CPU API improvements

Author: vit9696

Lilu/Headers/kern_cpu.hpp

@@ -400,8 +400,10 @@ namespace CPUInfo {
 	 *  @param b      ebx output pointer (optional)
 	 *  @param c      ecx output pointer (optional)
 	 *  @param d      edx output pointer (optional)
+	 *
+	 *  @return true if supported
 	 */
-	EXPORT void getCpuid(uint32_t no, uint32_t count, uint32_t *a, uint32_t *b=nullptr, uint32_t *c=nullptr, uint32_t *d=nullptr);
+	EXPORT bool getCpuid(uint32_t no, uint32_t count, uint32_t *a, uint32_t *b=nullptr, uint32_t *c=nullptr, uint32_t *d=nullptr);
 }
 
 #endif /* kern_cpu_h */

Lilu/Headers/kern_util.hpp

@@ -414,6 +414,21 @@ constexpr T getBitField(T so, T hi, T lo) {
 	return (so & getBitMask(hi, lo)) >> lo;
 }
 
+/**
+ *  Set bit field of size sizeof(T)
+ *  Warning, you are suggested to always pass the type explicitly!
+ *
+ *  @param va  value
+ *  @param hi  starting high bit
+ *  @param lo  ending low bit
+ *
+ *  @return bit field value
+ */
+template <typename T>
+constexpr T setBitField(T so, T hi, T lo) {
+	return (so << lo) & getBitMask(hi, lo);
+}
+
 /**
  *  This is an ugly replacement to std::find_if, allowing you
  *  to check whether a container consists only of value values.

Lilu/Sources/kern_cpu.cpp

@@ -18,18 +18,22 @@ static CPUInfo::CpuGeneration currentGeneration = CPUInfo::CpuGeneration::Unknow
 static uint32_t currentFamily = 0;
 static uint32_t currentModel = 0;
 static uint32_t currentStepping = 0;
+static uint32_t currentMaxLevel = 0;
+static uint32_t currentMaxLevelExt = 0x80000000;
 
 void CPUInfo::loadCpuInformation() {
 	// Start with detecting CPU vendor
 	uint32_t b = 0, c = 0, d = 0;
-	getCpuid(0, 0, nullptr, &b, &c, &d);
+	getCpuid(0, 0, &currentMaxLevel, &b, &c, &d);
 	if (b == signature_INTEL_ebx && c == signature_INTEL_ecx && d == signature_INTEL_edx)
 		currentVendor = CpuVendor::Intel;
 	else if (b == signature_AMD_ebx && c == signature_AMD_ecx && d == signature_AMD_edx)
 		currentVendor = CpuVendor::AMD;
 
-	// Only do extended model checking on Intel
-	if (currentVendor != CpuVendor::Intel)
+	getCpuid(0x80000000, 0, &currentMaxLevelExt);
+
+	// Only do extended model checking on Intel or when unsupported.
+	if (currentVendor != CpuVendor::Intel || currentMaxLevel < 1)
 		return;
 
 	// Detect CPU family and model
@@ -174,6 +178,7 @@ bool CPUInfo::getCpuTopology(CpuTopology &topology) {
 				topology.logicalCount[topology.packageCount]++;
 				lcpu = lcpu->next_in_core;
 			}
+			core = core->next_in_pkg;
 		}
 
 		topology.packageCount++;
@@ -183,16 +188,24 @@ bool CPUInfo::getCpuTopology(CpuTopology &topology) {
 	return true;
 }
 
-void CPUInfo::getCpuid(uint32_t no, uint32_t count, uint32_t *a, uint32_t *b, uint32_t *c, uint32_t *d) {
+bool CPUInfo::getCpuid(uint32_t no, uint32_t count, uint32_t *a, uint32_t *b, uint32_t *c, uint32_t *d) {
 	uint32_t eax = 0, ebx = 0, ecx = 0, edx = 0;
-	asm ("xchgq %%rbx, %q1\n"
-		 "cpuid\n"
-		 "xchgq %%rbx, %q1"
-		 : "=a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx)
-		 : "0" (no), "2" (count));
+
+	bool supported = (no & 0x80000000) ? currentMaxLevelExt >= no : currentMaxLevel >= no;
+
+	// At least pass zeroes on failure
+	if (supported) {
+		asm ("xchgq %%rbx, %q1\n"
+			 "cpuid\n"
+			 "xchgq %%rbx, %q1"
+			 : "=a" (eax), "=b" (ebx), "=c" (ecx), "=d" (edx)
+			 : "0" (no), "2" (count));
+	}
 
 	if (a) *a = eax;
 	if (b) *b = ebx;
 	if (c) *c = ecx;
 	if (d) *d = edx;
+
+	return supported;
 }

Lilu/Sources/kern_user.cpp

@@ -660,9 +660,10 @@ bool UserPatcher::loadDyldSharedCacheMapping() {
 	uint8_t *buffer {nullptr};
 	size_t bufferSize {0};
 	uint32_t ebx = 0;
-	CPUInfo::getCpuid(7, 0, nullptr, &ebx);
-	if ((ebx & CPUInfo::bit_AVX2) && getKernelVersion() >= KernelVersion::Yosemite)
-		buffer = FileIO::readFileToBuffer(SharedCacheMapHaswell, bufferSize);
+	if (CPUInfo::getCpuid(7, 0, nullptr, &ebx)) {
+		if ((ebx & CPUInfo::bit_AVX2) && getKernelVersion() >= KernelVersion::Yosemite)
+			buffer = FileIO::readFileToBuffer(SharedCacheMapHaswell, bufferSize);
+	}
 
 	if (!buffer)
 		buffer = FileIO::readFileToBuffer(SharedCacheMapLegacy, bufferSize);