W dniu 7.04.2023 o 15:02, Pedro Falcato pisze:
+EFI_STATUS
+EFIAPI
+UefiMain (
+ IN EFI_HANDLE ImageHandle,
+ IN EFI_SYSTEM_TABLE *SystemTable
+ )
+{
+ UINT64 aa64dfr0_el1 = read_aa64dfr0_el1 ();
+ UINT64 aa64dfr1_el1 = read_aa64dfr1_el1 ();
+ UINT64 aa64isar0_el1 = read_aa64isar0_el1 ();
+ UINT64 aa64isar1_el1 = read_aa64isar1_el1 ();
+ UINT64 aa64isar2_el1 = read_aa64isar2_el1 ();
+ UINT64 aa64mmfr0_el1 = read_aa64mmfr0_el1 ();
+ UINT64 aa64mmfr1_el1 = read_aa64mmfr1_el1 ();
+ UINT64 aa64mmfr2_el1 = read_aa64mmfr2_el1 ();
+ UINT64 aa64pfr0_el1 = read_aa64pfr0_el1 ();
+ UINT64 aa64pfr1_el1 = read_aa64pfr1_el1 ();
EDK2 requires separation between declarations and code (something
alike old C89 semantics, but stricter). so:
UINT64 aa64pfr1_el1;
<...>
aa64pfr1_el1 = read_aa64pfr1_el1 ();
done
+
+ /* UINT64 aa64smfr0_el1 = read_aa64smfr0_el1 ();*/
+ /* UINT64 aa64zfr0_el1 = read_aa64zfr0_el1 ();*/
Dead Code?
I ignore SVE/SME registers for now. Dropped code.
+
+ AsciiPrint ("ID_AA64MMFR0_EL1 = 0x%016lx\n", aa64mmfr0_el1);
+ AsciiPrint ("ID_AA64MMFR1_EL1 = 0x%016lx\n", aa64mmfr1_el1);
+ AsciiPrint ("ID_AA64MMFR2_EL1 = 0x%016lx\n", aa64mmfr2_el1);
+ AsciiPrint ("ID_AA64PFR0_EL1 = 0x%016lx\n", aa64pfr0_el1);
+ AsciiPrint ("ID_AA64PFR1_EL1 = 0x%016lx\n", aa64pfr1_el1);
+ AsciiPrint ("ID_AA64ISAR0_EL1 = 0x%016lx\n", aa64isar0_el1);
+ AsciiPrint ("ID_AA64ISAR1_EL1 = 0x%016lx\n", aa64isar1_el1);
+ AsciiPrint ("ID_AA64ISAR2_EL1 = 0x%016lx\n", aa64isar2_el1);
+ AsciiPrint ("ID_AA64DFR0_EL1 = 0x%016lx\n", aa64dfr0_el1);
+ AsciiPrint ("ID_AA64DFR1_EL1 = 0x%016lx\n", aa64dfr1_el1); //
ignore
Why ignore?
Was 0 on systems I checked. Dropped.
+ /* AsciiPrint ("ID_AA64SMFR0_EL1 = 0x%016lx\n", aa64smfr0_el1);*/
+ /* AsciiPrint ("ID_AA64ZFR0_EL1 = 0x%016lx\n", aa64zfr0_el1);*/
dead?
Dropped.
diff --git a/ArmPkg/Application/ArmCpuInfo/readregs.s
b/ArmPkg/Application/ArmCpuInfo/readregs.s
ASM files that require preprocessing should have a capital S here (.S vs .s)
After renaming it to readregs.S (and changing in ArmCpuInfo.inf) I get:
build.py...
/home/marcin/devel/linaro/sbsa-qemu/code/edk2/ArmPkg/Application/ArmCpuInfo/ArmCpuInfo.inf(29):
error 000E: File/directory not found in workspace
readregs.S is not found in packages path:
new file mode 100644
index 000000000000..052834b3c3f2
--- /dev/null
+++ b/ArmPkg/Application/ArmCpuInfo/readregs.s
@@ -0,0 +1,49 @@
+#include <AsmMacroIoLibV8.h>
+
+ASM_FUNC(read_aa64pfr0_el1)
+ mrs x0, ID_AA64PFR0_EL1;
+ ret;
ASM lines (for GAS at least) don't usually end in semicolons. so
+ASM_FUNC(read_aa64pfr0_el1)
+ mrs x0, ID_AA64PFR0_EL1
+ ret
You'd only need the semicolons if you had multiple instructions in one
line (like mrs x0, ID_AA...; ret)
thanks, dropped semicolons
+# ASM_FUNC(read_aa64zfr0_el1)
+# mrs x0, ID_AA64ZFR0_EL1;
+# ret;
+
+# ASM_FUNC(read_aa64smfr0_el1)
+# mrs x0, ID_AA64SMFR0_EL1;
+# ret;
Dead code?
Those two registers names were missing so got disabled. Dropped.
Brief comments in general:
1) You use binary literals extensively, which are not portable (GNU C
extension, AFAIK not in MSVC)
My C skills are rusty and the last time I used compiler other than GCC
was in Borland Turbo C 3 times.
2) Naming of identifiers (vars, functions, etc) needs to follow the
EDK2 style. eg:
description vs Description
read_aa64mmfr1_el1 vs ReadAa64MmfrEl1
etc...
ok
> Forgot mentioning: STATIC vs static, CONST vs const,
> VOID vs void, CHAR8 vs char, etc.
> All fun microsoftisms you need to use here.
Will be fun.
-=-=-=-=-=-=-=-=-=-=-=-
Groups.io Links: You receive all messages sent to this group.
View/Reply Online (#102699): https://edk2.groups.io/g/devel/message/102699
Mute This Topic: https://groups.io/mt/98123579/21656
Group Owner: devel+ow...@edk2.groups.io
Unsubscribe: https://edk2.groups.io/g/devel/unsub [arch...@mail-archive.com]
-=-=-=-=-=-=-=-=-=-=-=-
