QEMU: Fix RISC-V virt & spike machines' dtbs
v2 cover QEMU: Fix RISC-V virt & spike machines' dtbs
From: Conor Dooley <[email protected]>
The device trees produced automatically for the virt and spike machines fail dt-validate on several grounds. Some of these need to be fixed in the linux kernel's dt-bindings, but others are caused by bugs in QEMU.
I mostly opted for what appeared to be the smallest change that would fix the warnings, partly due to my inexperience with the QEMU codebase. A "sister" patchset for the kernel will clear the remaining warnings [0]. Thanks to Rob Herring for reporting these issues [1],
Conor.
Changes since v1:
- drop patch 1
To reproduce the errors:
./build/qemu-system-riscv64 -nographic -machine virt,dumpdtb=qemu.dtb
dt-validate -p /path/to/linux/kernel/Documentation/devicetree/bindings/processed-schema.json qemu.dtb (The processed schema needs to be generated first)
0 - https://lore.kernel.org/linux-riscv/[email protected]
1 - https://lore.kernel.org/linux-riscv/[email protected]
Conor Dooley (4):
hw/riscv: virt: fix uart node name
hw/riscv: virt: Fix the plic's address cells
hw/riscv: virt: fix syscon subnode paths
hw/core: fix platform bus node name
hw/core/sysbus-fdt.c | 2 +- hw/riscv/virt.c | 10 +++++++--- include/hw/riscv/virt.h | 1 + 3 files changed, 9 insertions(+), 4 deletions(-)
base-commit: 2480f3bbd03814b0651a1f74959f5c6631ee5819
Signature · 1 line
2.37.1
PATCH v2 1/4 hw/riscv: virt: fix uart node name
From: Conor Dooley <[email protected]>
"uart" is not a node name that complies with the dt-schema.
Change the node name to "serial" to ix warnings seen during
dt-validate on a dtbdump of the virt machine such as:
/stuff/qemu/qemu.dtb: uart@10000000: $nodename:0: 'uart@10000000' does not match '^serial(@.*)?$'
From schema: /stuff/linux/Documentation/devicetree/bindings/serial/8250.yaml
--- hw/riscv/virt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
hw/riscv/virt.c +1 -1
@@ -917,7 +917,7 @@ static void create_fdt_uart(RISCVVirtState *s, const MemMapEntry *memmap,char *name;MachineState *mc =MACHINE (s); - name =g_strdup_printf ("/soc/uart@%lx" , (long )memmap[VIRT_UART0 ].base); + name =g_strdup_printf ("/soc/serial@%lx" , (long )memmap[VIRT_UART0 ].base);qemu_fdt_add_subnode (mc->fdt, name);qemu_fdt_setprop_string (mc->fdt, name,"compatible" ,"ns16550a" ); qemu_fdt_setprop_cells(mc->fdt, name, "reg",
Signature · 1 line
2.37.1
PATCH v2 2/4 hw/riscv: virt: Fix the plic's address cells
In reply to Conor DooleyFrom: Conor Dooley <[email protected]>
When optional AIA PLIC support was added the to the virt machine, the
address cells property was removed leading the issues with dt-validate
on a dump from the virt machine:
/stuff/qemu/qemu.dtb: plic@c000000: '#address-cells' is a required property
From schema: /stuff/linux/Documentation/devicetree/bindings/interrupt-controller/sifive,plic-1.0.0.yaml
Add back the property to suppress the warning.
