Palacios Public Git Repository

To checkout Palacios execute

  git clone http://v3vee.org/palacios/palacios.web/palacios.git
This will give you the master branch. You probably want the devel branch or one of the release branches. To switch to the devel branch, simply execute
  cd palacios
  git checkout --track -b devel origin/devel
The other branches are similar.


Removed a lot of the highly specific multicore code from the general core VMM code
[palacios.git] / palacios / src / devices / apic.c
index 7434b7d..0254d67 100644 (file)
 
 #include <devices/apic.h>
 #include <devices/apic_regs.h>
+#include <devices/icc_bus.h>
 #include <palacios/vmm.h>
 #include <palacios/vmm_msr.h>
+#include <palacios/vmm_sprintf.h>
+#include <palacios/vm_guest.h>
+
+
+#ifndef CONFIG_DEBUG_APIC
+#undef PrintDebug
+#define PrintDebug(fmt, args...)
+#endif
 
 
 typedef enum { APIC_TMR_INT, APIC_THERM_INT, APIC_PERF_INT, 
@@ -34,7 +43,7 @@ typedef enum { APIC_TMR_INT, APIC_THERM_INT, APIC_PERF_INT,
 #define APIC_EXTINT_DELIVERY 0x7
 
 
-#define BASE_ADDR_MSR 0x0000001B
+#define BASE_ADDR_MSR     0x0000001B
 #define DEFAULT_BASE_ADDR 0xfee00000
 
 #define APIC_ID_OFFSET                    0x020
@@ -109,807 +118,1126 @@ typedef enum { APIC_TMR_INT, APIC_THERM_INT, APIC_PERF_INT,
 
 
 
+
+
 struct apic_msr {
-  union {
-    uint64_t val;
-    struct {
-      uchar_t rsvd;
-      uint_t bootstrap_cpu : 1;
-      uint_t rsvd2         : 2;
-      uint_t apic_enable   : 1;
-      ullong_t base_addr   : 40;
-      uint_t rsvd3         : 12;
+    union {
+       uint64_t value;
+       struct {
+           uint8_t rsvd;
+           uint8_t bootstrap_cpu : 1;
+           uint8_t rsvd2         : 2;
+           uint8_t apic_enable   : 1;
+           uint64_t base_addr   : 40;
+           uint32_t rsvd3         : 12;
+       } __attribute__((packed));
     } __attribute__((packed));
-  } __attribute__((packed));
 } __attribute__((packed));
 
 
 
+typedef enum {INIT, SIPI, STARTED} ipi_state_t; 
 
 struct apic_state {
-  addr_t base_addr;
-
-  /* MSRs */
-  v3_msr_t base_addr_msr;
-
-
-  /* memory map registers */
-
-  struct lapic_id_reg lapic_id;
-  struct apic_ver_reg apic_ver;
-  struct ext_apic_ctrl_reg ext_apic_ctrl;
-  struct local_vec_tbl_reg local_vec_tbl;
-  struct tmr_vec_tbl_reg tmr_vec_tbl;
-  struct tmr_div_cfg_reg tmr_div_cfg;
-  struct lint_vec_tbl_reg lint0_vec_tbl;
-  struct lint_vec_tbl_reg lint1_vec_tbl;
-  struct perf_ctr_loc_vec_tbl_reg perf_ctr_loc_vec_tbl;
-  struct therm_loc_vec_tbl_reg therm_loc_vec_tbl;
-  struct err_vec_tbl_reg err_vec_tbl;
-  struct err_status_reg err_status;
-  struct spurious_int_reg spurious_int;
-  struct int_cmd_reg int_cmd;
-  struct log_dst_reg log_dst;
-  struct dst_fmt_reg dst_fmt;
-  struct arb_prio_reg arb_prio;
-  struct task_prio_reg task_prio;
-  struct proc_prio_reg proc_prio;
-  struct ext_apic_feature_reg ext_apic_feature;
-  struct spec_eoi_reg spec_eoi;
+    addr_t base_addr;
+
+    /* MSRs */
+    struct apic_msr base_addr_msr;
+
+
+    /* memory map registers */
+
+    struct lapic_id_reg lapic_id;
+    struct apic_ver_reg apic_ver;
+    struct ext_apic_ctrl_reg ext_apic_ctrl;
+    struct local_vec_tbl_reg local_vec_tbl;
+    struct tmr_vec_tbl_reg tmr_vec_tbl;
+    struct tmr_div_cfg_reg tmr_div_cfg;
+    struct lint_vec_tbl_reg lint0_vec_tbl;
+    struct lint_vec_tbl_reg lint1_vec_tbl;
+    struct perf_ctr_loc_vec_tbl_reg perf_ctr_loc_vec_tbl;
+    struct therm_loc_vec_tbl_reg therm_loc_vec_tbl;
+    struct err_vec_tbl_reg err_vec_tbl;
+    struct err_status_reg err_status;
+    struct spurious_int_reg spurious_int;
+    struct int_cmd_reg int_cmd;
+    struct log_dst_reg log_dst;
+    struct dst_fmt_reg dst_fmt;
+    struct arb_prio_reg arb_prio;
+    struct task_prio_reg task_prio;
+    struct proc_prio_reg proc_prio;
+    struct ext_apic_feature_reg ext_apic_feature;
+    struct spec_eoi_reg spec_eoi;
   
 
-  uint32_t tmr_cur_cnt;
-  uint32_t tmr_init_cnt;
+    uint32_t tmr_cur_cnt;
+    uint32_t tmr_init_cnt;
+
 
+    struct local_vec_tbl_reg ext_intr_vec_tbl[4];
 
+    uint32_t rem_rd_data;
 
-  uint32_t rem_rd_data;
 
+    ipi_state_t ipi_state;
 
-  uchar_t int_req_reg[32];
-  uchar_t int_svc_reg[32];
-  uchar_t int_en_reg[32];
-  uchar_t trig_mode_reg[32];
+    uint8_t int_req_reg[32];
+    uint8_t int_svc_reg[32];
+    uint8_t int_en_reg[32];
+    uint8_t trig_mode_reg[32];
   
-  uint32_t eoi;
+    uint32_t eoi;
 
+    struct vm_device * icc_bus;
 
+    v3_lock_t  lock;
 };
 
 
-static void init_apic_state(struct apic_state * apic) {
-  apic->base_addr = DEFAULT_BASE_ADDR;
-  apic->base_addr_msr.value = 0x0000000000000900LL;
-  apic->base_addr_msr.value |= ((uint64_t)DEFAULT_BASE_ADDR); 
-
-  PrintDebug("Sizeof Interrupt Request Register %d, should be 32\n", 
-            (uint_t)sizeof(apic->int_req_reg));
-
-  memset(apic->int_req_reg, 0, sizeof(apic->int_req_reg));
-  memset(apic->int_svc_reg, 0, sizeof(apic->int_svc_reg));
-  memset(apic->int_en_reg, 0xff, sizeof(apic->int_en_reg));
-  memset(apic->trig_mode_reg, 0, sizeof(apic->trig_mode_reg));
-
-  apic->eoi = 0x00000000;
-  apic->rem_rd_data = 0x00000000;
-  apic->tmr_init_cnt = 0x00000000;
-  apic->tmr_cur_cnt = 0x00000000;
-
-  // TODO:
-  // We need to figure out what the APIC ID is....
-  apic->lapic_id.val = 0x00000000;
-
-  // The P6 has 6 LVT entries, so we set the value to (6-1)...
-  apic->apic_ver.val = 0x80050010;
-
-  apic->task_prio.val = 0x00000000;
-  apic->arb_prio.val = 0x00000000;
-  apic->proc_prio.val = 0x00000000;
-  apic->log_dst.val = 0x00000000;
-  apic->dst_fmt.val = 0xffffffff;
-  apic->spurious_int.val = 0x000000ff;
-  apic->err_status.val = 0x00000000;
-  apic->int_cmd.val = 0x0000000000000000LL;
-  apic->tmr_vec_tbl.val = 0x00010000;
-  apic->therm_loc_vec_tbl.val = 0x00010000;
-  apic->perf_ctr_loc_vec_tbl.val = 0x00010000;
-  apic->lint0_vec_tbl.val = 0x00010000;
-  apic->lint1_vec_tbl.val = 0x00010000;
-  apic->err_vec_tbl.val = 0x00010000;
-  apic->tmr_div_cfg.val = 0x00000000;
-  apic->ext_apic_feature.val = 0x00040007;
-  apic->ext_apic_ctrl.val = 0x00000000;
-  apic->spec_eoi.val = 0x00000000;
+
+
+
+static int apic_read(struct guest_info * core, addr_t guest_addr, void * dst, uint_t length, void * priv_data);
+static int apic_write(struct guest_info * core, addr_t guest_addr, void * src, uint_t length, void * priv_data);
+
+static void init_apic_state(struct apic_state * apic, uint32_t id, struct vm_device * icc) {
+    apic->base_addr = DEFAULT_BASE_ADDR;
+
+    if (id == 0) { 
+       // boot processor, enabled
+       apic->base_addr_msr.value = 0x0000000000000900LL;
+    } else {
+       // ap processor, enabled
+       apic->base_addr_msr.value = 0x0000000000000800LL;
+    }
+
+    // same base address regardless of ap or main
+    apic->base_addr_msr.value |= ((uint64_t)DEFAULT_BASE_ADDR); 
+
+    PrintDebug("apic %u: (init_apic_state): msr=0x%llx\n",id, apic->base_addr_msr.value);
+
+    PrintDebug("apic %u: (init_apic_state): Sizeof Interrupt Request Register %d, should be 32\n",
+              id, (uint_t)sizeof(apic->int_req_reg));
+
+    memset(apic->int_req_reg, 0, sizeof(apic->int_req_reg));
+    memset(apic->int_svc_reg, 0, sizeof(apic->int_svc_reg));
+    memset(apic->int_en_reg, 0xff, sizeof(apic->int_en_reg));
+    memset(apic->trig_mode_reg, 0, sizeof(apic->trig_mode_reg));
+
+    apic->eoi = 0x00000000;
+    apic->rem_rd_data = 0x00000000;
+    apic->tmr_init_cnt = 0x00000000;
+    apic->tmr_cur_cnt = 0x00000000;
+
+    apic->lapic_id.val = id;
+    
+    apic->icc_bus = icc;
+
+    apic->ipi_state = INIT;
+
+    // The P6 has 6 LVT entries, so we set the value to (6-1)...
+    apic->apic_ver.val = 0x80050010;
+
+    apic->task_prio.val = 0x00000000;
+    apic->arb_prio.val = 0x00000000;
+    apic->proc_prio.val = 0x00000000;
+    apic->log_dst.val = 0x00000000;
+    apic->dst_fmt.val = 0xffffffff;
+    apic->spurious_int.val = 0x000000ff;
+    apic->err_status.val = 0x00000000;
+    apic->int_cmd.val = 0x0000000000000000LL;
+    apic->tmr_vec_tbl.val = 0x00010000;
+    apic->therm_loc_vec_tbl.val = 0x00010000;
+    apic->perf_ctr_loc_vec_tbl.val = 0x00010000;
+    apic->lint0_vec_tbl.val = 0x00010000;
+    apic->lint1_vec_tbl.val = 0x00010000;
+    apic->err_vec_tbl.val = 0x00010000;
+    apic->tmr_div_cfg.val = 0x00000000;
+    //apic->ext_apic_feature.val = 0x00000007;
+    apic->ext_apic_feature.val = 0x00040007;
+    apic->ext_apic_ctrl.val = 0x00000000;
+    apic->spec_eoi.val = 0x00000000;
+
+    v3_lock_init(&(apic->lock));
 }
 
 
 
 
-static int read_apic_msr(uint_t msr, v3_msr_t * dst, void * priv_data) {
-  struct vm_device * dev = (struct vm_device *)priv_data;
-  struct apic_state * apic = (struct apic_state *)dev->private_data;
-  PrintDebug("READING APIC BASE ADDR: HI=%x LO=%x\n", apic->base_addr_msr.hi, apic->base_addr_msr.lo);
+static int read_apic_msr(struct guest_info * core, uint_t msr, v3_msr_t * dst, void * priv_data) {
+    struct vm_device * dev = (struct vm_device *)priv_data;
+    struct apic_state * apics = (struct apic_state *)(dev->private_data);
+    struct apic_state * apic = &(apics[core->cpu_id]);
 
-  return -1;
+    PrintDebug("apic %u: core %u: MSR read\n", apic->lapic_id.val, core->cpu_id);
+    v3_lock(apic->lock);
+    dst->value = apic->base_addr;
+    v3_unlock(apic->lock);
+    return 0;
 }
 
 
-static int write_apic_msr(uint_t msr, v3_msr_t src, void * priv_data) {
-  //  struct vm_device * dev = (struct vm_device *)priv_data;
-  //  struct apic_state * apic = (struct apic_state *)dev->private_data;
+static int write_apic_msr(struct guest_info * core, uint_t msr, v3_msr_t src, void * priv_data) {
+    struct vm_device * dev = (struct vm_device *)priv_data;
+    struct apic_state * apics = (struct apic_state *)(dev->private_data);
+    struct apic_state * apic = &(apics[core->cpu_id]);
+    struct v3_mem_region * old_reg = v3_get_mem_region(dev->vm, core->cpu_id, apic->base_addr);
+
+
+    PrintDebug("apic %u: core %u: MSR write\n",apic->lapic_id.val,core->cpu_id);
 
-  PrintDebug("WRITING APIC BASE ADDR: HI=%x LO=%x\n", src.hi, src.lo);
+    if (old_reg == NULL) {
+       // uh oh...
+       PrintError("apic %u: core %u: APIC Base address region does not exit...\n",
+                  apic->lapic_id.val, core->cpu_id);
+       return -1;
+    }
+    
+    v3_lock(apic->lock);
+
+    v3_delete_mem_region(dev->vm, old_reg);
+
+    apic->base_addr = src.value;
+
+    if (v3_hook_full_mem(dev->vm, core->cpu_id, apic->base_addr, apic->base_addr + PAGE_SIZE_4KB, apic_read, apic_write, dev) == -1) {
+       PrintError("apic %u: core %u: Could not hook new APIC Base address\n",
+                  apic->lapic_id.val, core->cpu_id);
+       v3_unlock(apic->lock);
+       return -1;
+    }
 
-  return -1;
+    v3_unlock(apic->lock);
+    return 0;
 }
 
 
 // irq_num is the bit offset into a 256 bit buffer...
 static int activate_apic_irq(struct apic_state * apic, uint32_t irq_num) {
-  int major_offset = irq_num & ~0x00000007;
-  int minor_offset = irq_num & 0x00000007;
-  uchar_t * req_location = apic->int_req_reg + major_offset;
-  uchar_t * en_location = apic->int_en_reg + major_offset;
-  uchar_t flag = 0x1 << minor_offset;
-
-  if (irq_num <= 15) {
-    PrintError("Attempting to raise an invalid interrupt: %d\n", irq_num);
-    return -1;
-  }
+    int major_offset = (irq_num & ~0x00000007) >> 3;
+    int minor_offset = irq_num & 0x00000007;
+    uint8_t * req_location = apic->int_req_reg + major_offset;
+    uint8_t * en_location = apic->int_en_reg + major_offset;
+    uint8_t flag = 0x1 << minor_offset;
 
-  if (*en_location & flag) {
-    *req_location |= flag;
-  } else {
-    return 0;
-  }
 
-  return 0;
-}
 
+    if (irq_num <= 15) {
+//     PrintError("apic %u: core ?: Attempting to raise an invalid interrupt: %d\n", apic->lapic_id.val,irq_num);
+       return -1;
+    }
 
-static int activate_internal_irq(struct apic_state * apic, apic_irq_type_t int_type) {
-  uint32_t vec_num = 0;
-  uint32_t del_mode = 0;
-  int masked = 0;
-
-
-  switch (int_type) {
-  case APIC_TMR_INT:
-    vec_num = apic->tmr_vec_tbl.vec;
-    del_mode = APIC_FIXED_DELIVERY;
-    masked = apic->tmr_vec_tbl.mask;
-    break;
-  case APIC_THERM_INT:
-    vec_num = apic->therm_loc_vec_tbl.vec;
-    del_mode = apic->therm_loc_vec_tbl.msg_type;
-    masked = apic->therm_loc_vec_tbl.mask;
-    break;
-  case APIC_PERF_INT:
-    vec_num = apic->perf_ctr_loc_vec_tbl.vec;
-    del_mode = apic->perf_ctr_loc_vec_tbl.msg_type;
-    masked = apic->perf_ctr_loc_vec_tbl.mask;
-    break;
-  case APIC_LINT0_INT:
-    vec_num = apic->lint0_vec_tbl.vec;
-    del_mode = apic->lint0_vec_tbl.msg_type;
-    masked = apic->lint0_vec_tbl.mask;
-    break;
-  case APIC_LINT1_INT:
-    vec_num = apic->lint1_vec_tbl.vec;
-    del_mode = apic->lint1_vec_tbl.msg_type;
-    masked = apic->lint1_vec_tbl.mask;
-    break;
-  case APIC_ERR_INT:
-    vec_num = apic->err_vec_tbl.vec;
-    del_mode = APIC_FIXED_DELIVERY;
-    masked = apic->err_vec_tbl.mask;
-    break;
-  default:
-    PrintError("Invalid APIC interrupt type\n");
-    return -1;
-  }
 
-  // interrupt is masked, don't send
-  if (masked == 1) {
+    PrintDebug("apic %u: core ?: Raising APIC IRQ %d\n", apic->lapic_id.val, irq_num);
+
+    if (*req_location & flag) {
+       //V3_Print("Interrupts coallescing\n");
+    }
+
+    if (*en_location & flag) {
+       *req_location |= flag;
+    } else {
+       PrintDebug("apic %u: core ?: Interrupt  not enabled... %.2x\n", 
+                  apic->lapic_id.val, *en_location);
+       return 0;
+    }
+
     return 0;
-  }
+}
+
+
+
+static int get_highest_isr(struct apic_state * apic) {
+    int i = 0, j = 0;
+
+    // We iterate backwards to find the highest priority
+    for (i = 31; i >= 0; i--) {
+       uint8_t  * svc_major = apic->int_svc_reg + i;
+    
+       if ((*svc_major) & 0xff) {
+           for (j = 7; j >= 0; j--) {
+               uint8_t flag = 0x1 << j;
+               if ((*svc_major) & flag) {
+                   return ((i * 8) + j);
+               }
+           }
+       }
+    }
 
-  if (del_mode == APIC_FIXED_DELIVERY) {
-    PrintDebug("Activating internal APIC IRQ\n");
-    return activate_apic_irq(apic, vec_num);
-  } else {
-    PrintError("Unhandled Delivery Mode\n");
     return -1;
-  }
 }
 
 
-static int apic_read(addr_t guest_addr, void * dst, uint_t length, void * priv_data) {
-  struct vm_device * dev = (struct vm_device *)priv_data;
-  struct apic_state * apic = (struct apic_state *)dev->private_data;
-  addr_t reg_addr  = guest_addr - apic->base_addr;
-  struct apic_msr * msr = (struct apic_msr *)&(apic->base_addr_msr.value);
-  uint32_t * val_ptr = (uint32_t *)dst;
+static int get_highest_irr(struct apic_state * apic) {
+    int i = 0, j = 0;
 
-  PrintDebug("Read apic address space (%p)\n", 
-            (void *)guest_addr);
+    // We iterate backwards to find the highest priority
+    for (i = 31; i >= 0; i--) {
+       uint8_t  * req_major = apic->int_req_reg + i;
+    
+       if ((*req_major) & 0xff) {
+           for (j = 7; j >= 0; j--) {
+               uint8_t flag = 0x1 << j;
+               if ((*req_major) & flag) {
+                   return ((i * 8) + j);
+               }
+           }
+       }
+    }
 
-  if (msr->apic_enable == 0) {
-    PrintError("Write to APIC address space with disabled APIC\n");
     return -1;
-  }
+}
 
 
-  if (length != 4) {
-    PrintError("Invalid apic readlength\n");
-    return -1;
-  }
 
-  switch (reg_addr) {
-  case EOI_OFFSET:
-    PrintError("Attempting to read from write only register\n");
-    return -1;
-    break;
-
-    // data registers
-  case APIC_ID_OFFSET:
-    *val_ptr = apic->lapic_id.val;
-    break;
-  case APIC_VERSION_OFFSET:
-    *val_ptr = apic->apic_ver.val;
-    break;
-  case TPR_OFFSET:
-    *val_ptr = apic->task_prio.val;
-    break;
-  case APR_OFFSET:
-    *val_ptr = apic->arb_prio.val;
-    break;
-  case PPR_OFFSET:
-    *val_ptr = apic->proc_prio.val;
-    break;
-  case REMOTE_READ_OFFSET:
-    *val_ptr = apic->rem_rd_data;
-    break;
-  case LDR_OFFSET:
-    *val_ptr = apic->log_dst.val;
-    break;
-  case DFR_OFFSET:
-    *val_ptr = apic->dst_fmt.val;
-    break;
-  case SPURIOUS_INT_VEC_OFFSET:
-    *val_ptr = apic->spurious_int.val;
-    break;
-  case ESR_OFFSET:
-    *val_ptr = apic->err_status.val;
-    break;
-  case TMR_LOC_VEC_TBL_OFFSET:
-    *val_ptr = apic->tmr_vec_tbl.val;
-    break;
-  case LINT0_VEC_TBL_OFFSET:
-    *val_ptr = apic->lint0_vec_tbl.val;
-    break;
-  case LINT1_VEC_TBL_OFFSET:
-    *val_ptr = apic->lint1_vec_tbl.val;
-    break;
-  case ERR_VEC_TBL_OFFSET:
-    *val_ptr = apic->err_vec_tbl.val;
-    break;
-  case TMR_INIT_CNT_OFFSET:
-    *val_ptr = apic->tmr_init_cnt;
-    break;
-  case TMR_DIV_CFG_OFFSET:
-    *val_ptr = apic->tmr_div_cfg.val;
-    break;
-
-  case IER_OFFSET0:
-    *val_ptr = *(uint32_t *)(apic->int_en_reg);
-    break;
-  case IER_OFFSET1:
-    *val_ptr = *(uint32_t *)(apic->int_en_reg + 4);
-    break;
-  case IER_OFFSET2:
-    *val_ptr = *(uint32_t *)(apic->int_en_reg + 8);
-    break;
-  case IER_OFFSET3:
-    *val_ptr = *(uint32_t *)(apic->int_en_reg + 12);
-    break;
-  case IER_OFFSET4:
-    *val_ptr = *(uint32_t *)(apic->int_en_reg + 16);
-    break;
-  case IER_OFFSET5:
-    *val_ptr = *(uint32_t *)(apic->int_en_reg + 20);
-    break;
-  case IER_OFFSET6:
-    *val_ptr = *(uint32_t *)(apic->int_en_reg + 24);
-    break;
-  case IER_OFFSET7:
-    *val_ptr = *(uint32_t *)(apic->int_en_reg + 28);
-    break;
-
-  case ISR_OFFSET0:
-    *val_ptr = *(uint32_t *)(apic->int_svc_reg);
-    break;
-  case ISR_OFFSET1:
-    *val_ptr = *(uint32_t *)(apic->int_svc_reg + 4);
-    break;
-  case ISR_OFFSET2:
-    *val_ptr = *(uint32_t *)(apic->int_svc_reg + 8);
-    break;
-  case ISR_OFFSET3:
-    *val_ptr = *(uint32_t *)(apic->int_svc_reg + 12);
-    break;
-  case ISR_OFFSET4:
-    *val_ptr = *(uint32_t *)(apic->int_svc_reg + 16);
-    break;
-  case ISR_OFFSET5:
-    *val_ptr = *(uint32_t *)(apic->int_svc_reg + 20);
-    break;
-  case ISR_OFFSET6:
-    *val_ptr = *(uint32_t *)(apic->int_svc_reg + 24);
-    break;
-  case ISR_OFFSET7:
-    *val_ptr = *(uint32_t *)(apic->int_svc_reg + 28);
-    break;
-   
-  case TRIG_OFFSET0:
-    *val_ptr = *(uint32_t *)(apic->trig_mode_reg);
-    break;
-  case TRIG_OFFSET1:
-    *val_ptr = *(uint32_t *)(apic->trig_mode_reg + 4);
-    break;
-  case TRIG_OFFSET2:
-    *val_ptr = *(uint32_t *)(apic->trig_mode_reg + 8);
-    break;
-  case TRIG_OFFSET3:
-    *val_ptr = *(uint32_t *)(apic->trig_mode_reg + 12);
-    break;
-  case TRIG_OFFSET4:
-    *val_ptr = *(uint32_t *)(apic->trig_mode_reg + 16);
-    break;
-  case TRIG_OFFSET5:
-    *val_ptr = *(uint32_t *)(apic->trig_mode_reg + 20);
-    break;
-  case TRIG_OFFSET6:
-    *val_ptr = *(uint32_t *)(apic->trig_mode_reg + 24);
-    break;
-  case TRIG_OFFSET7:
-    *val_ptr = *(uint32_t *)(apic->trig_mode_reg + 28);
-    break;
-
-  case IRR_OFFSET0:
-    *val_ptr = *(uint32_t *)(apic->int_req_reg);
-    break;
-  case IRR_OFFSET1:
-    *val_ptr = *(uint32_t *)(apic->int_req_reg + 4);
-    break;
-  case IRR_OFFSET2:
-    *val_ptr = *(uint32_t *)(apic->int_req_reg + 8);
-    break;
-  case IRR_OFFSET3:
-    *val_ptr = *(uint32_t *)(apic->int_req_reg + 12);
-    break;
-  case IRR_OFFSET4:
-    *val_ptr = *(uint32_t *)(apic->int_req_reg + 16);
-    break;
-  case IRR_OFFSET5:
-    *val_ptr = *(uint32_t *)(apic->int_req_reg + 20);
-    break;
-  case IRR_OFFSET6:
-    *val_ptr = *(uint32_t *)(apic->int_req_reg + 24);
-    break;
-  case IRR_OFFSET7:
-    *val_ptr = *(uint32_t *)(apic->int_req_reg + 28);
-    break;
-  case TMR_CUR_CNT_OFFSET:
-    *val_ptr = apic->tmr_cur_cnt;
-    break;
-
-    // We are not going to implement these....
-  case THERM_LOC_VEC_TBL_OFFSET:
-    *val_ptr = apic->therm_loc_vec_tbl.val;
-    break;
-  case PERF_CTR_LOC_VEC_TBL_OFFSET:
-    *val_ptr = apic->perf_ctr_loc_vec_tbl.val;
-    break;
+static int apic_do_eoi(struct apic_state * apic) {
+    int isr_irq = get_highest_isr(apic);
 
+    if (isr_irq != -1) {
+       int major_offset = (isr_irq & ~0x00000007) >> 3;
+       int minor_offset = isr_irq & 0x00000007;
+       uint8_t flag = 0x1 << minor_offset;
+       uint8_t * svc_location = apic->int_svc_reg + major_offset;
+       
+       PrintDebug("apic %u: core ?: Received APIC EOI for IRQ %d\n", apic->lapic_id.val,isr_irq);
+       
+       *svc_location &= ~flag;
+
+#ifdef CONFIG_CRAY_XT
+       
+       if ((isr_irq == 238) || 
+           (isr_irq == 239)) {
+           PrintDebug("apic %u: core ?: Acking IRQ %d\n", apic->lapic_id.val,isr_irq);
+       }
+       
+       if (isr_irq == 238) {
+           V3_ACK_IRQ(238);
+       }
+#endif
+    } else {
+       //PrintError("apic %u: core ?: Spurious EOI...\n",apic->lapic_id.val);
+    }
+       
+    return 0;
+}
  
 
-    // handled registers
-  case INT_CMD_LO_OFFSET:    
-    *val_ptr = apic->int_cmd.lo;
-    break;
-  case INT_CMD_HI_OFFSET:
-    *val_ptr = apic->int_cmd.hi;
-    break;
-
-    // handle current timer count
-
-    // Unhandled Registers
-  case EXT_INT_LOC_VEC_TBL_OFFSET0:
-  case EXT_INT_LOC_VEC_TBL_OFFSET1:
-  case EXT_INT_LOC_VEC_TBL_OFFSET2:
-  case EXT_INT_LOC_VEC_TBL_OFFSET3:
-  case EXT_APIC_FEATURE_OFFSET:
-  case EXT_APIC_CMD_OFFSET:
-  case SEOI_OFFSET:
-
-  default:
-    PrintError("Read from Unhandled APIC Register: %x\n", (uint32_t)reg_addr);
-    return -1;
-  }
+static int activate_internal_irq(struct apic_state * apic, apic_irq_type_t int_type) {
+    uint32_t vec_num = 0;
+    uint32_t del_mode = 0;
+    int masked = 0;
+
+
+    switch (int_type) {
+       case APIC_TMR_INT:
+           vec_num = apic->tmr_vec_tbl.vec;
+           del_mode = APIC_FIXED_DELIVERY;
+           masked = apic->tmr_vec_tbl.mask;
+           break;
+       case APIC_THERM_INT:
+           vec_num = apic->therm_loc_vec_tbl.vec;
+           del_mode = apic->therm_loc_vec_tbl.msg_type;
+           masked = apic->therm_loc_vec_tbl.mask;
+           break;
+       case APIC_PERF_INT:
+           vec_num = apic->perf_ctr_loc_vec_tbl.vec;
+           del_mode = apic->perf_ctr_loc_vec_tbl.msg_type;
+           masked = apic->perf_ctr_loc_vec_tbl.mask;
+           break;
+       case APIC_LINT0_INT:
+           vec_num = apic->lint0_vec_tbl.vec;
+           del_mode = apic->lint0_vec_tbl.msg_type;
+           masked = apic->lint0_vec_tbl.mask;
+           break;
+       case APIC_LINT1_INT:
+           vec_num = apic->lint1_vec_tbl.vec;
+           del_mode = apic->lint1_vec_tbl.msg_type;
+           masked = apic->lint1_vec_tbl.mask;
+           break;
+       case APIC_ERR_INT:
+           vec_num = apic->err_vec_tbl.vec;
+           del_mode = APIC_FIXED_DELIVERY;
+           masked = apic->err_vec_tbl.mask;
+           break;
+       default:
+           PrintError("apic %u: core ?: Invalid APIC interrupt type\n", apic->lapic_id.val);
+           return -1;
+    }
 
-  PrintDebug("Read finished (val=%x)\n", *(uint32_t *)dst);
+    // interrupt is masked, don't send
+    if (masked == 1) {
+       PrintDebug("apic %u: core ?: Inerrupt is masked\n", apic->lapic_id.val);
+       return 0;
+    }
 
-  return length;
+    if (del_mode == APIC_FIXED_DELIVERY) {
+       //PrintDebug("Activating internal APIC IRQ %d\n", vec_num);
+       return activate_apic_irq(apic, vec_num);
+    } else {
+       PrintError("apic %u: core ?: Unhandled Delivery Mode\n", apic->lapic_id.val);
+       return -1;
+    }
 }
 
 
-static int apic_write(addr_t guest_addr, void * src, uint_t length, void * priv_data) {
-  struct vm_device * dev = (struct vm_device *)priv_data;
-  struct apic_state * apic = (struct apic_state *)dev->private_data;
-  addr_t reg_addr  = guest_addr - apic->base_addr;
-  struct apic_msr * msr = (struct apic_msr *)&(apic->base_addr_msr.value);
-  uint32_t op_val = *(uint32_t *)src;
+static int apic_read(struct guest_info * core, addr_t guest_addr, void * dst, uint_t length, void * priv_data) {
+    struct apic_state * apic = (struct apic_state *)(priv_data);
+    addr_t reg_addr  = guest_addr - apic->base_addr;
+    struct apic_msr * msr = (struct apic_msr *)&(apic->base_addr_msr.value);
+    uint32_t val = 0;
 
-  PrintDebug("Write to apic address space (%p) (val=%x)\n", 
-            (void *)guest_addr, *(uint32_t *)src);
 
-  if (msr->apic_enable == 0) {
-    PrintError("Write to APIC address space with disabled APIC\n");
-    return -1;
-  }
+    PrintDebug("apic %u: core %u: at %p: Read apic address space (%p)\n",
+              apic->lapic_id.val, core->cpu_id, apic, (void *)guest_addr);
 
+    if (msr->apic_enable == 0) {
+       PrintError("apic %u: core %u: Read from APIC address space with disabled APIC, apic msr=0x%llx\n",
+                  apic->lapic_id.val, core->cpu_id, apic->base_addr_msr.value);
 
-  if (length != 4) {
-    PrintError("Invalid apic write length\n");
-    return -1;
-  }
+       return -1;
+    }
 
