hw: acpi: add UART clock freq to AArch64 SPCR
v2 cover hw: acpi: add UART clock freq to AArch64 SPCR
This series upgrades the ACPI SPCR table for AArch64 virt to revision 4, adding the UART clock frequency, which was not previously available via ACPI.
Regarding the NamespaceString: following the example of the SPCR table for RISC-V, I've set it to "." (indicating that there is no device corresponding to the UART in the ACPI namespace). However, looking at build_dsdt(), the UART is actually added to the DSDT, as "\_SB.COM0" (same as RISC-V). Should the NamespaceString be that instead?
Thanks,
Vadim
Vadim Chichikalyuk (3):
tests: acpi: whitelist expected blobs
hw: arm: acpi: upgrade SPCR table to revision 4
tests: acpi: update expected blobs
hw/arm/virt-acpi-build.c | 14 ++++++++------ tests/data/acpi/aarch64/virt/SPCR | Bin 80 -> 90 bytes 2 files changed, 8 insertions(+), 6 deletions(-)
Signature · 1 line
2.39.5 (Apple Git-154)
PATCH v2 1/3 tests: acpi: whitelist expected blobs
--- tests/qtest/bios-tables-test-allowed-diff.h | 1 + 1 file changed, 1 insertion(+)
tests/qtest/bios-tables-test-allowed-diff.h +1 -0
@@ -1 +1,2 @@/* List of comma-separated changed AML files to ignore */ +"tests/data/acpi/aarch64/virt/SPCR",
Signature · 1 line
2.39.5 (Apple Git-154)
PATCH v2 2/3 hw: arm: acpi: upgrade SPCR table to revision 4
In reply to Vadim ChichikalyukOn the ARM virt machine, there is currently no way to programmatically discover the frequency of the UART reference clock solely through the use of UEFI/ACPI (without the DTB). The SPCR table can include this information as of revision 4, but is currently at revision 2.
Bump the revision to 4, add the clock frequency of 24 MHz and complete the other new fields.
--- hw/arm/virt-acpi-build.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-)
hw/arm/virt-acpi-build.c +8 -6
@@ -536,6 +536,7 @@ build_iort(GArray *table_data, BIOSLinker *linker, VirtMachineState *vms)static void spcr_setup (GArray *table_data,BIOSLinker *linker,VirtMachineState *vms) { +const char name[] ="." ;AcpiSpcrData serial = { .interface_type =3 ,/* ARM PL011 UART */ .base_addr.id =AML_AS_SYSTEM_MEMORY ,
@@ -559,13 +560,14 @@ spcr_setup(GArray *table_data, BIOSLinker *linker, VirtMachineState *vms) .pci_function = 0, .pci_flags = 0, .pci_segment = 0, + .uart_clk_freq = 24000000, + .precise_baudrate = 0, + .namespace_string_length = sizeof(name), + .namespace_string_offset = 88, }; -/* -* Passing NULL as the SPCR Table for Revision 2 doesn't support -* NameSpaceString. -*/ -build_spcr (table_data, linker, &serial,2 , vms->oem_id, vms->oem_table_id, -NULL ); + +build_spcr (table_data, linker, &serial,4 , vms->oem_id, vms->oem_table_id, + name); }/*
Signature · 1 line
2.39.5 (Apple Git-154)
Re: [PATCH v2 2/3] hw: arm: acpi: upgrade SPCR table to revision 4
Quoted 8 lines · Vadim Chichikalyuk
On the ARM virt machine, there is currently no way to programmatically discover the frequency of the UART reference clock solely through the use of UEFI/ACPI (without the DTB). The SPCR table can include this information as of revision 4, but is currently at revision 2.
Bump the revision to 4, add the clock frequency of 24 MHz and complete the other new fields.
This commit message should say why we need to do this, i.e. why we can't continue to do what we've always done and not tell the guest about a clock frequency which is entirely imaginary in any case.
Quoted 18 lines
Signed-off-by Vadim Chichikalyuk <[email protected]>--- hw/arm/virt-acpi-build.c | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-)hw/arm/virt-acpi-build.c +1
-0@@ -536,6 +536,7 @@ build_iort(GArray *table_data, BIOSLinker *linker, VirtMachineState *vms)static void spcr_setup (GArray *table_data,BIOSLinker *linker,VirtMachineState *vms) { +const char name[] ="." ;AcpiSpcrData serial = { .interface_type =3 ,/* ARM PL011 UART */ .base_addr.id =AML_AS_SYSTEM_MEMORY ,
hw/arm/virt-acpi-build.c spcr_setup(GArray *table_data, BIOSLinker *linker, VirtMachineState *vms) · line 563
Quoted 4 lines
hw/arm/virt-acpi-build.c +1
-0@@ -559,13 +560,14 @@ spcr_setup(GArray *table_data, BIOSLinker *linker, VirtMachineState *vms) .pci_function = 0, .pci_flags = 0, .pci_segment = 0, + .uart_clk_freq = 24000000,
This is a number which ought to match the existing one in create_fdt() in hw/arm/virt.c. If we need it in two different files we should #define it as a constant in include/hw/arm/virt.h.
