save fpu state properly-ish, add a guard to doom_malloc_wrapper, fix QueryPerformance* and GetSystemTimeAsFileTime

code/kernel/ke_sched.c

@@ -53,6 +53,7 @@ void KeInitializeScheduler(void) {
     __stosq((unsigned __int64 *)ctx, 0, sizeof(KCONTEXT) / 8);
     ctx->Rip = (uint64_t)IdleLoop;
     idle->SavedContext = ctx;
+    KeInitThreadFpuContext(idle);
 
     gIdleThread    = idle;
     gCurrentThread = idle;

code/kernel/ke_switch.asm

@@ -9,6 +9,7 @@
 THREAD_SavedContext EQU 88
 THREAD_TssRsp0      EQU 104
 THREAD_Teb          EQU 112
+THREAD_FpuContext   EQU 160
 TSS_RSP0_OFFSET     EQU 4
 
 EXTERN gTss:QWORD          ; address of the single TSS, exported from ke_idt.c
@@ -20,6 +21,9 @@ EXTERN gTss:QWORD          ; address of the single TSS, exported from ke_idt.c
 ;   RCX = from,  RDX = to   (Microsoft x64)
 ; -------------------------------------------------------------------------
 KeSwitchContext PROC
+    mov     rax, [rcx + THREAD_FpuContext]
+    fxsave64 [rax]
+
     ; Save non-volatile registers onto the current (outgoing) kernel stack.
     push    rbx
     push    rbp
@@ -50,6 +54,9 @@ KeSwitchContext PROC
     mov     r8,  gTss
     mov     qword ptr [r8 + TSS_RSP0_OFFSET], rax
 
+    mov     rax, [rdx + THREAD_FpuContext]
+    fxrstor64 [rax]
+
     ; Restore non-volatile registers of the incoming thread.
     pop     r15
     pop     r14

code/kernel/ke_thread.c

@@ -7,6 +7,9 @@
 #include <mm_virt.h>
 #include <ob_object.h>
 
+#define FPU_FCW_DEFAULT   0x037Fu
+#define FPU_MXCSR_DEFAULT 0x1F80u
+
 static uint32_t gNextThreadId = 1;
 extern void KeThreadEntry(void);
 
@@ -80,8 +83,8 @@ __declspec(dllexport) KSTATUS KeCreateThread(PROCESS* owner, uintptr_t entry, ui
     // offset 0x10 is StackLimit, which __chkstk checks.
     uint64_t* teb = (uint64_t*)ustack;
     teb[0] = 0; // ExceptionList (GS:0x00)
-    teb[1] = 0; // StackBase     (GS:0x08)
-    teb[2] = 0; // StackLimit    (GS:0x10) - __chkstk will see 0 and safely exit
+    teb[1] = userStackTop;        // StackBase
+    teb[2] = (uintptr_t)ustack;   // StackLimit
     
     thread->Teb = teb;
 
@@ -124,6 +127,7 @@ __declspec(dllexport) KSTATUS KeCreateThread(PROCESS* owner, uintptr_t entry, ui
     owner->ThreadCount++;
     KeReleaseSpinlock(&owner->HandleLock, flags);
 
+    KeInitThreadFpuContext(thread);
     KeReadyThread(thread);
     *out = thread;
     return STATUS_SUCCESS;
@@ -211,6 +215,7 @@ __declspec(dllexport) KSTATUS KeCreateKernelThread(uintptr_t fn, uintptr_t arg,
 
     if (out) *out = thread;
 
+    KeInitThreadFpuContext(thread);
     KeReadyThread(thread);
     return STATUS_SUCCESS;
 }
@@ -233,3 +238,19 @@ void KeUnblockThread(THREAD* thread) {
     
     KeReadyThread(thread);
 }
+
+void KeInitThreadFpuContext(THREAD* thread)
+{
+    uintptr_t raw =
+        (uintptr_t)thread->FpuStorage;
+
+    uintptr_t aligned =
+        (raw + 15) & ~(uintptr_t)15;
+
+    thread->FpuContext = (void*)aligned;
+
+    memset(thread->FpuContext, 0, 512);
+
+    *(uint16_t*)((uint8_t*)thread->FpuContext + 0x00) = 0x037F;
+    *(uint32_t*)((uint8_t*)thread->FpuContext + 0x18) = 0x1F80;
+}

code/kernel/ke_thread.h

@@ -73,6 +73,9 @@ typedef struct THREAD {
     // Allocated in KeCreateThread, signaled in KeExitThread before teardown.
     // Waiters block on this via SYS_WAIT_THREAD -> KeWaitForEvent.
     EVENT*          ExitEvent;
+
+    void*   FpuContext;
+    uint8_t FpuStorage[528];
 } THREAD;
 
 __declspec(dllexport) KSTATUS KeCreateThread(PROCESS* owner, uintptr_t entry, uintptr_t arg,
@@ -94,3 +97,5 @@ void    KeUnblockThread(THREAD* thread);
 
 // Assembly stub - implemented in ke_switch.asm.
 void    KeSwitchContext(THREAD* from, THREAD* to);
+
+void    KeInitThreadFpuContext(THREAD* thread);