-  switch (reg_addr) {
-  case REMOTE_READ_OFFSET:
-  case APIC_VERSION_OFFSET:
-  case APR_OFFSET:
-  case IRR_OFFSET0:
-  case IRR_OFFSET1:
-  case IRR_OFFSET2:
-  case IRR_OFFSET3:
-  case IRR_OFFSET4:
-  case IRR_OFFSET5:
-  case IRR_OFFSET6:
-  case IRR_OFFSET7:
-  case ISR_OFFSET0:
-  case ISR_OFFSET1:
-  case ISR_OFFSET2:
-  case ISR_OFFSET3:
-  case ISR_OFFSET4:
-  case ISR_OFFSET5:
-  case ISR_OFFSET6:
-  case ISR_OFFSET7:
-  case TRIG_OFFSET0:
-  case TRIG_OFFSET1:
-  case TRIG_OFFSET2:
-  case TRIG_OFFSET3:
-  case TRIG_OFFSET4:
-  case TRIG_OFFSET5:
-  case TRIG_OFFSET6:
-  case TRIG_OFFSET7:
-  case PPR_OFFSET:
-  case EXT_APIC_FEATURE_OFFSET:
-    PrintError("Attempting to write to read only register\n");
-    return -1;
-    break;
-
-    // Data registers
-  case APIC_ID_OFFSET:
-    apic->lapic_id.val = op_val;
-    break;
-  case TPR_OFFSET:
-    apic->task_prio.val = op_val;
-    break;
-  case LDR_OFFSET:
-    apic->log_dst.val = op_val;
-    break;
-  case DFR_OFFSET:
-    apic->dst_fmt.val = op_val;
-    break;
-  case SPURIOUS_INT_VEC_OFFSET:
-    apic->spurious_int.val = op_val;
-    break;
-  case ESR_OFFSET:
-    apic->err_status.val = op_val;
-    break;
-  case TMR_LOC_VEC_TBL_OFFSET:
-    apic->tmr_vec_tbl.val = op_val;
-    break;
-  case THERM_LOC_VEC_TBL_OFFSET:
-    apic->therm_loc_vec_tbl.val = op_val;
-    break;
-  case PERF_CTR_LOC_VEC_TBL_OFFSET:
-    apic->perf_ctr_loc_vec_tbl.val = op_val;
-    break;
-  case LINT0_VEC_TBL_OFFSET:
-    apic->lint0_vec_tbl.val = op_val;
-    break;
-  case LINT1_VEC_TBL_OFFSET:
-    apic->lint1_vec_tbl.val = op_val;
-    break;
-  case ERR_VEC_TBL_OFFSET:
-    apic->err_vec_tbl.val = op_val;
-    break;
-  case TMR_INIT_CNT_OFFSET:
-    apic->tmr_init_cnt = op_val;
-    apic->tmr_cur_cnt = op_val;
-    break;
-  case TMR_CUR_CNT_OFFSET:
-    apic->tmr_cur_cnt = op_val;
-    break;
-  case TMR_DIV_CFG_OFFSET:
-    apic->tmr_div_cfg.val = op_val;
-    break;
-
-
-    // Enable mask (256 bits)
-  case IER_OFFSET0:
-    *(uint32_t *)(apic->int_en_reg) = op_val;
-    break;
-  case IER_OFFSET1:
-    *(uint32_t *)(apic->int_en_reg + 4) = op_val;
-    break;
-  case IER_OFFSET2:
-    *(uint32_t *)(apic->int_en_reg + 8) = op_val;
-    break;
-  case IER_OFFSET3:
-    *(uint32_t *)(apic->int_en_reg + 12) = op_val;
-    break;
-  case IER_OFFSET4:
-    *(uint32_t *)(apic->int_en_reg + 16) = op_val;
-    break;
-  case IER_OFFSET5:
-    *(uint32_t *)(apic->int_en_reg + 20) = op_val;
-    break;
-  case IER_OFFSET6:
-    *(uint32_t *)(apic->int_en_reg + 24) = op_val;
-    break;
-  case IER_OFFSET7:
-    *(uint32_t *)(apic->int_en_reg + 28) = op_val;
-    break;
+
+    /* Because "May not be supported" doesn't matter to Linux developers... */
+    /*   if (length != 4) { */
+    /*     PrintError("Invalid apic read length (%d)\n", length); */
+    /*     return -1; */
+    /*   } */
+
+    switch (reg_addr & ~0x3) {
+       case EOI_OFFSET:
+           // Well, only an idiot would read from a architectural write only register
+           // Oh, Hello Linux.
+           //    PrintError("Attempting to read from write only register\n");
+           //    return -1;
+           break;
+
+           // data registers
+       case APIC_ID_OFFSET:
+           val = apic->lapic_id.val;
+           break;
+       case APIC_VERSION_OFFSET:
+           val = apic->apic_ver.val;
+           break;
+       case TPR_OFFSET:
+           val = apic->task_prio.val;
+           break;
+       case APR_OFFSET:
+           val = apic->arb_prio.val;
+           break;
+       case PPR_OFFSET:
+           val = apic->proc_prio.val;
+           break;
+       case REMOTE_READ_OFFSET:
+           val = apic->rem_rd_data;
+           break;
+       case LDR_OFFSET:
+           val = apic->log_dst.val;
+           break;
+       case DFR_OFFSET:
+           val = apic->dst_fmt.val;
+           break;
+       case SPURIOUS_INT_VEC_OFFSET:
+           val = apic->spurious_int.val;
+           break;
+       case ESR_OFFSET:
+           val = apic->err_status.val;
+           break;
+       case TMR_LOC_VEC_TBL_OFFSET:
+           val = apic->tmr_vec_tbl.val;
+           break;
+       case LINT0_VEC_TBL_OFFSET:
+           val = apic->lint0_vec_tbl.val;
+           break;
+       case LINT1_VEC_TBL_OFFSET:
+           val = apic->lint1_vec_tbl.val;
+           break;
+       case ERR_VEC_TBL_OFFSET:
+           val = apic->err_vec_tbl.val;
+           break;
+       case TMR_INIT_CNT_OFFSET:
+           val = apic->tmr_init_cnt;
+           break;
+       case TMR_DIV_CFG_OFFSET:
+           val = apic->tmr_div_cfg.val;
+           break;
+
+       case IER_OFFSET0:
+           val = *(uint32_t *)(apic->int_en_reg);
+           break;
+       case IER_OFFSET1:
+           val = *(uint32_t *)(apic->int_en_reg + 4);
+           break;
+       case IER_OFFSET2:
+           val = *(uint32_t *)(apic->int_en_reg + 8);
+           break;
+       case IER_OFFSET3:
+           val = *(uint32_t *)(apic->int_en_reg + 12);
+           break;
+       case IER_OFFSET4:
+           val = *(uint32_t *)(apic->int_en_reg + 16);
+           break;
+       case IER_OFFSET5:
+           val = *(uint32_t *)(apic->int_en_reg + 20);
+           break;
+       case IER_OFFSET6:
+           val = *(uint32_t *)(apic->int_en_reg + 24);
+           break;
+       case IER_OFFSET7:
+           val = *(uint32_t *)(apic->int_en_reg + 28);
+           break;
+
+       case ISR_OFFSET0:
+           val = *(uint32_t *)(apic->int_svc_reg);
+           break;
+       case ISR_OFFSET1:
+           val = *(uint32_t *)(apic->int_svc_reg + 4);
+           break;
+       case ISR_OFFSET2:
+           val = *(uint32_t *)(apic->int_svc_reg + 8);
+           break;
+       case ISR_OFFSET3:
+           val = *(uint32_t *)(apic->int_svc_reg + 12);
+           break;
+       case ISR_OFFSET4:
+           val = *(uint32_t *)(apic->int_svc_reg + 16);
+           break;
+       case ISR_OFFSET5:
+           val = *(uint32_t *)(apic->int_svc_reg + 20);
+           break;
+       case ISR_OFFSET6:
+           val = *(uint32_t *)(apic->int_svc_reg + 24);
+           break;
+       case ISR_OFFSET7:
+           val = *(uint32_t *)(apic->int_svc_reg + 28);
+           break;
    
+       case TRIG_OFFSET0:
+           val = *(uint32_t *)(apic->trig_mode_reg);
+           break;
+       case TRIG_OFFSET1:
+           val = *(uint32_t *)(apic->trig_mode_reg + 4);
+           break;
+       case TRIG_OFFSET2:
+           val = *(uint32_t *)(apic->trig_mode_reg + 8);
+           break;
+       case TRIG_OFFSET3:
+           val = *(uint32_t *)(apic->trig_mode_reg + 12);
+           break;
+       case TRIG_OFFSET4:
+           val = *(uint32_t *)(apic->trig_mode_reg + 16);
+           break;
+       case TRIG_OFFSET5:
+           val = *(uint32_t *)(apic->trig_mode_reg + 20);
+           break;
+       case TRIG_OFFSET6:
+           val = *(uint32_t *)(apic->trig_mode_reg + 24);
+           break;
+       case TRIG_OFFSET7:
+           val = *(uint32_t *)(apic->trig_mode_reg + 28);
+           break;
+
+       case IRR_OFFSET0:
+           val = *(uint32_t *)(apic->int_req_reg);
+           break;
+       case IRR_OFFSET1:
+           val = *(uint32_t *)(apic->int_req_reg + 4);
+           break;
+       case IRR_OFFSET2:
+           val = *(uint32_t *)(apic->int_req_reg + 8);
+           break;
+       case IRR_OFFSET3:
+           val = *(uint32_t *)(apic->int_req_reg + 12);
+           break;
+       case IRR_OFFSET4:
+           val = *(uint32_t *)(apic->int_req_reg + 16);
+           break;
+       case IRR_OFFSET5:
+           val = *(uint32_t *)(apic->int_req_reg + 20);
+           break;
+       case IRR_OFFSET6:
+           val = *(uint32_t *)(apic->int_req_reg + 24);
+           break;
+       case IRR_OFFSET7:
+           val = *(uint32_t *)(apic->int_req_reg + 28);
+           break;
+       case TMR_CUR_CNT_OFFSET:
+           val = apic->tmr_cur_cnt;
+           break;
+
+           // We are not going to implement these....
+       case THERM_LOC_VEC_TBL_OFFSET:
+           val = apic->therm_loc_vec_tbl.val;
+           break;
+       case PERF_CTR_LOC_VEC_TBL_OFFSET:
+           val = apic->perf_ctr_loc_vec_tbl.val;
+           break;
 
