Author: mcayland
Date: Fri Jan 8 13:09:44 2016
New Revision: 1375
URL: http://tracker.coreboot.org/trac/openbios/changeset/1375
Log:
pci: enable AAPL,address property for all Apple PPC machines
The property appears in device trees for both Old World and New World Macs
so let's do the same.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk>
Reviewed-by: Alexander Graf <agraf(a)suse.de>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk>
Modified:
trunk/openbios-devel/drivers/pci.c
Modified: trunk/openbios-devel/drivers/pci.c
==============================================================================
--- trunk/openbios-devel/drivers/pci.c Fri Jan 8 13:09:41 2016 (r1374)
+++ trunk/openbios-devel/drivers/pci.c Fri Jan 8 13:09:44 2016 (r1375)
@@ -1020,7 +1020,10 @@
}
pci_set_assigned_addresses(phandle, config, num_bars);
- OLDWORLD(pci_set_AAPL_address(config));
+
+ if (is_apple()) {
+ pci_set_AAPL_address(config);
+ }
PCI_DPRINTF("\n");
}
Author: mcayland
Date: Fri Jan 8 13:09:41 2016
New Revision: 1374
URL: http://tracker.coreboot.org/trac/openbios/changeset/1374
Log:
pci: fix AAPL,address property for IO space mappings
According to "Designing PCI Cards and Drivers for Power Macintosh Computers" the
AAPL,address property should be an array of virtual addresses representing the
bus addresses in the assigned-addresses array.
While this works for PCI memory addresses, IO addresses are incorrect because
the calculated address is still the PCI bus address. Switch over to using the
new space-aware pci_bus_addr_to_host_addr() so that addresses in both memory
and IO space are correctly calculated.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk>
Reviewed-by: Alexander Graf <agraf(a)suse.de>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk>
Modified:
trunk/openbios-devel/drivers/pci.c
Modified: trunk/openbios-devel/drivers/pci.c
==============================================================================
--- trunk/openbios-devel/drivers/pci.c Fri Jan 8 13:09:38 2016 (r1373)
+++ trunk/openbios-devel/drivers/pci.c Fri Jan 8 13:09:41 2016 (r1374)
@@ -597,13 +597,18 @@
{
phandle_t dev = get_cur_dev();
cell props[7];
- int ncells, i;
+ uint32_t mask;
+ int ncells, i, flags, space_code;
ncells = 0;
for (i = 0; i < 6; i++) {
if (!config->assigned[i] || !config->sizes[i])
continue;
- props[ncells++] = config->assigned[i] & ~0x0000000F;
+ pci_decode_pci_addr(config->assigned[i],
+ &flags, &space_code, &mask);
+
+ props[ncells++] = pci_bus_addr_to_host_addr(space_code,
+ config->assigned[i] & ~mask);
}
if (ncells)
set_property(dev, "AAPL,address", (char *)props,
This patchset is a result of work to fix the rtl8139 card under Apple PPC
machines in QEMU (both g3beige and mac99).
Primarily it fixes up pci_bus_addr_to_host_addr() to support both memory
and IO spaces in order to generate a correct AAPL,address
property for IO space mappings. As well as this, it removes the
configuration space range from the PCI host bridge for all machines except
SPARC64 which fixes an issue with Darwin/OS X calculating the wrong address
spaces for PCI accesses.
As a final tidy-up the last patch in the series is a nice rework which
properly generates the PCI node interrupt-map property from the device tree
rather than using hardcoded (and incorrect) properties for openpic.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk>
v2:
- Updated cover letter
- Remove OLDWORLD() macro typo
- Add constants suggested by Alex
- Include extra patch to tidy up interrupt-map property
Mark Cave-Ayland (6):
pci: switch ob_pci_map_in() to use physical addresses rather than
region and size
pci: support PCI spaces other than memory in
pci_bus_addr_to_host_addr()
pci: fix AAPL,address property for IO space mappings
pci: enable AAPL,address property for all Apple PPC machines
pci: remove the configuration space range from the PCI host bridge by
default
pci: don't hard-code PCI interrupt maps by default
openbios-devel/drivers/pci.c | 130 ++++++++++++++++++++++++-----------------
openbios-devel/drivers/pci.fs | 68 +++------------------
openbios-devel/drivers/pci.h | 9 +++
openbios-devel/drivers/vga.fs | 11 ++--
4 files changed, 100 insertions(+), 118 deletions(-)
--
1.7.10.4
On Jan 3, 2016, at 2:27 PM, Mark Cave-Ayland wrote:
> On 03/01/16 19:01, Programmingkid wrote:
>
>>>> The rtl8139 did not work on the mac99 target. I didn't see the timeout message.
>>>> I did use your rtl8139 patchset fix.
>>>
>>> I think this might be because we're missing an interrupt mapping as they
>>> are currently hard-coded in OpenBIOS.