--- hw/riscv/virt.c | 2 ++ include/hw/riscv/virt.h | 1 + 2 files changed, 3 insertions(+)
hw/riscv/virt.c +2 -0
@@ -465,6 +465,8 @@ static void create_fdt_socket_plic(RISCVVirtState *s,qemu_fdt_add_subnode (mc->fdt, plic_name);qemu_fdt_setprop_cell (mc->fdt, plic_name,"#interrupt-cells" ,FDT_PLIC_INT_CELLS ); +qemu_fdt_setprop_cell (mc->fdt, plic_name, +"#address-cells" ,FDT_PLIC_ADDR_CELLS );qemu_fdt_setprop_string_array (mc->fdt, plic_name,"compatible" , (char **)&plic_compat,ARRAY_SIZE (plic_compat));
include/hw/riscv/virt.h +1 -0
@@ -111,6 +111,7 @@ enum {#define FDT_PCI_ADDR_CELLS 3#define FDT_PCI_INT_CELLS 1 +#define FDT_PLIC_ADDR_CELLS 0#define FDT_PLIC_INT_CELLS 1#define FDT_APLIC_INT_CELLS 2#define FDT_IMSIC_INT_CELLS 0
Signature · 1 line
2.37.1
PATCH v2 3/4 hw/riscv: virt: fix syscon subnode paths
In reply to Conor DooleyFrom: Conor Dooley <[email protected]>
The subnodes of the syscon have been added to the incorrect paths. Rather than add them as subnodes, they were originally added to "/foo" and a later patch moved them to "/soc/foo". Both are incorrect & they should have been added as "/soc/test@###/foo" as "/soc/test" is the syscon node. Fix both the reboot and poweroff subnodes to avoid errors such as:
/stuff/qemu/qemu.dtb: soc: poweroff: {'value': [[21845]], 'offset': [[0]], 'regmap': [[4]], 'compatible': ['syscon-poweroff']} should not be valid under {'type': 'object'}
From schema: /home/conor/.local/lib/python3.9/site-packages/dtschema/schemas/simple-bus.yaml
/stuff/qemu/qemu.dtb: soc: reboot: {'value': [[30583]], 'offset': [[0]], 'regmap': [[4]], 'compatible': ['syscon-reboot']} should not be valid under {'type': 'object'}
From schema: /home/conor/.local/lib/python3.9/site-packages/dtschema/schemas/simple-bus.yaml
--- hw/riscv/virt.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
hw/riscv/virt.c +4 -2
@@ -896,7 +896,8 @@ static void create_fdt_reset(RISCVVirtState *s, const MemMapEntry *memmap, test_phandle =qemu_fdt_get_phandle (mc->fdt, name);g_free (name); - name =g_strdup_printf ("/soc/reboot" ); + name =g_strdup_printf ("/soc/test@%lx/reboot" , + (long )memmap[VIRT_TEST ].base);qemu_fdt_add_subnode (mc->fdt, name);qemu_fdt_setprop_string (mc->fdt, name,"compatible" ,"syscon-reboot" );qemu_fdt_setprop_cell (mc->fdt, name,"regmap" , test_phandle);
@@ -904,7 +905,8 @@ static void create_fdt_reset(RISCVVirtState *s, const MemMapEntry *memmap,qemu_fdt_setprop_cell (mc->fdt, name,"value" ,FINISHER_RESET );g_free (name); - name =g_strdup_printf ("/soc/poweroff" ); + name =g_strdup_printf ("/soc/test@%lx/poweroff" , + (long )memmap[VIRT_TEST ].base);qemu_fdt_add_subnode (mc->fdt, name);qemu_fdt_setprop_string (mc->fdt, name,"compatible" ,"syscon-poweroff" );qemu_fdt_setprop_cell (mc->fdt, name,"regmap" , test_phandle);
Signature · 1 line
2.37.1
Re: [PATCH v2 3/4] hw/riscv: virt: fix syscon subnode paths
Quoted 15 lines · Conor Dooley
From: Conor Dooley <[email protected]>
The subnodes of the syscon have been added to the incorrect paths. Rather than add them as subnodes, they were originally added to "/foo" and a later patch moved them to "/soc/foo". Both are incorrect & they should have been added as "/soc/test@###/foo" as "/soc/test" is the syscon node. Fix both the reboot and poweroff subnodes to avoid errors such as:
/stuff/qemu/qemu.dtb: soc: poweroff: {'value': [[21845]], 'offset': [[0]], 'regmap': [[4]], 'compatible': ['syscon-poweroff']} should not be valid under {'type': 'object'}
From schema:/home/conor/.local/lib/python3.9/site-packages/dtschema/schemas/simple-bus.yaml
/stuff/qemu/qemu.dtb: soc: reboot: {'value': [[30583]], 'offset': [[0]], 'regmap': [[4]], 'compatible': ['syscon-reboot']} should not be valid under {'type': 'object'}
From schema:/home/conor/.local/lib/python3.9/site-packages/dtschema/schemas/simple-bus.yaml
Quoted 6 lines
Reported-by Rob Herring <[email protected]>Fixes 18df0b4695 ("hw/riscv: virt: Allow creating multiple NUMA sockets")Fixes 0e404da007 ("riscv/virt: Add syscon reboot and poweroff DT nodes")Reviewed-by Alistair Francis <[email protected]>Signed-off-by Conor Dooley <[email protected]>
This breaks FreeBSD’s driver (well, it just won’t probe/attach), which is written to handle syscon-poweroff/reboot existing as a child of a simplebus not a syscon. Moreover, what is the point of regmap in this case? Its existence suggests the point is for them to *not* be children of the syscon, otherwise you wouldn’t need an explicit phandle, you’d just look at the parent. Moving the nodes whilst keeping the property doesn’t make sense to me.