-    // Action Registers
-  case INT_CMD_LO_OFFSET:
-  case INT_CMD_HI_OFFSET:
-  case EOI_OFFSET:
-    // do eoi
-
-    // Unhandled Registers
-  case EXT_INT_LOC_VEC_TBL_OFFSET0:
-  case EXT_INT_LOC_VEC_TBL_OFFSET1:
-  case EXT_INT_LOC_VEC_TBL_OFFSET2:
-  case EXT_INT_LOC_VEC_TBL_OFFSET3:
-  case EXT_APIC_CMD_OFFSET:
-  case SEOI_OFFSET:
-  default:
-    PrintError("Write to Unhandled APIC Register: %x\n", (uint32_t)reg_addr);
-    return -1;
-  }
+
+           // handled registers
+       case INT_CMD_LO_OFFSET:    
+           val = apic->int_cmd.lo;
+           break;
+       case INT_CMD_HI_OFFSET:
+           val = apic->int_cmd.hi;
+           break;
+
+           // handle current timer count
+
+           // Unhandled Registers
+       case EXT_INT_LOC_VEC_TBL_OFFSET0:
+           val = apic->ext_intr_vec_tbl[0].val;
+           break;
+       case EXT_INT_LOC_VEC_TBL_OFFSET1:
+           val = apic->ext_intr_vec_tbl[1].val;
+           break;
+       case EXT_INT_LOC_VEC_TBL_OFFSET2:
+           val = apic->ext_intr_vec_tbl[2].val;
+           break;
+       case EXT_INT_LOC_VEC_TBL_OFFSET3:
+           val = apic->ext_intr_vec_tbl[3].val;
+           break;
+    
+
+       case EXT_APIC_FEATURE_OFFSET:
+       case EXT_APIC_CMD_OFFSET:
+       case SEOI_OFFSET:
+
+       default:
+           PrintError("apic %u: core %u: Read from Unhandled APIC Register: %x (getting zero)\n", 
+                      apic->lapic_id.val, core->cpu_id, (uint32_t)reg_addr);
+           return -1;
+    }
 
-  PrintDebug("Write finished\n");
 
-  return length;
+    if (length == 1) {
+       uint_t byte_addr = reg_addr & 0x3;
+       uint8_t * val_ptr = (uint8_t *)dst;
+    
+       *val_ptr = *(((uint8_t *)&val) + byte_addr);
+
+    } else if ((length == 2) && 
+              ((reg_addr & 0x3) == 0x3)) {
+       uint_t byte_addr = reg_addr & 0x3;
+       uint16_t * val_ptr = (uint16_t *)dst;
+       *val_ptr = *(((uint16_t *)&val) + byte_addr);
+
+    } else if (length == 4) {
+       uint32_t * val_ptr = (uint32_t *)dst;
+       *val_ptr = val;
+
+    } else {
+       PrintError("apic %u: core %u: Invalid apic read length (%d)\n", 
+                  apic->lapic_id.val, core->cpu_id, length);
+       return -1;
+    }
+
+    PrintDebug("apic %u: core %u: Read finished (val=%x)\n", 
+              apic->lapic_id.val, core->cpu_id, *(uint32_t *)dst);
+
+    return length;
 }
 
 
+/**
+ *
+ */
+static int apic_write(struct guest_info * core, addr_t guest_addr, void * src, uint_t length, void * priv_data) {
+    struct apic_state * apic = (struct apic_state *)(priv_data);
+    addr_t reg_addr  = guest_addr - apic->base_addr;
+    struct apic_msr * msr = (struct apic_msr *)&(apic->base_addr_msr.value);
+    uint32_t op_val = *(uint32_t *)src;
+
+    PrintDebug("apic %u: core %u: at %p and priv_data is at %p\n",
+              apic->lapic_id.val, core->cpu_id, apic, priv_data);
+
+    PrintDebug("Write to address space (%p) (val=%x)\n", 
+              (void *)guest_addr, *(uint32_t *)src);
+
+    if (msr->apic_enable == 0) {
+       PrintError("apic %u: core %u: Write to APIC address space with disabled APIC, apic msr=0x%llx\n",
+                  apic->lapic_id.val, core->cpu_id, apic->base_addr_msr.value);
+       return -1;
+    }
 
-/* Interrupt Controller Functions */
 
-static int apic_intr_pending(void * private_data) {
-  struct vm_device * dev = (struct vm_device *)private_data;
-  struct apic_state * apic = (struct apic_state *)dev->private_data;
-  int i = 0;
+    if (length != 4) {
+       PrintError("apic %u: core %u: Invalid apic write length (%d)\n", 
+                  apic->lapic_id.val, length, core->cpu_id);
+       return -1;
+    }
 
-  // just scan the request register looking for any set bit
-  // we should probably just do this with uint64 casts 
-  for (i = 0; i < 32; i++) {
-    if (apic->int_req_reg[i] & 0xff) {
-      return 1;
+    switch (reg_addr) {
+       case REMOTE_READ_OFFSET:
+       case APIC_VERSION_OFFSET:
+       case APR_OFFSET:
+       case IRR_OFFSET0:
+       case IRR_OFFSET1:
+       case IRR_OFFSET2:
+       case IRR_OFFSET3:
+       case IRR_OFFSET4:
+       case IRR_OFFSET5:
+       case IRR_OFFSET6:
+       case IRR_OFFSET7:
+       case ISR_OFFSET0:
+       case ISR_OFFSET1:
+       case ISR_OFFSET2:
+       case ISR_OFFSET3:
+       case ISR_OFFSET4:
+       case ISR_OFFSET5:
+       case ISR_OFFSET6:
+       case ISR_OFFSET7:
+       case TRIG_OFFSET0:
+       case TRIG_OFFSET1:
+       case TRIG_OFFSET2:
+       case TRIG_OFFSET3:
+       case TRIG_OFFSET4:
+       case TRIG_OFFSET5:
+       case TRIG_OFFSET6:
+       case TRIG_OFFSET7:
+       case PPR_OFFSET:
+       case EXT_APIC_FEATURE_OFFSET:
+
+           PrintError("apic %u: core %u: Attempting to write to read only register %p (error)\n", 
+                      apic->lapic_id.val, core->cpu_id, (void *)reg_addr);
+           //  return -1;
+
+           break;
+
+           // Data registers
+       case APIC_ID_OFFSET:
+           PrintDebug("apic %u: core %u: my id is being changed to %u\n", 
+                      apic->lapic_id.val, core->cpu_id, op_val);
+
+           apic->lapic_id.val = op_val;
+           break;
+       case TPR_OFFSET:
+           apic->task_prio.val = op_val;
+           break;
+       case LDR_OFFSET:
+           PrintDebug("apic %u: core %u: setting log_dst.val to 0x%x\n",
+                      apic->lapic_id.val, core->cpu_id, op_val);
+           apic->log_dst.val = op_val;
+           break;
+       case DFR_OFFSET:
+           apic->dst_fmt.val = op_val;
+           break;
+       case SPURIOUS_INT_VEC_OFFSET:
+           apic->spurious_int.val = op_val;
+           break;
+       case ESR_OFFSET:
+           apic->err_status.val = op_val;
+           break;
+       case TMR_LOC_VEC_TBL_OFFSET:
+           apic->tmr_vec_tbl.val = op_val;
+           break;
+       case THERM_LOC_VEC_TBL_OFFSET:
+           apic->therm_loc_vec_tbl.val = op_val;
+           break;
+       case PERF_CTR_LOC_VEC_TBL_OFFSET:
+           apic->perf_ctr_loc_vec_tbl.val = op_val;
+           break;
+       case LINT0_VEC_TBL_OFFSET:
+           apic->lint0_vec_tbl.val = op_val;
+           break;
+       case LINT1_VEC_TBL_OFFSET:
+           apic->lint1_vec_tbl.val = op_val;
+           break;
+       case ERR_VEC_TBL_OFFSET:
+           apic->err_vec_tbl.val = op_val;
+           break;
+       case TMR_INIT_CNT_OFFSET:
+           apic->tmr_init_cnt = op_val;
+           apic->tmr_cur_cnt = op_val;
+           break;
+       case TMR_CUR_CNT_OFFSET:
+           apic->tmr_cur_cnt = op_val;
+           break;
+       case TMR_DIV_CFG_OFFSET:
+           apic->tmr_div_cfg.val = op_val;
+           break;
+
+
+           // Enable mask (256 bits)
+       case IER_OFFSET0:
+           *(uint32_t *)(apic->int_en_reg) = op_val;
+           break;
+       case IER_OFFSET1:
+           *(uint32_t *)(apic->int_en_reg + 4) = op_val;
+           break;
+       case IER_OFFSET2:
+           *(uint32_t *)(apic->int_en_reg + 8) = op_val;
+           break;
+       case IER_OFFSET3:
+           *(uint32_t *)(apic->int_en_reg + 12) = op_val;
+           break;
+       case IER_OFFSET4:
+           *(uint32_t *)(apic->int_en_reg + 16) = op_val;
+           break;
+       case IER_OFFSET5:
+           *(uint32_t *)(apic->int_en_reg + 20) = op_val;
+           break;
+       case IER_OFFSET6:
+           *(uint32_t *)(apic->int_en_reg + 24) = op_val;
+           break;
+       case IER_OFFSET7:
+           *(uint32_t *)(apic->int_en_reg + 28) = op_val;
+           break;
+
+       case EXT_INT_LOC_VEC_TBL_OFFSET0:
+           apic->ext_intr_vec_tbl[0].val = op_val;
+           break;
+       case EXT_INT_LOC_VEC_TBL_OFFSET1:
+           apic->ext_intr_vec_tbl[1].val = op_val;
+           break;
+       case EXT_INT_LOC_VEC_TBL_OFFSET2:
+           apic->ext_intr_vec_tbl[2].val = op_val;
+           break;
+       case EXT_INT_LOC_VEC_TBL_OFFSET3:
+           apic->ext_intr_vec_tbl[3].val = op_val;
+           break;
+
+
+           // Action Registers
+       case EOI_OFFSET:
+           // do eoi
+           apic_do_eoi(apic);
+           break;
+
+       case INT_CMD_LO_OFFSET:
+           apic->int_cmd.lo = op_val;
+
+           // ICC???
+           PrintDebug("apic %u: core %u: sending cmd 0x%llx to apic %u\n", 
+                      apic->lapic_id.val, core->cpu_id,
+                      apic->int_cmd.val, apic->int_cmd.dst);
+           if (v3_icc_send_ipi(apic->icc_bus, apic->lapic_id.val, apic->int_cmd.val, 0)==-1) { 
+               return -1;
+           }
+           break;
+
+       case INT_CMD_HI_OFFSET:
+           apic->int_cmd.hi = op_val;
+           break;
+
+
+       // Unhandled Registers
+       case EXT_APIC_CMD_OFFSET:
+       case SEOI_OFFSET:
+       default:
+           PrintError("apic %u: core %u: Write to Unhandled APIC Register: %x (ignored)\n", 
+                      apic->lapic_id.val, core->cpu_id, (uint32_t)reg_addr);
+
+           return -1;
     }
-  }
 
