On Mon, 2024-10-21 at 16:17 +0206, John Ogness wrote:
> On 2024-10-21, Petr Mladek <pmladek@xxxxxxxx> wrote:
> > > That will not work because migrate_enable() can only be called
> > > from
> > > can_sleep context. Instead, the migrate_disable()/enable() should
> > > be at
> > > the few (one?) call sites where
> > > printk_loud_console_enter()/exit() is
> > > used from task context.
> >
> > Hmm, if I get it correctly, we could not use migrate_disable() in
> > __handle_sysrq() because it can be called also in atomic context,
>
> I am talking about callers of __handle_sysrq() and/or their callers.
>
> For example write_sysrq_trigger() could do:
>
> migrate_disable();
> __handle_sysrq(c, false);
> migrate_enable();
>
> Or a new wrapper could be introduced for this purpose:
>
> static inline void wrapper handle_sysrq_task(u8 key, bool check_mask)
> {
> migrate_disable();
> __handle_sysrq(key, check_mask);
> migrate_enable();
> }
>
> A quick grep shows about 25 call sites to check.
>
> > I do not see any easy way how to distinguish whether it was called
> > in
> > an atomic context or not.
>
> There is no clean way to do that. If this information is needed, it
> must
> be tracked by the call chain.
>
> > So, I see three possibilities:
> >
> > 1. Explicitly call preempt_disable() in __handle_sysrq().
> >
> > It would be just around the the single line or the help. But
> > still,
> > I do not like it much.
>
> Not acceptable for PREEMPT_RT since sysrq is exposed to external
> inputs.
>
> > 2. Avoid the per-CPU variable. Force adding the
> > LOUD_CON/FORCE_CON
> > flag using a global variable, e.g. printk_force_console.
> >
> > The problem is that it might affect also messages printed by
> > other CPUs. And there might be many.
> >
> > Well, console_loglevel is a global variable. The original code
> > had a similar problem.
>
> If disabling migration is not an option for you, this would be my
> second
> choice. I assume tagging too many messages is better than not tagging
> enough. And, as you say, this is effectively what the current code is
> trying to do.
Thanks for your input John. I talked with Petr and he suggested to
follow this option. I'll prepare the changes and send them after
reviewing them with Petr.
Thanks,
Marcos
>
> > 3. Add the LOUD_CON/FLUSH_CON flag via a parameter. For example,
> > by a special LOGLEVEL_FORCE_CON, similar to LOGLEVEL_SCHED.
> >
> > I might work well for __handle_sysrq() which calls the
> > affected
> > printk() directly.
> >
> > But it won't work, for example, for kdb_show_stack(). It wants
> > to show messages printed by a nested functions.
>
> Right, this has limited usefulness and might miss the important
> things,
> which tend to be within helper functions.
>
> John