Jess
Quoted 36 lines
---
hw/riscv/virt.c| 6 ++++--1 file changed, 4 insertions(+), 2 deletions(-)hw/riscv/virt.c +4
-2@@ -896,7 +896,8 @@ static void create_fdt_reset(RISCVVirtState *s, const MemMapEntry *memmap, test_phandle =qemu_fdt_get_phandle (mc->fdt, name);g_free (name); - name =g_strdup_printf ("/soc/reboot" ); + name =g_strdup_printf ("/soc/test@%lx/reboot" , + (long )memmap[VIRT_TEST ].base);qemu_fdt_add_subnode (mc->fdt, name);qemu_fdt_setprop_string (mc->fdt, name,"compatible" ,"syscon-reboot" );qemu_fdt_setprop_cell (mc->fdt, name,"regmap" , test_phandle);@@ -904,7 +905,8 @@ static void create_fdt_reset(RISCVVirtState *s, const MemMapEntry *memmap,qemu_fdt_setprop_cell (mc->fdt, name,"value" ,FINISHER_RESET );g_free (name); - name =g_strdup_printf ("/soc/poweroff" ); + name =g_strdup_printf ("/soc/test@%lx/poweroff" , + (long )memmap[VIRT_TEST ].base);qemu_fdt_add_subnode (mc->fdt, name);qemu_fdt_setprop_string (mc->fdt, name,"compatible" ,"syscon-poweroff" );qemu_fdt_setprop_cell (mc->fdt, name,"regmap" , test_phandle);Signature · 7 lines
2.37.1 _______________________________________________ linux-riscv mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-riscv
Re: [PATCH v2 3/4] hw/riscv: virt: fix syscon subnode paths
Quoted 23 lines · Jessica Clarke
Quoted 15 lines · Conor Dooley
From: Conor Dooley <[email protected]>
The subnodes of the syscon have been added to the incorrect paths. Rather than add them as subnodes, they were originally added to "/foo" and a later patch moved them to "/soc/foo". Both are incorrect & they should have been added as "/soc/test@###/foo" as "/soc/test" is the syscon node. Fix both the reboot and poweroff subnodes to avoid errors such as:
/stuff/qemu/qemu.dtb: soc: poweroff: {'value': [[21845]], 'offset': [[0]], 'regmap': [[4]], 'compatible': ['syscon-poweroff']} should not be valid under {'type': 'object'}
From schema:/home/conor/.local/lib/python3.9/site-packages/dtschema/schemas/simple-bus.yaml
/stuff/qemu/qemu.dtb: soc: reboot: {'value': [[30583]], 'offset': [[0]], 'regmap': [[4]], 'compatible': ['syscon-reboot']} should not be valid under {'type': 'object'}
From schema:/home/conor/.local/lib/python3.9/site-packages/dtschema/schemas/simple-bus.yamlQuoted 6 lines
Reported-by Rob Herring <[email protected]>Fixes 18df0b4695 ("hw/riscv: virt: Allow creating multiple NUMA sockets")Fixes 0e404da007 ("riscv/virt: Add syscon reboot and poweroff DT nodes")Reviewed-by Alistair Francis <[email protected]>Signed-off-by Conor Dooley <[email protected]>
Quoted 3 lines
This breaks FreeBSD’s driver (well, it just won’t probe/attach), which is written to handle syscon-poweroff/reboot existing as a child of a simplebus not a syscon.
I know next to nothing about FreeBSD unfortunately or how it handles buses. My understanding of simple-bus was that it is supposed to represent a bus with "things" mapped to addresses, relying on the "reg" property. And then syscon is used when there is some multifunction register region that controls multiple features of the hardware. Since simple-bus defines a reg property and the function nodes do not define one, I'd like to know how FreeBSD's driver handles that.
