[kernel] Initial XSAVE support implementation

Initial support for XSAVE, but not XSAVEOPT or XSAVEC:

- Enable XSAVE and set up xcr0 for all CPUs
- Allocate XSAVE area for all non-kernel threads
- Call XSAVE and XRSTOR on task switch

src/kernel/objects/process.cpp

@@ -88,6 +88,9 @@ process::create_thread(uintptr_t rsp3, uint8_t priority)
     if (rsp3)
         th->tcb()->rsp3 = rsp3;
 
+    if (this != &g_kernel_process)
+        th->init_xsave_area();
+
     m_threads.append(th);
     scheduler::get().add_thread(th->tcb());
     return th;

src/kernel/objects/thread.cpp

@@ -8,6 +8,7 @@
 #include "objects/process.h"
 #include "objects/vm_area.h"
 #include "scheduler.h"
+#include "xsave.h"
 
 extern "C" void initialize_user_cpu();
 extern obj::vm_area_guarded &g_kernel_stacks;
@@ -37,6 +38,9 @@ thread::thread(process &parent, uint8_t pri, uintptr_t rsp0) :
 
 thread::~thread()
 {
+    if (m_tcb.xsave)
+        delete [] reinterpret_cast<uint8_t*>(m_tcb.xsave);
+
     g_kernel_stacks.return_section(m_tcb.kernel_stack);
     m_parent.handle_release();
 }
@@ -158,6 +162,14 @@ thread::add_thunk_user(uintptr_t rip3, uint64_t arg0, uint64_t arg1, uintptr_t r
     add_thunk_kernel(rip0 ? rip0 : trampoline);
 }
 
+void
+thread::init_xsave_area()
+{
+    void *xsave_area = new uint8_t [xsave_size];
+    memset(xsave_area, 0, xsave_size);
+    m_tcb.xsave = reinterpret_cast<uintptr_t>(xsave_area);
+}
+
 void
 thread::setup_kernel_stack()
 {

src/kernel/objects/thread.h

@@ -26,6 +26,7 @@ struct TCB
     uintptr_t rsp3;
     uintptr_t rflags3;
     uintptr_t pml4;
+    uintptr_t xsave;
     // End of area used by asembly
 
     obj::thread* thread;
@@ -182,6 +183,9 @@ private:
     /// \arg rsp0    The existing kernel stack rsp, 0 for none
     thread(process &parent, uint8_t pri, uintptr_t rsp0 = 0);
 
+    /// Set up the XSAVE saved processor state area for this thread
+    void init_xsave_area();
+
     /// Set up a new empty kernel stack for this thread.
     void setup_kernel_stack();