->name, newcon->index);
- for_each_console(bcon)
- if (bcon->flags & CON_BOOT)
+ for_each_console(bcon) {
+ if (bcon->flags & CON_BOOT) {
+ printk(KERN_INFO "console [%s%d] enabled, "
+ "bootconsole [%s%d] disabled\n",
+ newcon->name, newcon->index,
+ bcon->name, bcon->index);
unregister_console(bcon);
+ }
+ }
} else {
printk(KERN_INFO "%sconsole [%s%d] enabled\n",
(newcon->flags & CON_BOOT) ? "boot" : "" ,
--
1.7.9.5


--
Kees Cook
Chrome OS Security

2013-09-21 23:52:32

by Frederic Weisbecker

Subject: Re: [PATCH RESEND] printk: report boot console names during cut-over

On Fri, Sep 20, 2013 at 08:34:01AM -0700, Kees Cook wrote:
> This reports the names of boot consoles as they're being disabled to
> help identify which is which during cut-over. Helps answer the question
> "which boot console actually got activated?" once the regular console
> is running, mostly when debugging boot console failures.
>
> Signed-off-by: Kees Cook <[email protected]>
> ---
> kernel/printk/printk.c | 16 ++++++++++------
> 1 file changed, 10 insertions(+), 6 deletions(-)
>
> diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
> index b4e8500..0dc54c9 100644
> --- a/kernel/printk/printk.c
> +++ b/kernel/printk/printk.c
> @@ -2361,14 +2361,18 @@ void register_console(struct console *newcon)
> if (bcon &&
> ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
> !keep_bootcon) {
> - /* we need to iterate through twice, to make sure we print
> - * everything out, before we unregister the console(s)
> + /* We need to iterate through all boot consoles, to make
> + * sure we print everything out, before we unregister them.
> */
> - printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n",
> - newcon->name, newcon->index);
> - for_each_console(bcon)
> - if (bcon->flags & CON_BOOT)
> + for_each_console(bcon) {
> + if (bcon->flags & CON_BOOT) {
> + printk(KERN_INFO "console [%s%d] enabled, "
> + "bootconsole [%s%d] disabled\n",
> + newcon->name, newcon->index,
> + bcon->name, bcon->index);
> unregister_console(bcon);

Looks useful! But I think the "console %foo disabled" message should move to unregister_console()
instead. That way it also handles printk_late_init() and all other console unregisters.

Thanks.

2013-09-23 21:25:11

by Kees Cook

Subject: Re: [PATCH RESEND] printk: report boot console names during cut-over

On Sat, Sep 21, 2013 at 4:52 PM, Frederic Weisbecker <[email protected]> wrote:
> On Fri, Sep 20, 2013 at 08:34:01AM -0700, Kees Cook wrote:
>> This reports the names of boot consoles as they're being disabled to
>> help identify which is which during cut-over. Helps answer the question
>> "which boot console actually got activated?" once the regular console
>> is running, mostly when debugging boot console failures.
>>
>> Signed-off-by: Kees Cook <[email protected]>
>> ---
>> kernel/printk/printk.c | 16 ++++++++++------
>> 1 file changed, 10 insertions(+), 6 deletions(-)
>>
>> diff --git a/kernel/printk/printk.c b/kernel/printk/printk.c
>> index b4e8500..0dc54c9 100644
>> --- a/kernel/printk/printk.c
>> +++ b/kernel/printk/printk.c
>> @@ -2361,14 +2361,18 @@ void register_console(struct console *newcon)
>> if (bcon &&
>> ((newcon->flags & (CON_CONSDEV | CON_BOOT)) == CON_CONSDEV) &&
>> !keep_bootcon) {
>> - /* we need to iterate through twice, to make sure we print
>> - * everything out, before we unregister the console(s)
>> + /* We need to iterate through all boot consoles, to make
>> + * sure we print everything out, before we unregister them.
>> */
>> - printk(KERN_INFO "console [%s%d] enabled, bootconsole disabled\n",
>> - newcon->name, newcon->index);
>> - for_each_console(bcon)
>> - if (bcon->flags & CON_BOOT)
>> + for_each_console(bcon) {
>> + if (bcon->flags & CON_BOOT) {
>> + printk(KERN_INFO "console [%s%d] enabled, "
>> + "bootconsole [%s%d] disabled\n",
>> + newcon->name, newcon->index,
>> + bcon->name, bcon->index);
>> unregister_console(bcon);
>
> Looks useful! But I think the "console %foo disabled" message should move to unregister_console()
> instead. That way it also handles printk_late_init() and all other console unregisters.

Yeah, that's a much cleaner way to do this. I'll send a new patch. Thanks!

-Kees

--
Kees Cook
Chrome OS Security