[PATCH][2.5] fix 'pci=noacpi' was: Re: [patch] remove mount_root_failed_msg()

[Thomas Schlichter <schlicht@uni-mannheim.de>]

[-- Attachment #1: Type: text/plain, Size: 1385 bytes --]

On Thursday 21 August 2003 01:41, Greg KH wrote:
> On Wed, Aug 20, 2003 at 07:20:45PM -0400, Jeff Garzik wrote:
> > * overall, I disagree with adding messages like this.  The number one
> >   bug report, by far, for networking drivers is ACPI-related (no
> >   interrupts delivered).  You don't see me adding "boot with acpi=off"
> >   messages to the net subsystem.
>
> It's the number one bug report for USB drivers too :(
>
> It has gotten smaller, and I would hope that is due to advances in the
> ACPI code, but they still do crop up quite frequently.

Well, I've got a board with these problems, too... (Epox 8K9A)
It seems the ACPI IRQ override tables are buggy, so it cannot work... :-(
But with 'pci=noacpi' it should work as the Documentation states:

       noacpi                  [IA-32] Do not use ACPI for IRQ routing.

Unfortunately the ACPI IO-APIC setup is not overridden and without the 
attached patch, which corrects this issue, I had to use 'acpi=off'...

I think this will help people easily working around their APCI interrupt 
routing problems, as I think most of these are due to buggy ACPI IRQ override 
tables. Windows seems not to use them and so they are often badly 
tested... :-(

Best regards
   Thomas Schlichter

BTW: Hunk 3 of this patch is only a simplification, because(acpi_lapic && 
acpi_ioapic) is always true here when that code is reached...

[-- Attachment #2: fix_noacpi.diff --]
[-- Type: text/x-diff, Size: 1504 bytes --]

--- linux-2.6.0-test3-mm3/arch/i386/kernel/acpi/boot.c.orig	Wed Aug 20 03:42:13 2003
+++ linux-2.6.0-test3-mm3/arch/i386/kernel/acpi/boot.c	Wed Aug 20 04:03:56 2003
@@ -39,6 +39,7 @@
 #define PREFIX			"ACPI: "
 
 extern int acpi_disabled;
+extern int acpi_irq;
 extern int acpi_ht;
 
 /* --------------------------------------------------------------------------
@@ -416,7 +417,7 @@
 	 * If MPS is present, it will handle them,
 	 * otherwise the system will stay in PIC mode
 	 */
-	if (acpi_disabled) {
+	if (acpi_disabled || !acpi_irq) {
 		return 1;
         }
 
@@ -451,15 +452,13 @@
 
 	acpi_ioapic = 1;
 
+#ifdef CONFIG_X86_LOCAL_APIC
+	smp_found_config = 1;
+	clustered_apic_check();
+#endif
+
 #endif /*CONFIG_ACPI*/
 #endif /*CONFIG_X86_IO_APIC*/
-
-#ifdef CONFIG_X86_LOCAL_APIC
-	if (acpi_lapic && acpi_ioapic) {
-		smp_found_config = 1;
-		clustered_apic_check();
-	}
-#endif
 
 	return 0;
 }
--- linux-2.6.0-test3-mm3/arch/i386/kernel/setup.c.orig	Wed Aug 20 03:41:56 2003
+++ linux-2.6.0-test3-mm3/arch/i386/kernel/setup.c	Wed Aug 20 04:03:03 2003
@@ -70,6 +70,7 @@
 EXPORT_SYMBOL(acpi_disabled);
 
 #ifdef	CONFIG_ACPI_BOOT
+	int acpi_irq __initdata = 1;	/* enable IRQ */
 	int acpi_ht __initdata = 1;	/* enable HT */
 #endif
 
@@ -541,6 +542,11 @@
 		else if (!memcmp(from, "acpi=ht", 7)) {
 			acpi_ht = 1;
 			if (!acpi_force) acpi_disabled = 1;
+		}
+
+		/* "pci=noacpi" disables ACPI interrupt routing */
+		else if (!memcmp(from, "pci=noacpi", 10)) {
+			acpi_irq = 0;
 		}
 #endif