-  return 0;
+    PrintDebug("apic %u: core %u: Write finished\n", apic->lapic_id.val, core->cpu_id);
+
+    return length;
 }
 
-static int apic_get_intr_number(void * private_data) {
-  struct vm_device * dev = (struct vm_device *)private_data;
-  struct apic_state * apic = (struct apic_state *)dev->private_data;
-  int i = 0, j = 0;
 
 
-  // We iterate backwards to find the highest priority
-  for (i = 31; i >= 0; i--) {
-    uchar_t req_major = apic->int_req_reg[i];
-    
-    if (req_major & 0xff) {
-      for (j = 7; j >= 0; j--) {
-       if ((req_major >> j) == 0x1) {
-         return (i * 32) + j;
-       }
-      }
+/* Interrupt Controller Functions */
+
+// returns 1 if an interrupt is pending, 0 otherwise
+static int apic_intr_pending(struct guest_info * info, void * private_data) {
+    struct apic_state * apic = (struct apic_state *)private_data;
+    int req_irq = get_highest_irr(apic);
+    int svc_irq = get_highest_isr(apic);
+
+    //    PrintDebug("apic %u: core %u: req_irq=%d, svc_irq=%d\n",apic->lapic_id.val,info->cpu_id,req_irq,svc_irq);
+
+    if ((req_irq >= 0) && 
+       (req_irq > svc_irq)) {
+       return 1;
     }
-  }
 
-  return -1;
+    return 0;
 }
 
-static int apic_raise_intr(void * private_data, int irq) {
-  struct vm_device * dev = (struct vm_device *)private_data;
-  struct apic_state * apic = (struct apic_state *)dev->private_data;
+static int apic_get_intr_number(struct guest_info * info, void * private_data) {
+    struct apic_state * apic = (struct apic_state *)private_data;
+    int req_irq = get_highest_irr(apic);
+    int svc_irq = get_highest_isr(apic);
 
-  return activate_apic_irq(apic, irq);
+    if (svc_irq == -1) {
+       return req_irq;
+    } else if (svc_irq < req_irq) {
+       return req_irq;
+    }
+
+    return -1;
 }
 
-static int apic_lower_intr(void * private_data, int irq) {
-  struct vm_device * dev = (struct vm_device *)private_data;
-  struct apic_state * apic = (struct apic_state *)dev->private_data;
-  int major_offset = irq & ~0x00000007;
-  int minor_offset = irq & 0x00000007;
-  uchar_t * req_location = apic->int_req_reg + major_offset;
-  uchar_t flag = 0x01 << minor_offset;
 
-  *req_location &= ~flag;
+static int apic_raise_intr(struct guest_info * info, int irq, void * private_data) {
+  struct apic_state * apic = (struct apic_state *)private_data;
 
-  return 0;
+  return activate_apic_irq(apic, irq);
 }
 
-static int apic_begin_irq(void * private_data, int irq) {
-  struct vm_device * dev = (struct vm_device *)private_data;
-  struct apic_state * apic = (struct apic_state *)dev->private_data;
-  int major_offset = irq & ~0x00000007;
-  int minor_offset = irq & 0x00000007;
-  uchar_t * req_location = apic->int_req_reg + major_offset;
-  uchar_t * svc_location = apic->int_svc_reg + major_offset;
-  uchar_t flag = 0x01 << minor_offset;
 
-  *svc_location |= flag;
-  *req_location &= ~flag;
 
-  return 0;
+static int apic_begin_irq(struct guest_info * info, void * private_data, int irq) {
+    struct apic_state * apic = (struct apic_state *)private_data;
+    int major_offset = (irq & ~0x00000007) >> 3;
+    int minor_offset = irq & 0x00000007;
+    uint8_t * req_location = apic->int_req_reg + major_offset;
+    uint8_t * svc_location = apic->int_svc_reg + major_offset;
+    uint8_t flag = 0x01 << minor_offset;
+
+    if (*req_location & flag) {
+       // we will only pay attention to a begin irq if we
+       // know that we initiated it!
+       *svc_location |= flag;
+       *req_location &= ~flag;
+    } else {
+       // do nothing... 
+       PrintDebug("apic %u: core %u: begin irq for %d ignored since I don't own it\n",
+                  apic->lapic_id.val, info->cpu_id, irq);
+    }
+
+    return 0;
 }
 
 
 
+
 /* Timer Functions */
-static void apic_update_time(ullong_t cpu_cycles, ullong_t cpu_freq, void * priv_data) {
-  struct vm_device * dev = (struct vm_device *)priv_data;
-  struct apic_state * apic = (struct apic_state *)dev->private_data;
-  // The 32 bit GCC runtime is a pile of shit
+static void apic_update_time(struct guest_info * info, 
+                            uint64_t cpu_cycles, uint64_t cpu_freq, 
+                            void * priv_data) {
+    struct apic_state * apic = (struct apic_state *)(priv_data);
+    // The 32 bit GCC runtime is a pile of shit
 #ifdef __V3_64BIT__
-  uint64_t tmr_ticks = 0;
+    uint64_t tmr_ticks = 0;
 #else 
-  uint32_t tmr_ticks = 0;
+    uint32_t tmr_ticks = 0;
 #endif
 
-  uchar_t tmr_div = *(uchar_t *)&(apic->tmr_div_cfg.val);
-  uint_t shift_num = 0;
+    uint8_t tmr_div = *(uint8_t *)&(apic->tmr_div_cfg.val);
+    uint_t shift_num = 0;
 
