On Thu, 2025-06-12 at 13:46 +0200, Petr Mladek wrote:
> On Fri 2025-06-06 23:53:43, Marcos Paulo de Souza wrote:
> > Since commit 9e70a5e109a4 ("printk: Add per-console suspended
> > state") the
> > CON_SUSPENDED flag was introced, and this flag was being checked on
> > console_is_usable function, which returns false if the console is
> > suspended.
> >
> > No functional changes.
>
> I double checked potential functional changes. In particular, I
> checked where the CON_ENABLED and CON_SUSPENDED flags were used.
>
> Both flags seems to have the same effect in most situations,
> for example, in console_is_usable() or console_unblank().
>
> But there seems to be two exceptions: kdb_msg_write() and
> show_cons_active(). These two functions check only
> the CON_ENABLED flag. And they think that the console is
> usable when the flag is set.
>
> The change in this patch would change the behavior of the two
> functions during suspend. It is later fixed by the 3rd and 4th
> patch. But it might cause regressions during bisections.
>
> It is probably not a big deal because the system is not much
> usable during the suspend anyway. But still, I would feel more
> comfortable if we prevented the "temporary" regression.
>
> I see two possibilities:
>
> 1. Merge the 3rd and 4th patch into this one. It would change
> the semantic in a single patch.
I liked the suggestion, I'll do it in the next revision. Thanks for
looking into it!
>
> 2. First update kdb_msg_write() and show_cons_active()
> to check both CON_ENABLE and CON_SUSPENDED flags.
>
> The 1st solution probably makes more sense because we are going
> to remove the CON_ENABLE flag in the end. And even the merged
> patch is small enough.
>
> Best Regards,
> Petr
