Support KFD interrupt protocol in second-level trap handler

If M0[23] is set then the driver will interpret the interrupt as a
debug event, rather than a signal event.

Clear M0 before sending the interrupt. All paths here are terminal so
it's not necessary to save/restore M0.

Change-Id: Ibd85b8cc6f8556941f2308a2c3fa3c68702cd606

runtime/hsa-runtime/core/inc/amd_gpu_shaders.h

@@ -202,6 +202,8 @@ static const unsigned int kCodeTrapHandler9[] = {
     s_waitcnt            lgkmcnt(0)
 
     // Send an interrupt to trigger event notification.
+    s_mov_b32            m0, 0x0
+    s_nop                0
     s_sendmsg            sendmsg(MSG_INTERRUPT)
 
   .signal_done:
@@ -226,15 +228,16 @@ static const unsigned int kCodeTrapHandler9[] = {
 */
     0xb8eef803, 0x8670ff6e, 0x10000100, 0xbf06ff70, 0x00000100, 0xbef000ff,
     0x20000000, 0xbf85000e, 0x8670ff6e, 0x00000800, 0xbef000f4, 0xbf85000a,
-    0x92eeff6d, 0x00080010, 0xbf84002a, 0xbf06826e, 0xbef000ff, 0x80000000,
-    0xbf850003, 0x806c846c, 0x826d806d, 0xbf820023, 0xbefe00ff, 0x80000000,
+    0x92eeff6d, 0x00080010, 0xbf84002c, 0xbf06826e, 0xbef000ff, 0x80000000,
+    0xbf850003, 0x806c846c, 0x826d806d, 0xbf820025, 0xbefe00ff, 0x80000000,
     0xbf90000a, 0xbf800007, 0xbf0c9f7e, 0xbf84fffd, 0x866eff7e, 0x000003ff,
     0x8e6e836e, 0xc0051bbd, 0x0000006e, 0xbf8cc07f, 0xc0071bb7, 0x000000c0,
     0xbf8cc07f, 0xbef10080, 0xc2831c37, 0x00000008, 0xbf8cc07f, 0x87707170,
-    0xbf85000c, 0xc0071c37, 0x00000010, 0xbf8cc07f, 0x86f07070, 0xbf840007,
+    0xbf85000e, 0xc0071c37, 0x00000010, 0xbf8cc07f, 0x86f07070, 0xbf840009,
     0xc0031bb7, 0x00000018, 0xbf8cc07f, 0xc0431bb8, 0x00000000, 0xbf8cc07f,
-    0xbf900001, 0x8778ff78, 0x00002000, 0x8f6e8b77, 0x866eff6e, 0x001f8000,
-    0xb96ef807, 0x86fe7e7e, 0x86ea6a6a, 0xb978f802, 0xbe801f6c,
+    0xbefc0080, 0xbf800000, 0xbf900001, 0x8778ff78, 0x00002000, 0x8f6e8b77,
+    0x866eff6e, 0x001f8000, 0xb96ef807, 0x86fe7e7e, 0x86ea6a6a, 0xb978f802,
+    0xbe801f6c,
 };
 
 static const unsigned int kCodeCopyAligned8[] = {