On Sun, 19 Apr 2020 at 17:40:41 +0200, Mark Kettenis wrote:
It's crashing on line 166 of acpivout.c:
if (level < sc->sc_bcl[0])
(Note there was a diff in snaps that is now committed).
Apparently sc_bcl is NULL. That shouldn't happen but we have no
protection against it. Diff below should make sure things don't go
off the rails.
Bigger question is why the _BCL method isn't returning a proper list
of values. Judging from the included acpi dump it should...
I think these devices have some bogus AML that is causing _BCL not
to return a proper package.
Method (_BCL, 0, NotSerialized) // _BCL: Brightness Control Levels
{
If (CondRefOf (\PBCL))
{
PBCL ()
}
Else
{
Return (Package (0x67)
{
0x50,
[...]
I think that "PBCL()" needs to be "Return (PBCL)", otherwise it
doesn't return what PBCL() returns.
ACPICA seems to agree:
- eval _SB.PCI0.GFX0.DD1F._BCL
Evaluating \_SB.PCI0.GFX0.DD1F._BCL
ACPI Warning: \_SB.PCI0.GFX0.DD1F._BCL: Missing expected return value
(20200214/nsrepair-347)
ACPI Warning: \_SB.PCI0.GFX0.DD1F._BCL: Expected return object of type Package
(20200214/nspredef-416)
No object was returned from evaluation of \_SB.PCI0.GFX0.DD1F._BCL
The acpivout fix looks correct to me, most other methods in there
check for sc_bcl_len.
Index: dev/acpi/acpivout.c
===================================================================
RCS file: /cvs/src/sys/dev/acpi/acpivout.c,v
retrieving revision 1.22
diff -u -p -r1.22 acpivout.c
--- dev/acpi/acpivout.c 19 Apr 2020 15:05:14 -0000 1.22
+++ dev/acpi/acpivout.c 19 Apr 2020 15:38:31 -0000
@@ -114,12 +114,14 @@ acpivout_attach(struct device *parent, s
ws_get_param || ws_set_param)
return;
- ws_get_param = acpivout_get_param;
- ws_set_param = acpivout_set_param;
-
acpivout_get_bcl(sc);
+ if (sc->sc_bcl_len == 0)
+ return;
sc->sc_brightness = acpivout_get_brightness(sc);
+
+ ws_get_param = acpivout_get_param;
+ ws_set_param = acpivout_set_param;
}
int
