diff --git a/reactos/ntoskrnl/include/internal/i386/ke.h b/reactos/ntoskrnl/include/internal/i386/ke.h
index 0dd6a866d73..7d0e521a54e 100644
--- a/reactos/ntoskrnl/include/internal/i386/ke.h
+++ b/reactos/ntoskrnl/include/internal/i386/ke.h
@@ -164,6 +164,9 @@ KiThreadStartup(PKSYSTEM_ROUTINE SystemRoutine,
                                      __d; \
                                  })
 #define _Ke386SetCr(N,X)         __asm__ __volatile__("movl %0,%%cr" #N : :"r" (X));
+#define Ke386SetTr(X)         __asm__ __volatile__("ltr %%ax" : :"a" (X));
+
+#define _Ke386SetSeg(N,X)         __asm__ __volatile__("movl %0,%%" #N : :"r" (X));
 
 #define Ke386GetCr0()            _Ke386GetCr(0)
 #define Ke386SetCr0(X)           _Ke386SetCr(0,X)
@@ -172,6 +175,9 @@ KiThreadStartup(PKSYSTEM_ROUTINE SystemRoutine,
 #define Ke386GetCr4()            _Ke386GetCr(4)
 #define Ke386SetCr4(X)           _Ke386SetCr(4,X)
 #define Ke386GetSs()             _Ke386GetSeg(ss)
+#define Ke386SetFs(X)            _Ke386SetSeg(fs, X)
+#define Ke386SetDs(X)            _Ke386SetSeg(ds, X)
+#define Ke386SetEs(X)            _Ke386SetSeg(es, X)
 
 static inline LONG Ke386TestAndClearBit(ULONG BitPos, volatile PULONG Addr)
 {
diff --git a/reactos/ntoskrnl/include/internal/ke.h b/reactos/ntoskrnl/include/internal/ke.h
index cab282b1c43..06d785692f4 100644
--- a/reactos/ntoskrnl/include/internal/ke.h
+++ b/reactos/ntoskrnl/include/internal/ke.h
@@ -121,6 +121,7 @@ extern ULONG KiMask32Array[MAXIMUM_PRIORITY];
 extern ULONG IdleProcessorMask;
 extern ULONG trap_stack_top;
 extern VOID KiTrap8(VOID);
+extern VOID KiTrap2(VOID);
 
 /* MACROS *************************************************************************/
 
diff --git a/reactos/ntoskrnl/ke/i386/cpu.c b/reactos/ntoskrnl/ke/i386/cpu.c
index cf14ca98a9c..daa2e252a90 100644
--- a/reactos/ntoskrnl/ke/i386/cpu.c
+++ b/reactos/ntoskrnl/ke/i386/cpu.c
@@ -26,6 +26,9 @@ KTSS KiBootTss;
 /* The TSS to use for Double Fault Traps (INT 0x9) */
 UCHAR KiDoubleFaultTSS[KTSS_IO_MAPS];
 
+/* The TSS to use for NMI Fault Traps (INT 0x2) */
+UCHAR KiNMITSS[KTSS_IO_MAPS];
+
 /* The Boot GDT (FIXME: should have more entries */
 KGDTENTRY KiBootGdt[12] =
 {
@@ -563,7 +566,7 @@ Ki386InitializeTss(VOID)
     TssEntry->HighWord.Bytes.BaseHi = (UCHAR)((ULONG_PTR)Tss >> 24);
 
     /* Load the task register */
-    __asm__("ltr %%ax":: "a" (KGDT_TSS));
+    Ke386SetTr(KGDT_TSS);
 
     /* Setup the Task Gate for Double Fault Traps */
     TaskGateEntry = &KiIdt[8];
@@ -596,6 +599,37 @@ Ki386InitializeTss(VOID)
     TssEntry->HighWord.Bytes.BaseMid = (UCHAR)((ULONG_PTR)Tss >> 16);
     TssEntry->HighWord.Bytes.BaseHi = (UCHAR)((ULONG_PTR)Tss >> 24);
     TssEntry->LimitLow = KTSS_IO_MAPS;
+
+    /* Now setup the NMI Task Gate */
+    TaskGateEntry = &KiIdt[2];
+    TaskGateAccess = (PKIDT_ACCESS)&TaskGateEntry->Access;
+#if 0
+    TaskGateAccess->SegmentType = I386_TASK_GATE;
+    TaskGateAccess->Present = 1;
+    TaskGateEntry->Selector = KGDT_NMI_TSS;
+#endif
+
+    /* Initialize the actual TSS */
+    Tss = (PKTSS)KiNMITSS;
+    KiInitializeTSS(Tss);
+    Tss->CR3 = _Ke386GetCr(3);
+    Tss->Esp0 = trap_stack_top;
+    Tss->Eip = PtrToUlong(KiTrap2);
+    Tss->Cs = KGDT_R0_CODE;
+    Tss->Fs = KGDT_R0_PCR;
+    Tss->Ss = Ke386GetSs();
+    Tss->Es = KGDT_R3_DATA | RPL_MASK;
+    Tss->Ds = KGDT_R3_DATA | RPL_MASK;
+
+    /* And its associated TSS Entry */
+    TssEntry = &KiBootGdt[KGDT_NMI_TSS / sizeof(KGDTENTRY)];
+    TssEntry->HighWord.Bits.Type = I386_TSS;
+    TssEntry->HighWord.Bits.Pres = 1;
+    TssEntry->HighWord.Bits.Dpl = 0;
+    TssEntry->BaseLow = (USHORT)((ULONG_PTR)Tss & 0xFFFF);
+    TssEntry->HighWord.Bytes.BaseMid = (UCHAR)((ULONG_PTR)Tss >> 16);
+    TssEntry->HighWord.Bytes.BaseHi = (UCHAR)((ULONG_PTR)Tss >> 24);
+    TssEntry->LimitLow = KTSS_IO_MAPS;
 }
 
