Re: [Qemu-devel] os x boot broken by commit 11948748495841bd54721b250d68c7b3cb0475ef

Re: [Qemu-devel] os x boot broken by commit 11948748495841bd54721b250d68c7b3cb0475ef

[Gabriel L. Somlo]

Added qemu-devel, since that is where this stuff belongs now. Everyone
else, sorry for the dupe...

On Thu, Nov 21, 2013 at 07:14:27PM +0100, Paolo Bonzini wrote:
> Can you remind us about your DSDT modifications?  It should be possible
> to patch the HPET and applesmc bits appropriately from QEMU (or to move
> them from the DSDT to an SSDT that is built entirely in QEMU).
> 
> It actually isn't impossible that Mac OS X would boot just fine with 1.8...

My current DSDT patch (against QEMU) is enclosed below. The HPET
basically needs "IRQNoFlags() {2, 8}", which causes XP to bluescreen.

So, I've made it conditional on the SMC STA method returning success
(0x0B).

The SMC node's STA method returns 0x0B unconditionally on real
hardware. So I was planning on figuring out what's easier in the
context of the most recent QEMU code base:

	1. dynamically generating (during qemu runtime initialization)
	 a DSDT entry for SMC with hardcoded 0x0B STA method, whenever
	"--device isa-applesmc" is present on the qemu command line

or

	2. writing a static (compile-time) SMC node but with a slightly
	smarter _STA method, which returns 0x0B when "--device isa-applesmc"
	was given on the cmdline, or which returns 0x00 in the absence
	of "--device isa-applesmc".

Either 1. or 2. could be used with HPET -- I can make inclusion of
IRQNoFlags dependent on either the success or on the presence of
SMC._STA() :)

Let me know what you think.

Thanks,
--Gabriel

###############################################################################
# Modify DSDT entry for HPET: conditionally insert "IRQNoFlags() {2, 8}" into
# _CRS method only if an AppleSMC DSDT node is also present and enabled (it
# otherwise causes WinXP to BSOD).
###############################################################################
diff --git a/hw/i386/acpi-dsdt-hpet.dsl b/hw/i386/acpi-dsdt-hpet.dsl
index dfde174..205cf05 100644
--- a/hw/i386/acpi-dsdt-hpet.dsl
+++ b/hw/i386/acpi-dsdt-hpet.dsl
@@ -38,14 +38,23 @@ Scope(\_SB) {
             }
             Return (0x0F)
         }
-        Name(_CRS, ResourceTemplate() {
-#if 0       /* This makes WinXP BSOD for not yet figured reasons. */
-            IRQNoFlags() {2, 8}
-#endif
+        Name(RESP, ResourceTemplate() {
             Memory32Fixed(ReadOnly,
                 0xFED00000,         // Address Base
                 0x00000400,         // Address Length
                 )
         })
+        Name(RESI, ResourceTemplate() {
+            IRQNoFlags() {2, 8}
+        })
+        Method(_CRS, 0) {
+            Store(\_SB.PCI0.ISA.SMC._STA(), Local0)
+            If (LEqual(Local0, 0x0B)) {        // AppleSMC present, add IRQ
+                ConcatenateResTemplate(RESP, RESI, Local1)
+                Return (Local1)
+            } else {
+                Return (RESP)
+            }
+        }
     }
 }
###############################################################################
# Add DSDT entry for AppleSMC;
# TODO: find a way to make the _STA method return 0x0b only if QEMU command
# line contains "-device isa-applesmc", and 0x00 otherwise!
###############################################################################
diff --git a/hw/i386/acpi-dsdt-isa.dsl b/hw/i386/acpi-dsdt-isa.dsl
index 89caa16..b7a27bb 100644
--- a/hw/i386/acpi-dsdt-isa.dsl
+++ b/hw/i386/acpi-dsdt-isa.dsl
@@ -16,6 +16,28 @@
 /* Common legacy ISA style devices. */
 Scope(\_SB.PCI0.ISA) {
 
+    Device (SMC) {
+        Name(_HID, EisaId("APP0001"))
+        OperationRegion(SMC, SystemIO, 0x0300, 0x20)
+        Field(SMC, ByteAcc, NoLock, Preserve) {
+            Offset(0x04),
+            CMDP, 8,
+        }
+        Method(_STA, 0) {
+//            Store(0x10, CMDP)    // APPLESMC_READ_CMD
+//            Store(CMDP, Local0)
+//            If (LEqual(Local0, 0x0c)) {
+                Return (0x0B)
+//            } Else {
+//                Return (0x00)
+//            }
+        }
+        Name (_CRS, ResourceTemplate () {
+            IO (Decode16, 0x0300, 0x0300, 0x01, 0x20)
+            IRQNoFlags() { 6 }
+        })
+    }
+
     Device(RTC) {
         Name(_HID, EisaId("PNP0B00"))
         Name(_CRS, ResourceTemplate() {

Re: [Qemu-devel] os x boot broken by commit 11948748495841bd54721b250d68c7b3cb0475ef

[Paolo Bonzini]

Il 21/11/2013 23:02, Gabriel L. Somlo ha scritto:
> On Thu, Nov 21, 2013 at 07:14:27PM +0100, Paolo Bonzini wrote:
>> > Can you remind us about your DSDT modifications?  It should be possible
>> > to patch the HPET and applesmc bits appropriately from QEMU (or to move
>> > them from the DSDT to an SSDT that is built entirely in QEMU).
>> > 
>> > It actually isn't impossible that Mac OS X would boot just fine with 1.8...
> My current DSDT patch (against QEMU) is enclosed below. The HPET
> basically needs "IRQNoFlags() {2, 8}", which causes XP to bluescreen.

The IRQNoFlags(){2,8} setting makes sense if the general configuration
register of the HPET has bits 0..1=1 (HPET enabled = 1 and HPET legacy
replacement route = 1).

That would be something like

     Field(HPTM, DWordAcc, Lock, Preserve) {
         VEND, 32,
         PRD, 32,
         UNUS, 32
         GCNF, 32
     }

     ...

     Method(_CRS, 0) {
        Store(GCNF, Local0)
        If (LEqual(LAnd(Local0, 3), 3)) {   // Legacy replacement route
             ConcatenateResTemplate(RESP, RESI, Local1)
             Return (Local1)
        } else {
             Return (RESP)
        }
    }

If that doesn't work, there are various choices here...

(1) Does Mac OS work if you add a _PRS with IRQNoFlags and
Memory32Fixed, but leave _CRS as it is?

(2) does it work with -no-hpet?

(3) you could also make that dependent on _OSI("Darwin").  It's unlikely
that Linux and/or Windows expose _OSI("Darwin"), and anyway the BSOD is
only there for Windows XP as I understand it.

> So, I've made it conditional on the SMC STA method returning success
> (0x0B).

That would mean that running Windows XP on "Mac OS X hardware" breaks,
though.

> The SMC node's STA method returns 0x0B unconditionally on real
> hardware. So I was planning on figuring out what's easier in the
> context of the most recent QEMU code base:
> 
> 	1. dynamically generating (during qemu runtime initialization)
> 	 a DSDT entry for SMC with hardcoded 0x0B STA method, whenever
> 	"--device isa-applesmc" is present on the qemu command line
> 
> or
> 
> 	2. writing a static (compile-time) SMC node but with a slightly
> 	smarter _STA method, which returns 0x0B when "--device isa-applesmc"
> 	was given on the cmdline, or which returns 0x00 in the absence
> 	of "--device isa-applesmc".

Either would work.  See acpi_get_misc_info and patch_ssdt in
hw/i386/acpi-build.c.

I think device-dependent ACPI stuff should become a QOM interface, but
you need not do that.

Paolo

Re: [Qemu-devel] os x boot broken by commit 11948748495841bd54721b250d68c7b3cb0475ef

[Michael S. Tsirkin]

On Thu, Nov 21, 2013 at 05:02:07PM -0500, Gabriel L. Somlo wrote:
> Added qemu-devel, since that is where this stuff belongs now. Everyone
> else, sorry for the dupe...
> 
> On Thu, Nov 21, 2013 at 07:14:27PM +0100, Paolo Bonzini wrote:
> > Can you remind us about your DSDT modifications?  It should be possible
> > to patch the HPET and applesmc bits appropriately from QEMU (or to move
> > them from the DSDT to an SSDT that is built entirely in QEMU).
> > 
> > It actually isn't impossible that Mac OS X would boot just fine with 1.8...
> 
> My current DSDT patch (against QEMU) is enclosed below. The HPET
> basically needs "IRQNoFlags() {2, 8}", which causes XP to bluescreen.
> 
> So, I've made it conditional on the SMC STA method returning success
> (0x0B).
> 
> The SMC node's STA method returns 0x0B unconditionally on real
> hardware. So I was planning on figuring out what's easier in the
> context of the most recent QEMU code base:
> 
> 	1. dynamically generating (during qemu runtime initialization)
> 	 a DSDT entry for SMC with hardcoded 0x0B STA method, whenever
> 	"--device isa-applesmc" is present on the qemu command line
> 
> or
> 
> 	2. writing a static (compile-time) SMC node but with a slightly
> 	smarter _STA method, which returns 0x0B when "--device isa-applesmc"
> 	was given on the cmdline, or which returns 0x00 in the absence
> 	of "--device isa-applesmc".
> 
> Either 1. or 2. could be used with HPET -- I can make inclusion of
> IRQNoFlags dependent on either the success or on the presence of
> SMC._STA() :)
> 
> Let me know what you think.
> 
> Thanks,
> --Gabriel