>>>
>>> Quick hack alert: take a look at OpenBIOS's drivers/pci.c and in
>>> particular ob_pci_host_set_interrupt_map(). See how at the moment we
>>> patch in hard-coded interrupt-parent nodes to link each device node to
>>> the interrupt controller? Maybe try adding something like the following
>>> for the rtl8139 card:
>>>
>>> target_node = find_dev("/pci/pci10ec,8139");
>>> set_int_property(target_node, "interrupt-parent", dnode);
>>>
>>> Does that then get the rtl8139 card to work under -M mac99 correctly?
>>
>> Sorry it doesn't. I paste your code right below the "if (dnode) {".
>
> Ah well. Thought it was worth a quick shot.
I did find out that the driver's init() and start() methods are never called. Using the ioreg command I can see the rtl8139 in the PCI slot. The NIC does work in Linux using the mac99 target. Note: I implemented your second set of patches for the rtl8139 and the booting from cd patchset.
This patchset primarily fixes up pci_bus_addr_to_host_addr() to support
both memory and IO spaces in order to generate a correct AAPL,address
property for IO space mappings. As well as this, it removes the
configuration space range from the PCI host bridge for all machines except
SPARC64 which fixes an issue with Darwin/OS X calculating the wrong address
spaces for PCI accesses.
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk>
Mark Cave-Ayland (5):
pci: switch ob_pci_map_in() to use physical addresses rather than
region and size
pci: support PCI spaces other than memory in
pci_bus_addr_to_host_addr()
pci: fix AAPL,address property for IO space mappings
pci: enable AAPL,address property for all Apple PPC machines
pci: remove the configuration space range from the PCI host bridge by
default
openbios-devel/drivers/pci.c | 56 ++++++++++++++++++++++++---------
openbios-devel/drivers/pci.fs | 68 +++++------------------------------------
openbios-devel/drivers/vga.fs | 11 ++++---
3 files changed, 56 insertions(+), 79 deletions(-)
--
1.7.10.4
Author: mcayland
Date: Mon Jan 4 13:03:40 2016
New Revision: 1371
URL: http://tracker.coreboot.org/trac/openbios/changeset/1371
Log:
macio: switch over to use 2 ata-3 ide controllers
This better matches real device trees and fixes Darwin/OS X boot from CDROM
under the QEMU mac99 machine (it seems that IOKit will only detect macio IDE
devices connected to an ata-3 ide controller rather than the previous ata-4
ide controller).
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk>
Reviewed-by: Alexander Graf <agraf(a)suse.de>
Signed-off-by: Mark Cave-Ayland <mark.cave-ayland(a)ilande.co.uk>
Modified:
trunk/openbios-devel/drivers/ide.c
trunk/openbios-devel/drivers/pci.c
Modified: trunk/openbios-devel/drivers/ide.c
==============================================================================
--- trunk/openbios-devel/drivers/ide.c Mon Jan 4 13:03:37 2016 (r1370)
+++ trunk/openbios-devel/drivers/ide.c Mon Jan 4 13:03:40 2016 (r1371)
@@ -1542,10 +1542,10 @@
struct ide_channel *chan;
/* IDE ports on Macs are numbered from 3.
- * Also see comments in macio.c:openpic_init() */
+ * Also see comments in pci.c:ob_pci_host_set_interrupt_map() */
current_channel = 3;
- for (i = 0; i < nb_channels; i++, current_channel++) {
+ for (i = 0; i < nb_channels; i++) {
chan = malloc(sizeof(struct ide_channel));
@@ -1651,7 +1651,7 @@
OLDWORLD(set_property(dnode, "AAPL,address",
(char *)&props, 2*sizeof(props[0])));
- props[0] = 0;
+ props[0] = i;
set_property(dnode, "AAPL,bus-id", (char*)props,
1 * sizeof(props[0]));
IDE_DPRINTF(DEV_NAME": [io ports 0x%lx]\n",
Modified: trunk/openbios-devel/drivers/pci.c
==============================================================================
--- trunk/openbios-devel/drivers/pci.c Mon Jan 4 13:03:37 2016 (r1370)
+++ trunk/openbios-devel/drivers/pci.c Mon Jan 4 13:03:40 2016 (r1371)
@@ -1430,12 +1430,12 @@
/* On a new world Mac these are not numbered but named by the
* ATA version they support. Thus we have: ata-3, ata-3, ata-4
* On g3beige they all called just ide.
- * We take ata-3 and ata-4 which seems to work for both
- * at least for clients we care about */
- target_node = find_dev("/pci/mac-io/ata-3");
+ * We take 2 x ata-3 buses which seems to work for
+ * at least the clients we care about */
+ target_node = find_dev("/pci/mac-io/ata-3@20000");
set_int_property(target_node, "interrupt-parent", dnode);
- target_node = find_dev("/pci/mac-io/ata-4");
+ target_node = find_dev("/pci/mac-io/ata-3@21000");
set_int_property(target_node, "interrupt-parent", dnode);
target_node = find_dev("/pci/mac-io/via-cuda");
