PPC: Initialize OpenPIC properly
PATCH 1/1 PPC: Initialize OpenPIC properly
When lowering an IRQ line, we search for the line we're supposed to lower.
Usually we run into an optimization there that queues up interrupts. This queue ends with -1. Unfortunately we didn't set the first item to -1.
This patch fixes this, making interrupts work on PPC64.
--- hw/openpic.c | 2 ++ 1 files changed, 2 insertions(+), 0 deletions(-)
hw/openpic.c +2 -0
@@ -442,7 +442,9 @@ static void openpic_reset (void *opaque) opp->dst[i].pctp =0x0000000F ; opp->dst[i].pcsr =0x00000000 ;memset (&opp->dst[i].raised,0 ,sizeof (IRQ_queue_t)); + opp->dst[i].raised.next =-1 ;memset (&opp->dst[i].servicing,0 ,sizeof (IRQ_queue_t)); + opp->dst[i].servicing.next =-1 ; }/* Initialise timers */ for (i =0 ; i <MAX_TMR ; i++) {
Signature · 1 line
1.6.0.2
PATCH 1/1 PPC: Use interrupts for escc
When using the escc with Linux, we need interrupts. So instead of creating a dummy device, let's just map them to the openpic we have anyways.
This makes Linux on PPC64 with console=ttyPZ0 work.
Obviously, this change needs to be reflected in openbios. Patch for that follows this one. Please update the binary then.
--- hw/ppc_newworld.c | 6 +----- 1 files changed, 1 insertions(+), 5 deletions(-)
hw/ppc_newworld.c +1 -5
@@ -108,7 +108,6 @@ static void ppc_core99_init (ram_addr_t ram_size,MacIONVRAMState *nvr;int nvram_mem_index;int vga_bios_size, bios_size; -qemu_irq *dummy_irq;int pic_mem_index, dbdma_mem_index, cuda_mem_index, escc_mem_index;int ppc_boot_device;DriveInfo *hd[MAX_IDE_BUS *MAX_IDE_DEVS ];
@@ -318,10 +317,7 @@ static void ppc_core99_init (ram_addr_t ram_size,/* init basic PC hardware */ pci_vga_init (pci_bus, vga_bios_offset, vga_bios_size); -/* XXX: suppress that */ - dummy_irq =i8259_init (NULL ); - - escc_mem_index =escc_init (0x80013000 , dummy_irq[4 ], dummy_irq[5 ], + escc_mem_index =escc_init (0x80013000 , pic[0x25 ], pic[0x24 ], serial_hds[0 ], serial_hds[1 ],ESCC_CLOCK ,4 );for (i =0 ; i < nb_nics; i++)
Signature · 1 line
1.6.0.2
Re: [Qemu-devel] [PATCH] PPC: Use interrupts for escc
Am 18.12.2009 um 23:37 schrieb Alexander Graf:
Quoted 4 lines
When using the escc with Linux, we need interrupts. So instead of creating a dummy device, let's just map them to the openpic we have anyways.
This makes Linux on PPC64 with console=ttyPZ0 work.
With these patches and OpenBIOS r646, booting 'install64' on the following Debian CDs worked for me:
debian-40r4a-powerpc-netinst.iso
debian-503-powerpc-netinst.iso
By default however you get the following error:
invalid/unsupported opcode: 1e - 12 - 1b (782106e4) 00000000014080e4 1 invalid/unsupported opcode: 00 - 00 - 00 (00000000) 0000000000008734 0
which translates to: qemu-system-ppc64 is using a 32-bit CPU by default.
'-cpu ppc64' (970FX according to OpenBIOS) worked, while '-cpu 970' was not recognized by OpenBIOS.
If you change only the CPU to 'ppc64', it will be rejected:
qemu: hardware error: Bus model not supported on OldWorld Mac machine
That's why Alex mentioned '-M mac99' elsewhere, I guess. Can't we change the defaults? I believe the G5 was the first 64-bit Mac and is considered New World, so Old World shouldn't even be selectable IMO.
install64 console=ttyPZ0 only gets me to the penguin and appears to hang...install64 gets me through to the installer screen on Linux/amd64 host. There are merely some FATALs visible while trying to load drivers for therm_* and windfarm_* devices ("No such device").
Andreas
Quoted 40 lines
Obviously, this change needs to be reflected in openbios. Patch for that follows this one. Please update the binary then.