 /* This is a rather naive implementation of Ke(Save/Restore)FloatingPointState
diff --git a/reactos/ntoskrnl/ke/i386/kernel.c b/reactos/ntoskrnl/ke/i386/kernel.c
index dd820c83369..07c69b8d4f7 100644
--- a/reactos/ntoskrnl/ke/i386/kernel.c
+++ b/reactos/ntoskrnl/ke/i386/kernel.c
@@ -435,9 +435,6 @@ KiInitializeKernel(IN PKPROCESS InitProcess,
     Prcb->NextThread = NULL;
     //Prcb->IdleThread = InitThread;
 
-    /* Initialize the Debugger */
-    KdInitSystem (0, &KeLoaderBlock);
-
     /* Initialize the Kernel Executive */
     ExpInitializeExecutive();
 
@@ -489,11 +486,26 @@ VOID
 NTAPI
 KiSystemStartup(IN PROS_LOADER_PARAMETER_BLOCK LoaderBlock)
 {
-    /* Currently hacked for CPU 0 only */
-    ULONG Cpu = 0;
+    ULONG Cpu;
     PKIPCR Pcr = (PKIPCR)KPCR_BASE;
     PKPRCB Prcb;
 
+    /* Save the loader block and get the current CPU */
+    //KeLoaderBlock = LoaderBlock;
+    Cpu = KeNumberProcessors;
+    if (!Cpu)
+    {
+        /* If this is the boot CPU, set FS and the CPU Number*/
+        Ke386SetFs(KGDT_R0_PCR);
+        KeGetPcr()->Number = Cpu;
+    }
+
+    /* Skip initial setup if this isn't the Boot CPU */
+    if (Cpu) goto AppCpuInit;
+
+    /* Setup the boot (Freeldr should've done), double fault and NMI TSS */
+    Ki386InitializeTss();
+
     /* Initialize the PCR */
     RtlZeroMemory(Pcr, PAGE_SIZE);
     KiInitializePcr(Cpu,
@@ -503,7 +515,6 @@ KiSystemStartup(IN PROS_LOADER_PARAMETER_BLOCK LoaderBlock)
                     &KiBootTss,
                     &KiInitialThread.Tcb,
                     trap_stack);
-    Prcb = Pcr->Prcb;
 
     /* Set us as the current process */
     KiInitialThread.Tcb.ApcState.Process = &KiInitialProcess.Pcb;
@@ -512,10 +523,13 @@ KiSystemStartup(IN PROS_LOADER_PARAMETER_BLOCK LoaderBlock)
     Pcr->PrcbData.ProcessorState.SpecialRegisters.KernelDr6 = 0;
     Pcr->PrcbData.ProcessorState.SpecialRegisters.KernelDr7 = 0;
 
-    /* Setup the boot (Freeldr should've done), double fault and NMI TSS */
-    Ki386InitializeTss();
+    /* Load Ring 3 selectors for DS/ES */
+    Ke386SetDs(KGDT_R3_DATA | RPL_MASK);
+    Ke386SetEs(KGDT_R3_DATA | RPL_MASK);
 
     /* Setup CPU-related fields */
+AppCpuInit:
+    Prcb = Pcr->Prcb;
     Pcr->Number = Cpu;
     Pcr->SetMember = 1 << Cpu;
     Pcr->SetMemberCopy = 1 << Cpu;
@@ -528,6 +542,12 @@ KiSystemStartup(IN PROS_LOADER_PARAMETER_BLOCK LoaderBlock)
     KeActiveProcessors |= Pcr->SetMember;
     KeNumberProcessors++;
 
+    /* Initialize the Debugger for the Boot CPU */
+    if (!Cpu) KdInitSystem (0, &KeLoaderBlock);
+
+    /* Check for break-in */
+    if (KdPollBreakIn()) DbgBreakPointWithStatus(1);
+
     /* Raise to HIGH_LEVEL */
     KfRaiseIrql(HIGH_LEVEL);
 
diff --git a/reactos/ntoskrnl/ke/i386/main_asm.S b/reactos/ntoskrnl/ke/i386/main_asm.S
index 22187d073c3..93e2834dbb5 100644
--- a/reactos/ntoskrnl/ke/i386/main_asm.S
+++ b/reactos/ntoskrnl/ke/i386/main_asm.S
@@ -36,19 +36,6 @@ _NtProcessStartup:
     lgdt _KiGdtDescriptor
     lidt _KiIdtDescriptor
 
-    /* Load the PCR selector */
-    movl $KGDT_R0_PCR, %eax
-    movl %eax, %fs
-
-    /* Load DS/ES (with override) */
-    .intel_syntax noprefix
-    mov eax, KGDT_R3_DATA + RPL_MASK
-    .byte 0x66
-    mov ds, ax
-    .byte 0x66
-    mov es, ax
-    .att_syntax
-
     /* Load the initial kernel stack */
     lea _kernel_stack_top, %eax
     sub $(SIZEOF_FX_SAVE_AREA), %eax
diff --git a/reactos/ntoskrnl/ke/i386/trap.s b/reactos/ntoskrnl/ke/i386/trap.s
index d312ed5438e..cb17cd448b4 100644
--- a/reactos/ntoskrnl/ke/i386/trap.s
+++ b/reactos/ntoskrnl/ke/i386/trap.s
@@ -826,6 +826,7 @@ V86Int1:
     int 3
 .endfunc
 
+.globl _KiTrap2
 .func KiTrap2
 _KiTrap2: