[regression]: acpi_aiboost fails to attach on Asus P5Q3 Deluxe in 8.0 (Was: [regression] : 8.0-BETA3 (and BETA2?) acpi_hpet0 fails on TYAN H2000M)

[John Baldwin]

On Saturday 19 September 2009 1:09:19 pm Olivier Smedts wrote:
> 2009/8/24 John Baldwin <jhb at freebsd.org>
> >
> > On Monday 24 August 2009 4:54:13 pm Olivier Smedts wrote:
> > > 2009/8/24 John Baldwin <jhb at freebsd.org>:
> > > > On Saturday 22 August 2009 1:20:18 pm Arno J. Klaassen wrote:
> > > >>
> > > >> Hello,
> > > >>
> > > >> I have a regression with acpi_hpet on a Tyan H2000M MB :
> > > >>
> > > >>   acpi_hpet0: <High Precision Event Timer> iomem 0xfed00000-0xfed03fff on
> > acpi0
> > > >>   acpi_hpet0: HPET never increments, disabling
> > > >>   device_attach: acpi_hpet0 attach returned 6
> > > >>   [twice]
> > > >>
> > > >> it exists at least since Aug16 sources (I just looked at
> > > >> the 'netif' problems on this board which BTW are indeed fixed
> > > >> by recent flowtable init changes)
> > > >
> > > > Perhaps the recent ACPI-CA update is when this broke?
> > >
> > > Sorry for hijacking the thread (and not reporting this before) but
> > > maybe my issues with acpi_aiboost are related. It won't attach on my
> > > ASUS P5Q3 Deluxe (latest BIOS) since the ACPICA update (which also
> > > made "est" attach successfully on this ASUS board) :
> > >
> > > acpi_aiboost0: <ASUStek AIBOOSTER> on acpi0
> > > NOREF
> > > BAD DATA
> > > device_attach: acpi_aiboost0 attach returned 22
> >
> > I would figure out what function in it's attach routine is failing with EINVAL
> > and drill down from there.  It looks like it is dying trying to
> > evaluate "TSIF", "VSIF", or "FSIF".  Maybe add some printfs to see which one
> > and print out the 'elem->Type' in the "NOREF" error message.  You could also
> > just print the 'name' there as well which would help perhaps.
> >
> > --
> > John Baldwin

Ok, try this.  I believe that the newer ACPI-CA is expanding the package
inline and there is no reason the subpackages have to be external
references rather than inline.

Index: acpi_aiboost.c
===================================================================
--- acpi_aiboost.c	(revision 197402)
+++ acpi_aiboost.c	(working copy)
@@ -46,7 +46,6 @@
 
 #define DESCSTRLEN 32
 struct acpi_aiboost_element{
-	ACPI_HANDLE h;
 	uint32_t id;
 	char desc[DESCSTRLEN];
 };
@@ -127,22 +126,23 @@
 	
 	for(i = 1 ; i < o->Package.Count; i++){
 		elem = &o->Package.Elements[i];
-		if(elem->Type != ACPI_TYPE_ANY){
-			printf("NOREF\n");
-			goto error;
-		}
-		c->elem[ i - 1].h = elem->Reference.Handle;
+		if (elem->type == ACPI_TYPE_ANY) {
+			buf2.Pointer = NULL;
+			buf2.Length = ACPI_ALLOCATE_BUFFER;
 
-		buf2.Pointer = NULL;
-		buf2.Length = ACPI_ALLOCATE_BUFFER;
-		
-		status = AcpiEvaluateObject(c->elem[i - 1].h, NULL, NULL,
-					    &buf2);
-		if(ACPI_FAILURE(status)){
-			printf("FETCH OBJECT\n");
+			status = AcpiEvaluateObject(elem->Reference.Handle,
+			    NULL, NULL, &buf2);
+			if (ACPI_FAILURE(status)){
+				printf("FETCH OBJECT\n");
+				goto error;
+			}
+			subobj = buf2.Pointer;
+		} else if (elem->type == ACPI_TYPE_PACKAGE)
+			subobj = elem;
+		else {
+			printf("NO PACKAGE\n");
 			goto error;
 		}
-		subobj = buf2.Pointer;
 		if(ACPI_FAILURE(acpi_PkgInt32(subobj,0, &c->elem[i -1].id))){
 			printf("ID FAILED\n");
 			goto error;
@@ -151,15 +151,17 @@
 				     sizeof(c->elem[i - 1].desc));
 		if(ACPI_FAILURE(status)){
 			if(status == E2BIG){
-				c->elem[i-1].desc[DESCSTRLEN-1] = 0;
+				c->elem[i - 1].desc[DESCSTRLEN-1] = 0;
 			}else{
 				printf("DESC FAILED %d\n", i-1);
 				goto error;
 			}
 		}
 		
-		if(buf2.Pointer)
-		  AcpiOsFree(buf2.Pointer);
+		if (buf2.Pointer) {
+			AcpiOsFree(buf2.Pointer);
+			buf2.Pointer = NULL;
+		}
 	}
 
 	if(buf.Pointer)

-- 
John Baldwin