Signed-off-by Alexander Graf <[email protected]>---
hw/ppc_newworld.c| 6 +-----1 files changed, 1 insertions(+), 5 deletions(-)hw/ppc_newworld.c +0
-1@@ -108,7 +108,6 @@ static void ppc_core99_init (ram_addr_t ram_size,MacIONVRAMState *nvr;int nvram_mem_index;int vga_bios_size, bios_size; -qemu_irq *dummy_irq;int pic_mem_index, dbdma_mem_index, cuda_mem_index, escc_mem_index;int ppc_boot_device;DriveInfo *hd[MAX_IDE_BUS * MAX_IDE_DEVS];hw/ppc_newworld.c +1
-4@@ -318,10 +317,7 @@ static void ppc_core99_init (ram_addr_t ram_size,/* init basic PC hardware */ pci_vga_init (pci_bus, vga_bios_offset, vga_bios_size); -/* XXX: suppress that */ - dummy_irq =i8259_init (NULL ); - - escc_mem_index =escc_init (0x80013000 , dummy_irq[4 ], dummy_irq[5 ], + escc_mem_index =escc_init (0x80013000 , pic[0x25 ], pic[0x24 ], serial_hds[0 ], serial_hds[1 ],ESCC_CLOCK ,4 );for(i = 0; i < nb_nics; i++)Signature · 1 line
1.6.0.2
Re: [Qemu-devel] [PATCH] PPC: Use interrupts for escc
Am 19.12.2009 um 22:58 schrieb Andreas Färber <[email protected]>:
Quoted 11 lines
Am 18.12.2009 um 23:37 schrieb Alexander Graf:
Quoted 4 lines
When using the escc with Linux, we need interrupts. So instead of creating a dummy device, let's just map them to the openpic we have anyways.
This makes Linux on PPC64 with console=ttyPZ0 work.
With these patches and OpenBIOS r646, booting 'install64' on the following Debian CDs worked for me:
debian-40r4a-powerpc-netinst.iso
debian-503-powerpc-netinst.iso
Nice, interesting!
Quoted 8 lines
By default however you get the following error:
invalid/unsupported opcode: 1e - 12 - 1b (
782106e4)00000000014080e41 invalid/unsupported opcode: 00 - 00 - 00 (00000000) 0000000000008734 0which translates to: qemu-system-ppc64 is using a 32-bit CPU by default.
'-cpu ppc64' (970FX according to OpenBIOS) worked, while '-cpu 970' was not recognized by OpenBIOS.
Yes, I only test whatever I do with -cpu 970fx.
Quoted 5 lines
If you change only the CPU to 'ppc64', it will be rejected:
qemu: hardware error: Bus model not supported on OldWorld Mac machineThat's why Alex mentioned '-M mac99' elsewhere, I guess. Can't we change the defaults? I believe the G5 was the first 64-bit Mac and is considered New World, so Old World shouldn't even be selectable IMO.
I agree. Please send a patch, CC me and I'll ack it.
Quoted 2 lines
install64 console=ttyPZ0only gets me to the penguin and appears to hang...
Well, on older kernels this was console=ttyS0.
Quoted 1 line
install64gets me through to the installer screen on Linux/amd64 host. There are merely some FATALs visible while trying to load drivers for therm_* and windfarm_* devices ("No such device").
Yesh, ADB shouldn't work either so you don't get input devices :-(.
Alex
Quoted 45 lines
Andreas
Quoted 41 lines
Obviously, this change needs to be reflected in openbios. Patch for that follows this one. Please update the binary then.
Signed-off-by Alexander Graf <[email protected]>---
hw/ppc_newworld.c| 6 +-----1 files changed, 1 insertions(+), 5 deletions(-)hw/ppc_newworld.c +0
-1@@ -108,7 +108,6 @@ static void ppc_core99_init (ram_addr_t ram_size,MacIONVRAMState *nvr;int nvram_mem_index;int vga_bios_size, bios_size; -qemu_irq *dummy_irq;int pic_mem_index, dbdma_mem_index, cuda_mem_index, escc_mem_index;int ppc_boot_device;DriveInfo *hd[MAX_IDE_BUS * MAX_IDE_DEVS];hw/ppc_newworld.c +1
-4@@ -318,10 +317,7 @@ static void ppc_core99_init (ram_addr_t ram_size,/* init basic PC hardware */ pci_vga_init(pci_bus, vga_bios_offset, vga_bios_size); -/* XXX: suppress that */ - dummy_irq =i8259_init (NULL ); - - escc_mem_index = escc_init(0x80013000, dummy_irq[4], dummy_irq [5 ], + escc_mem_index = escc_init(0x80013000, pic[0x25], pic[0x24], serial_hds[0], serial_hds[1],
ESCC_CLOCK, 4);for(i = 0; i < nb_nics; i++)Signature · 1 line
1.6.0.2