Quoted 5 lines
Moreover, what is the point of regmap in this
case? Its existence suggests the point is for them to *not* be children of the syscon, otherwise you wouldn’t need an explicit phandle, you’d just look at the parent. Moving the nodes whilst keeping the property doesn’t make sense to me.
That's how syscon bindings are constructed, makes it easier to follow I suppose if they functions are children of the syscon node. Strictly I think they don't need to be under the syscon itself, I think they can also go at the top level - they just aren't valid under the /soc node as it has been defined as a "simple-bus".
It would appear that the original patch 0e404da007 ("riscv/virt: Add syscon reboot and poweroff DT nodes") that added them put them at the top level and it was in the refactor that they got moved to the soc bus.* Maybe the solution would be to put them back at the top level?
dt-validate will consider it valid, but what does the FreeBSD driver think of that?
Thanks for your input Jess,
Conor.
* On reflection it looks like my fixes tags are not correct and that 0e404da007 was actually correct but the refactor broke things.
Quoted 40 lines
Jess
Quoted 36 lines
---
hw/riscv/virt.c| 6 ++++--1 file changed, 4 insertions(+), 2 deletions(-)hw/riscv/virt.c +4
-2@@ -896,7 +896,8 @@ static void create_fdt_reset(RISCVVirtState *s, const MemMapEntry *memmap, test_phandle =qemu_fdt_get_phandle (mc->fdt, name);g_free (name); - name =g_strdup_printf ("/soc/reboot" ); + name =g_strdup_printf ("/soc/test@%lx/reboot" , + (long )memmap[VIRT_TEST ].base);qemu_fdt_add_subnode (mc->fdt, name);qemu_fdt_setprop_string (mc->fdt, name,"compatible" ,"syscon-reboot" );qemu_fdt_setprop_cell (mc->fdt, name,"regmap" , test_phandle);@@ -904,7 +905,8 @@ static void create_fdt_reset(RISCVVirtState *s, const MemMapEntry *memmap,qemu_fdt_setprop_cell (mc->fdt, name,"value" ,FINISHER_RESET );g_free (name); - name =g_strdup_printf ("/soc/poweroff" ); + name =g_strdup_printf ("/soc/test@%lx/poweroff" , + (long )memmap[VIRT_TEST ].base);qemu_fdt_add_subnode (mc->fdt, name);qemu_fdt_setprop_string (mc->fdt, name,"compatible" ,"syscon-poweroff" );qemu_fdt_setprop_cell (mc->fdt, name,"regmap" , test_phandle);Signature · 7 lines
2.37.1 _______________________________________________ linux-riscv mailing list [email protected] http://lists.infradead.org/mailman/listinfo/linux-riscv
Re: [PATCH v2 3/4] hw/riscv: virt: fix syscon subnode paths
Quoted 25 lines · On Mon, Aug 8, 2022 at 4:10 PM <[email protected]> wrote:
Quoted 23 lines · Jessica Clarke
Quoted 21 lines · Conor Dooley
From: Conor Dooley <[email protected]>
The subnodes of the syscon have been added to the incorrect paths. Rather than add them as subnodes, they were originally added to "/foo" and a later patch moved them to "/soc/foo". Both are incorrect & they should have been added as "/soc/test@###/foo" as "/soc/test" is the syscon node. Fix both the reboot and poweroff subnodes to avoid errors such as:
/stuff/qemu/qemu.dtb: soc: poweroff: {'value': [[21845]], 'offset': [[0]], 'regmap': [[4]], 'compatible': ['syscon-poweroff']} should not be valid under {'type': 'object'}
From schema:/home/conor/.local/lib/python3.9/site-packages/dtschema/schemas/simple-bus.yaml
/stuff/qemu/qemu.dtb: soc: reboot: {'value': [[30583]], 'offset': [[0]], 'regmap': [[4]], 'compatible': ['syscon-reboot']} should not be valid under {'type': 'object'}
From schema:/home/conor/.local/lib/python3.9/site-packages/dtschema/schemas/simple-bus.yamlReported-by Rob Herring <[email protected]>Fixes 18df0b4695 ("hw/riscv: virt: Allow creating multiple NUMA sockets")Fixes 0e404da007 ("riscv/virt: Add syscon reboot and poweroff DT nodes")Reviewed-by Alistair Francis <[email protected]>Signed-off-by Conor Dooley <[email protected]>
Quoted 3 lines
Quoted 3 lines
This breaks FreeBSD’s driver (well, it just won’t probe/attach), which is written to handle syscon-poweroff/reboot existing as a child of a simplebus not a syscon.