hw/arm/virt-acpi-build.c spcr_setup(GArray *table_data, BIOSLinker *linker, VirtMachineState *vms) · line 566
Quoted 3 lines
hw/arm/virt-acpi-build.c +3
-0+ .precise_baudrate = 0, + .namespace_string_length = sizeof(name), + .namespace_string_offset = 88,
What is this magic number 88 for? The spec says that it's the offset from the beginning of the structure to wherever the namespace string is put, but we don't decide that, the build_spcr() function does. The build_spcr() function knows what the offset of the name string is, so it should write this field; the callers shouldn't have to.
hw/arm/virt-acpi-build.c spcr_setup(GArray *table_data, BIOSLinker *linker, VirtMachineState *vms) · line 567
Quoted 1 line
hw/arm/virt-acpi-build.c +0
-0};
thanks
-- PMM
PATCH v2 3/3 tests: acpi: update expected blobs
In reply to Vadim ChichikalyukPrevious patch changed the SPCR ACPI table for AArch64 virt. Hexdump diff:
@@ -1 +1 @@
-00000000 53 50 43 52 50 00 00 00 02 b1 42 4f 43 48 53 20 +00000000 53 50 43 52 5a 00 00 00 04 78 42 4f 43 48 53 20 @@ -5,2 +5,3 @@ -00000040 ff ff ff ff 00 00 00 00 00 00 00 00 00 00 00 00 -00000050 +00000040 ff ff ff ff 00 00 00 00 00 00 00 00 00 36 6e 01 +00000050 00 00 00 00 02 00 58 00 2e 00
+0000005a
--- tests/data/acpi/aarch64/virt/SPCR | Bin 80 -> 90 bytes tests/qtest/bios-tables-test-allowed-diff.h | 1 - 2 files changed, 1 deletion(-)
tests/data/acpi/aarch64/virt/SPCR binary +0 -0
tests/qtest/bios-tables-test-allowed-diff.h +0 -1
@@ -1,2 +1 @@/* List of comma-separated changed AML files to ignore */ -"tests/data/acpi/aarch64/virt/SPCR",
Signature · 1 line
2.39.5 (Apple Git-154)
Quoted 3 lines · Vadim Chichikalyuk
This series upgrades the ACPI SPCR table for AArch64 virt to revision 4, adding the UART clock frequency, which was not previously available via ACPI.
So, what's the rationale here? QEMU's UART doesn't care about clocks at all. We provide a clock in the DTB because the kernel refuses to boot if you don't, but if the ACPI spec didn't even have a way to pass the clock frequency until rev 4 this obviously isn't a problem on that side.
thanks
-- PMM
Quoted 8 lines · Peter Maydell
So, what's the rationale here? QEMU's UART doesn't care about clocks at all. We provide a clock in the DTB because the kernel refuses to boot if you don't, but if the ACPI spec didn't even have a way to pass the clock frequency until rev 4 this obviously isn't a problem on that side.
The rationale is the same – just as the (Linux?) kernel expects a clock frequency if a UART is present in the DTB, guests using ACPI may expect a valid value for the clock frequency in the SPCR record when one is present.
Since we already have to have a notional value in the DTB, it makes sense to expose the same value via ACPI, too, for consistency. The RISC-V virt machine, for example, has it both in its DTB and ACPI DSDT.
I’ll fix the other patches per your comments if you think this is worth adding.
Thanks,
Vadim
Quoted 10 lines · Vadim Chichikalyuk
Quoted 9 lines · Peter Maydell
So, what's the rationale here? QEMU's UART doesn't care about clocks at all. We provide a clock in the DTB because the kernel refuses to boot if you don't, but if the ACPI spec didn't even have a way to pass the clock frequency until rev 4 this obviously isn't a problem on that side.
Quoted 4 lines
The rationale is the same – just as the (Linux?) kernel expects a clock frequency if a UART is present in the DTB, guests using ACPI may expect a valid value for the clock frequency in the SPCR record when one is present.
But the ACPI spec prior to rev 4 doesn't even provide a way to specify the clock frequency, so why would a guest expect it, and how could it do so?
thanks
-- PMM