On 2025/9/23 下午12:22, Huacai Chen wrote:
Hi, Bibo,
On Tue, Sep 23, 2025 at 11:40 AM Bibo Mao <[email protected]> wrote:
Hi huacai,
It breaks with compatible issue since acpi table is changed, and test
case qtest-loongarch64/bios-tables-test fails to run.
LoongArch VM compatibility is not perfect now, one method is to modify
test case at the same time, another method is to add extra option in
order to support aligned GED ACPI address.
Does this issue must be fixed now? With ACPI spec, 5.2.12.20 Core
Programmable Interrupt Controller (CORE PIC) Structure, ACPI Processor
ID is not aligned also, its size is 4 byte and offset is 3 bytes.
They are different.
ACPI tables are probably packed (so the members may not be aligned),
such as CORE PIC you mentioned. Linux kernel defines two kinds of
accessors to parse members. You can see the code in
drivers/acpi/acpica/acmacros.h from the kernel as an example.
The problem mentioned in this patch is the alignment of a struct as a
whole. If the struct itself isn't aligned, Linux kernel cannot handle
it on hardware without UAL, even with two kinds of accessors.
If it must be fixed, the test case should be modified also together with
the patch. If not, it can be record as pending bug, will solve it if VM
compatibility method is decided.
Generally speaking, I think this should be fixed, because it is
allowed for qemu to emulate a machine without UAL. I will take a look
ok, loongarch VM compatibility is not good now, maybe we can skip it
now. There will be compatibility problem every time new feature and
hardware is added.
at qtest-loongarch64/bios-tables-test (maybe you means
tests/qtest/bios-tables-test.c?), but it seeams a separate patch is
better?
yeap, it is this test case, it fails with command "make check" or in
qemu CI. In general it should be in one patch set such as:
https://lists.nongnu.org/archive/html/qemu-riscv/2025-07/msg00155.html
Regards
Bibo Mao
Huacai
Regards
Bibo Mao
On 2025/9/22 下午10:15, Huacai Chen wrote:
From: Huacai Chen <[email protected]>
Now VIRT_GED_CPUHP_ADDR is not aligned to 4 bytes, but if Linux kernel
is built with ACPI_MISALIGNMENT_NOT_SUPPORTED, it assumes the alignment,
otherwise we get ACPI errors at boot phase:
ACPI Error: AE_AML_ALIGNMENT, Returned by Handler for [SystemMemory]
(20250404/evregion-301)
ACPI Error: Aborting method \_SB.CPUS.CSTA due to previous error
(AE_AML_ALIGNMENT) (20250404/psparse-529)
ACPI Error: Aborting method \_SB.CPUS.C000._STA due to previous error
(AE_AML_ALIGNMENT) (20250404/psparse-529)
ACPI Error: Method execution failed \_SB.CPUS.C000._STA due to previous error
(AE_AML_ALIGNMENT) (20250404/uteval-68)
VIRT_GED_MEM_ADDR and VIRT_GED_REG_ADDR are already aligned now, but use
QEMU_ALIGN_UP() to explicitly align them can make code more robust.
Reported-by: Nathan Chancellor <[email protected]>
Suggested-by: WANG Rui <[email protected]>
Signed-off-by: Huacai Chen <[email protected]>
---
include/hw/loongarch/virt.h | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/include/hw/loongarch/virt.h b/include/hw/loongarch/virt.h
index 602feab0f0..be4f5d603f 100644
--- a/include/hw/loongarch/virt.h
+++ b/include/hw/loongarch/virt.h
@@ -28,9 +28,9 @@
#define VIRT_LOWMEM_SIZE 0x10000000
#define VIRT_HIGHMEM_BASE 0x80000000
#define VIRT_GED_EVT_ADDR 0x100e0000
-#define VIRT_GED_MEM_ADDR (VIRT_GED_EVT_ADDR + ACPI_GED_EVT_SEL_LEN)
-#define VIRT_GED_REG_ADDR (VIRT_GED_MEM_ADDR + MEMORY_HOTPLUG_IO_LEN)
-#define VIRT_GED_CPUHP_ADDR (VIRT_GED_REG_ADDR + ACPI_GED_REG_COUNT)
+#define VIRT_GED_MEM_ADDR QEMU_ALIGN_UP(VIRT_GED_EVT_ADDR +
ACPI_GED_EVT_SEL_LEN, 4)
+#define VIRT_GED_REG_ADDR QEMU_ALIGN_UP(VIRT_GED_MEM_ADDR +
MEMORY_HOTPLUG_IO_LEN, 4)
+#define VIRT_GED_CPUHP_ADDR QEMU_ALIGN_UP(VIRT_GED_REG_ADDR +
ACPI_GED_REG_COUNT, 4)
#define COMMAND_LINE_SIZE 512