It probably breaks Linux, too.
Quoted 20 lines
I know next to nothing about FreeBSD unfortunately or how it handles buses. My understanding of simple-bus was that it is supposed to represent a bus with "things" mapped to addresses, relying on the "reg" property. And then syscon is used when there is some multifunction register region that controls multiple features of the hardware. Since simple-bus defines a reg property and the function nodes do not define one, I'd like to know how FreeBSD's driver handles that.
Quoted 5 lines
Moreover, what is the point of regmap in this
case? Its existence suggests the point is for them to *not* be children of the syscon, otherwise you wouldn’t need an explicit phandle, you’d just look at the parent. Moving the nodes whilst keeping the property doesn’t make sense to me.That's how syscon bindings are constructed, makes it easier to follow I suppose if they functions are children of the syscon node. Strictly I think they don't need to be under the syscon itself, I think they can also go at the top level - they just aren't valid under the /soc node as it has been defined as a "simple-bus".
Quoted 4 lines
It would appear that the original patch
0e404da007("riscv/virt: Add syscon reboot and poweroff DT nodes") that added them put them at the top level and it was in the refactor that they got moved to the soc bus.* Maybe the solution would be to put them back at the top level?
Perhaps.
The other option is adding 'simple-mfd' to the 'test' node compatible. That would work for Linux. Not sure for FreeBSD.
Rob
Re: [PATCH v2 3/4] hw/riscv: virt: fix syscon subnode paths
Quoted 22 lines · Rob Herring
EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
Quoted 14 lines · On Mon, Aug 8, 2022 at 4:10 PM <[email protected]> wrote:
Quoted 5 lines · Jessica Clarke
Moreover, what is the point of regmap in this
case? Its existence suggests the point is for them to *not* be children of the syscon, otherwise you wouldn’t need an explicit phandle, you’d just look at the parent. Moving the nodes whilst keeping the property doesn’t make sense to me.That's how syscon bindings are constructed, makes it easier to follow I suppose if they functions are children of the syscon node. Strictly I think they don't need to be under the syscon itself, I think they can also go at the top level - they just aren't valid under the /soc node as it has been defined as a "simple-bus".
Quoted 4 lines
It would appear that the original patch
0e404da007("riscv/virt: Add syscon reboot and poweroff DT nodes") that added them put them at the top level and it was in the refactor that they got moved to the soc bus.* Maybe the solution would be to put them back at the top level?
Quoted 4 lines
Perhaps.
The other option is adding 'simple-mfd' to the 'test' node compatible. That would work for Linux. Not sure for FreeBSD.
Right, of course I was missing something in my understanding. The probe flow on Linux came back to me on the bike this morning after reading this and I felt like an idiot for missing that in the devicetrees I looked at!
@Jess, which does FreeBSD prefer? top level or add the extra compatible?
Thanks,
Conor.
PATCH v2 4/4 hw/core: fix platform bus node name
In reply to Conor DooleyFrom: Conor Dooley <[email protected]>
"platform" is not a valid name for a bus node in dt-schema, so warnings can be see in dt-validate on a dump of the riscv virt dtb:
/stuff/qemu/qemu.dtb: platform@4000000: $nodename:0: 'platform@4000000' does not match '^([a-z][a-z0-9\\-]+-bus|bus|soc|axi|ahb|apb)(@[0-9a-f]+)?$'
From schema: /home/conor/.local/lib/python3.9/site-packages/dtschema/schemas/simple-bus.yaml
"platform-bus" is a valid name, so use that instead.
--- hw/core/sysbus-fdt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
hw/core/sysbus-fdt.c +1 -1
@@ -539,7 +539,7 @@ void platform_bus_add_all_fdt_nodes(void *fdt, const char *intc, hwaddr addr,assert (fdt); - node =g_strdup_printf ("/platform@%" PRIx64, addr); + node =g_strdup_printf ("/platform-bus@%" PRIx64, addr);/* Create a /platform node that we can put all devices into */ qemu_fdt_add_subnode (fdt, node);
Signature · 1 line
2.37.1