Without discussing whether this is the right thing to do,
main options to implement this would be:

        - write code in ASL, supply entry in ACPI always, patch some
          fields
          to enable/disable it dynamically

        see how pvpanic entry is disabled for an example

        - write code in ASL append it to ACPI if necessary

        see how CPU entries are added depending on number of CPUs for an example

        - generate code in AML directly

        see acpi based pci hotplug on pci branch in my tree for an example

Re: [Qemu-devel] os x boot broken by commit 11948748495841bd54721b250d68c7b3cb0475ef

[Michael S. Tsirkin]

On Fri, Nov 22, 2013 at 10:00:20AM +0100, Paolo Bonzini wrote:
> Il 21/11/2013 23:02, Gabriel L. Somlo ha scritto:
> > On Thu, Nov 21, 2013 at 07:14:27PM +0100, Paolo Bonzini wrote:
> >> > Can you remind us about your DSDT modifications?  It should be possible
> >> > to patch the HPET and applesmc bits appropriately from QEMU (or to move
> >> > them from the DSDT to an SSDT that is built entirely in QEMU).
> >> > 
> >> > It actually isn't impossible that Mac OS X would boot just fine with 1.8...
> > My current DSDT patch (against QEMU) is enclosed below. The HPET
> > basically needs "IRQNoFlags() {2, 8}", which causes XP to bluescreen.
> 
> The IRQNoFlags(){2,8} setting makes sense if the general configuration
> register of the HPET has bits 0..1=1 (HPET enabled = 1 and HPET legacy
> replacement route = 1).
> 
> That would be something like
> 
>      Field(HPTM, DWordAcc, Lock, Preserve) {
>          VEND, 32,
>          PRD, 32,
>          UNUS, 32
>          GCNF, 32
>      }
> 
>      ...
> 
>      Method(_CRS, 0) {
>         Store(GCNF, Local0)
>         If (LEqual(LAnd(Local0, 3), 3)) {   // Legacy replacement route
>              ConcatenateResTemplate(RESP, RESI, Local1)
>              Return (Local1)
>         } else {
>              Return (RESP)
>         }
>     }

Which reminds me. We run C preprocessor over the source so there's
no good reason to use 4-byte names anymore really (iasl also has
an integrated preprocessor but that only appeared in 2012, not
sure it's wise to rely on that).

So simply
#define HPET_MEMORY HPTM

and use HPET_MEMORY everywhere.

> If that doesn't work, there are various choices here...
> 
> (1) Does Mac OS work if you add a _PRS with IRQNoFlags and
> Memory32Fixed, but leave _CRS as it is?
> 
> (2) does it work with -no-hpet?
> 
> (3) you could also make that dependent on _OSI("Darwin").  It's unlikely
> that Linux and/or Windows expose _OSI("Darwin"), and anyway the BSOD is
> only there for Windows XP as I understand it.
> 
> > So, I've made it conditional on the SMC STA method returning success
> > (0x0B).
> 
> That would mean that running Windows XP on "Mac OS X hardware" breaks,
> though.
> 
> > The SMC node's STA method returns 0x0B unconditionally on real
> > hardware. So I was planning on figuring out what's easier in the
> > context of the most recent QEMU code base:
> > 
> > 	1. dynamically generating (during qemu runtime initialization)
> > 	 a DSDT entry for SMC with hardcoded 0x0B STA method, whenever
> > 	"--device isa-applesmc" is present on the qemu command line
> > 
> > or
> > 
> > 	2. writing a static (compile-time) SMC node but with a slightly
> > 	smarter _STA method, which returns 0x0B when "--device isa-applesmc"
> > 	was given on the cmdline, or which returns 0x00 in the absence
> > 	of "--device isa-applesmc".
> 
> Either would work.  See acpi_get_misc_info and patch_ssdt in
> hw/i386/acpi-build.c.
> 
> I think device-dependent ACPI stuff should become a QOM interface, but
> you need not do that.
> 
> Paolo