Skip to contentSkip to conversation

alpha 2.4.11-pre3: delay disabling early boot messages

PATCH Richard Henderson 3 participants 3 messages Oct 5, 2001 Alpha port

lore
Share & tools
lore
#0 lore raw
Quoted 1 line

From Jay Estabrook:

Alpha has this nice way to output console messages via SRM callbacks which we can enable immediately on bootup, and so get debugging output before console_init. This patch delays when we turn that off, and so narrows the window in which we can't get debug output.

r~

arch/alpha/kernel/irq_alpha.c +0 -5
diff -rup 2.4.10-dist/arch/alpha/kernel/irq_alpha.c 2.4.10/arch/alpha/kernel/irq_alpha.c
@@ -108,11 +108,6 @@ init_IRQ(void)
 	wrent(entInt, 0);
 
 	alpha_mv.init_irq();
-
-	/* If we had wanted SRM console printk echoing early, undo it now. */
-	if (alpha_using_srm && srmcons_output) {
-		unregister_srm_console();
-	}
 }
 
 /*
arch/alpha/kernel/setup.c +14 -6
diff -rup 2.4.10-dist/arch/alpha/kernel/setup.c 2.4.10/arch/alpha/kernel/setup.c
@@ -63,12 +63,20 @@ unsigned long srm_hae;
 /* Which processor we booted from.  */
 int boot_cpuid;
 
-/* Using SRM callbacks for initial console output. This works from
-   setup_arch() time through the end of init_IRQ(), as those places
-   are under our control.
-
-   By default, OFF; set it with a bootcommand arg of "srmcons".
-*/
+/*
+ * Using SRM callbacks for initial console output. This works from
+ * setup_arch() time through the end of time_init(), as those places
+ * are under our (Alpha) control.
+
+ * "srmcons" specified in the boot command arguments allows us to
+ * see kernel messages during the period of time before the true
+ * console device is "registered" during console_init(). As of this
+ * version (2.4.10), time_init() is the last Alpha-specific code
+ * called before console_init(), so we put "unregister" code
+ * there to prevent schizophrenic console behavior later... ;-}
+ *
+ * By default, OFF; set it with a bootcommand arg of "srmcons".
+ */
 int srmcons_output = 0;
 
 /* Enforce a memory size limit; useful for testing. By default, none. */
arch/alpha/kernel/time.c +15 -0
diff -rup 2.4.10-dist/arch/alpha/kernel/time.c 2.4.10/arch/alpha/kernel/time.c
@@ -332,6 +333,21 @@ time_init(void)
 	alpha_mv.init_rtc();
 
 	do_get_fast_time = do_gettimeofday;
+
+	/*
+	 * If we had wanted SRM console printk echoing early, undo it now.
+	 *
+	 * "srmcons" specified in the boot command arguments allows us to
+	 * see kernel messages during the period of time before the true
+	 * console device is "registered" during console_init(). As of this
+	 * version (2.4.10), time_init() is the last Alpha-specific code
+	 * called before console_init(), so we put this "unregister" code
+	 * here to prevent schizophrenic console behavior later... ;-}
+	 */
+	if (alpha_using_srm && srmcons_output) {
+		unregister_srm_console();
+		srmcons_output = 0;
+	}
 }
 
 /*
#1 lore raw
Quoted 4 lines · [email protected] said:

Alpha has this nice way to output console messages via SRM callbacks which we can enable immediately on bootup, and so get debugging output before console_init. This patch delays when we turn that off, and so narrows the window in which we can't get debug output.

Putting it in time_init() because we happen to know that time_init() is the last thing called from init/main.c before console_init() is a bit of a hack :)

I've wanted this on other platforms too - couldn't we add an unregister_boot_console() to console_init() instead?

--
dwmw2

#2 lore raw
In reply to Richard Henderson
Quoted 2 lines · David Woodhouse

I've wanted this on other platforms too - couldn't we add an unregister_boot_console() to console_init() instead?

Seems reasonable. I wouldn't want to touch such a think in 2.4 though.

r~