Conflict with FORCE_CON: Re: [PATCH v6 04/11] printk: Support toggling per-console loglevel via syslog() and cmdline

[Petr Mladek]

On Mon 2024-10-28 16:45:40, Chris Down wrote:
> A new module parameter (ignore_per_console_loglevel) is added, which can
> be set via the kernel command line or at runtime through
> /sys/module/printk/parameters/ignore_per_console_loglevel. When set, the
> per-console loglevels are ignored, and the global console loglevel
> (console_loglevel) is used for all consoles.
> 
> During sysrq, we temporarily disable per-console loglevels to ensure all
> requisite messages are printed to the console. This is necessary because
> sysrq is often used in dire circumstances where access to
> /sys/class/console may not be trivially possible.

I have just pushed a patchset which removed the console_loglevel
manipulation from sysrq, see
https://lore.kernel.org/r/20241105-printk-loud-con-v2-0-bd3ecdf7b0e4@xxxxxxxx

As a result, the change in drivers/tty/sysrq.c is not needed anymore.

Note that the other patchset causes some conflict with this patchset.
But they does not seem to be hard to resolved:


1st conflict is in boot_delay_msec(). But the affected logic
has actually been moved to printk_delay(). As a result,
boot_delay_msec() might stay as it is:

static void boot_delay_msec(void)
{
	unsigned long long k;
	unsigned long timeout;

	if (boot_delay == 0 || system_state >= SYSTEM_RUNNING)
		return;

	k = (unsigned long long)loops_per_msec * boot_delay;

	timeout = jiffies + msecs_to_jiffies(boot_delay);
	while (k) {
		k--;
		cpu_relax();
		/*
		 * use (volatile) jiffies to prevent
		 * compiler reduction; loop termination via jiffies
		 * is secondary and may or may not happen.
		 */
		if (time_after(jiffies, timeout))
			break;
		touch_nmi_watchdog();
	}
}


Instead, we should add check of is_printk_force_console() into
printk_delay(). I suggest to do it the following way:

static bool suppress_message_printing_everywhere(int level)
{
	bool suppress_everywhere = true;
	struct console *con;
	int cookie;

	cookie = console_srcu_read_lock();

	for_each_console_srcu(con) {
		if (!suppress_message_printing(level, con)) {
			suppress_everywhere = false;
			break;
		}
	}
	console_srcu_read_unlock(cookie);

	return suppress_everywhere;
}

static inline void printk_delay(int level)
{
	if (!is_printk_force_console() &&
	    suppress_message_printing_everywhere(level))
		return;

	boot_delay_msec();

	if (unlikely(printk_delay_msec)) {
		int m = printk_delay_msec;

		while (m--) {
			mdelay(1);
			touch_nmi_watchdog();
		}
	}
}


2nd conflict is in printk_get_next_message(). I suggest to do
something like:

	force_con = r.info->flags & LOG_FORCE_CON;

	/*
	 * Skip records that are not forced to be printed on consoles and that
	 * has level above the console loglevel. Never suppress when used in
	 * devkmsg_read().
	 */
	if (!force_con && con && suppress_message_printing(r.info->level, con))
		goto out;


Actually, I suggested to pass @con_level instead of @con here.
In which case, we might need something like:

	if (con) {
		is_extended = console_srcu_read_flags(con) & CON_EXTENDED;
		con_level = console_srcu_read_loglevel(con);
	} else {
		/* Used only by devkmsg_read(). Show all messages there. */
		is_extended = true;
		con_level = CONSOLE_LOGLEVEL_MOTORMOUTH;
	}
[...]
	force_con = r.info->flags & LOG_FORCE_CON;

	/*
	 * Skip records that are not forced to be printed on consoles and that
	 * has level above the console loglevel.
	 */
	if (!force_con && suppress_message_printing(r.info->level, con_level))
		goto out;



I hope that you find there code snippets useful. I provide them
because I feel a bit guilty that I have already merged the other
patchset... ;-)

Best Regards,
Petr