 
-  // Check whether this is true:
-  //   -> If the Init count is zero then the timer is disabled
-  //      and doesn't just blitz interrupts to the CPU
-  if ((apic->tmr_init_cnt == 0) || 
-      ( (apic->tmr_vec_tbl.tmr_mode == APIC_TMR_ONESHOT) &&
-       (apic->tmr_cur_cnt == 0))) {
-    //PrintDebug("APIC timer not yet initialized\n");
-    return;
-  }
+    // Check whether this is true:
+    //   -> If the Init count is zero then the timer is disabled
+    //      and doesn't just blitz interrupts to the CPU
+    if ((apic->tmr_init_cnt == 0) || 
+       ( (apic->tmr_vec_tbl.tmr_mode == APIC_TMR_ONESHOT) &&
+         (apic->tmr_cur_cnt == 0))) {
+       //PrintDebug("apic %u: core %u: APIC timer not yet initialized\n",apic->lapic_id.val,info->cpu_id);
+       return;
+    }
 
 
-  switch (tmr_div) {
-  case APIC_TMR_DIV1:
-    shift_num = 0;
-    break;
-  case APIC_TMR_DIV2:
-    shift_num = 1;
-    break;
-  case APIC_TMR_DIV4:
-    shift_num = 2;
-    break;
-  case APIC_TMR_DIV8:
-    shift_num = 3;
-    break;
-  case APIC_TMR_DIV16:
-    shift_num = 4;
-    break;
-  case APIC_TMR_DIV32:
-    shift_num = 5;
-    break;
-  case APIC_TMR_DIV64:
-    shift_num = 6;
-    break;
- case APIC_TMR_DIV128:
-    shift_num = 7;
-    break;
-  default:
-    PrintError("Invalid Timer Divider configuration\n");
-    return;
-  }
+    switch (tmr_div) {
+       case APIC_TMR_DIV1:
+           shift_num = 0;
+           break;
+       case APIC_TMR_DIV2:
+           shift_num = 1;
+           break;
+       case APIC_TMR_DIV4:
+           shift_num = 2;
+           break;
+       case APIC_TMR_DIV8:
+           shift_num = 3;
+           break;
+       case APIC_TMR_DIV16:
+           shift_num = 4;
+           break;
+       case APIC_TMR_DIV32:
+           shift_num = 5;
+           break;
+       case APIC_TMR_DIV64:
+           shift_num = 6;
+           break;
+       case APIC_TMR_DIV128:
+           shift_num = 7;
+           break;
+       default:
+           PrintError("apic %u: core %u: Invalid Timer Divider configuration\n",
+                      apic->lapic_id.val, info->cpu_id);
+           return;
+    }
 
-  tmr_ticks = cpu_cycles >> shift_num;
-  PrintDebug("Timer Ticks: %p\n", (void *)tmr_ticks);
+    tmr_ticks = cpu_cycles >> shift_num;
+    //    PrintDebug("Timer Ticks: %p\n", (void *)tmr_ticks);
 
-  if (tmr_ticks <= apic->tmr_cur_cnt) {
-    apic->tmr_cur_cnt -= tmr_ticks;
-  } else {
-    tmr_ticks -= apic->tmr_cur_cnt;
-    apic->tmr_cur_cnt = 0;
+    if (tmr_ticks < apic->tmr_cur_cnt) {
+       apic->tmr_cur_cnt -= tmr_ticks;
+    } else {
+       tmr_ticks -= apic->tmr_cur_cnt;
+       apic->tmr_cur_cnt = 0;
+
+       // raise irq
+       PrintDebug("apic %u: core %u: Raising APIC Timer interrupt (periodic=%d) (icnt=%d) (div=%d)\n",
+                  apic->lapic_id.val, info->cpu_id,
+                  apic->tmr_vec_tbl.tmr_mode, apic->tmr_init_cnt, shift_num);
 
-    // raise irq
-    if (activate_internal_irq(apic, APIC_TMR_INT) == -1) {
-      PrintError("Could not raise Timer interrupt\n");
+       if (apic_intr_pending(info, priv_data)) {
+           PrintDebug("apic %u: core %u: Overriding pending IRQ %d\n", 
+                      apic->lapic_id.val, info->cpu_id, 
+                      apic_get_intr_number(info, priv_data));
+       }
+
+       if (activate_internal_irq(apic, APIC_TMR_INT) == -1) {
+           PrintError("apic %u: core %u: Could not raise Timer interrupt\n",
+                      apic->lapic_id.val, info->cpu_id);
+       }
+    
+       if (apic->tmr_vec_tbl.tmr_mode == APIC_TMR_PERIODIC) {
+           tmr_ticks = tmr_ticks % apic->tmr_init_cnt;
+           apic->tmr_cur_cnt = apic->tmr_init_cnt - tmr_ticks;
+       }
     }
-  }
 
-  if (apic->tmr_vec_tbl.tmr_mode == APIC_TMR_PERIODIC) {
-    tmr_ticks = tmr_ticks % apic->tmr_init_cnt;
-    apic->tmr_init_cnt = apic->tmr_init_cnt - tmr_ticks;
-  }
-}
 
+}
 
 
 static struct intr_ctrl_ops intr_ops = {
-  .intr_pending = apic_intr_pending,
-  .get_intr_number = apic_get_intr_number,
-  .raise_intr = apic_raise_intr,
-  .begin_irq = apic_begin_irq,
-  .lower_intr = apic_lower_intr, 
+    .intr_pending = apic_intr_pending,
+    .get_intr_number = apic_get_intr_number,
+    .begin_irq = apic_begin_irq,
 };
 
 
 static struct vm_timer_ops timer_ops = {
-  .update_time = apic_update_time,
+    .update_timer = apic_update_time,
 };
 
 
-static int apic_init(struct vm_device * dev) {
-  struct guest_info * info = dev->vm;
-  struct apic_state * apic = (struct apic_state *)(dev->private_data);
 
-  v3_set_intr_controller(dev->vm, &intr_ops, dev);
-  v3_add_timer(dev->vm, &timer_ops, dev);
 
-  init_apic_state(apic);
+static int apic_free(struct vm_device * dev) {
 
-  v3_hook_msr(info, BASE_ADDR_MSR, read_apic_msr, write_apic_msr, dev);
+    /* TODO: This should crosscall to force an unhook on each CPU */
 
-  v3_hook_full_mem(info, apic->base_addr, apic->base_addr + PAGE_SIZE_4KB, apic_read, apic_write, dev);
+    //   struct apic_state * apic = (struct apic_state *)dev->private_data;
 
-  return 0;
+    v3_unhook_msr(dev->vm, BASE_ADDR_MSR);
+
+    return 0;
 }
 
-static int apic_deinit(struct vm_device * dev) {
-  struct guest_info * info = dev->vm;
 
-  v3_unhook_msr(info, BASE_ADDR_MSR);
+static struct v3_device_ops dev_ops = {
+    .free = apic_free,
+    .reset = NULL,
+    .start = NULL,
+    .stop = NULL,
+};
+
+static int apic_should_deliver_flat(struct guest_info * core, uint8_t mda, void * private_data)
+{
+  struct apic_state * apic = (struct apic_state *)private_data;
+
+  if (mda==0xff ||                         // broadcast or
+      (apic->log_dst.dst_log_id & mda)) {  // I am in the set 
+       PrintDebug("apic %u core %u: accepting flat IRQ (mda 0x%x == log_dst 0x%x)\n",
+                  apic->lapic_id.val, core->cpu_id, mda, apic->log_dst.dst_log_id);
+      return 1;
+  } else {
+       PrintDebug("apic %u core %u: rejecting flat IRQ (mda 0x%x != log_dst 0x%x)\n",
+                  apic->lapic_id.val, core->cpu_id, mda, apic->log_dst.dst_log_id);
+      return 0;
+  }
+}
+
+static int apic_should_deliver_cluster(struct guest_info * core, uint8_t mda, void * private_data)
+{
+  struct apic_state * apic = (struct apic_state *)private_data;
 
-  return 0;
+  if (mda==0xff ||                                                 // broadcast or
+      ( ((mda & 0xf0) == (apic->log_dst.dst_log_id & 0xf0)) &&     // (I am in the cluster and
+        ((mda & 0x0f)  & (apic->log_dst.dst_log_id & 0x0f)) ) ) {  //  I am in the set)
+       PrintDebug("apic %u core %u: accepting clustered IRQ (mda 0x%x == log_dst 0x%x)\n",
+                  apic->lapic_id.val, core->cpu_id, mda, apic->log_dst.dst_log_id);
+                  
+      return 1;
+  } else {
+       PrintDebug("apic %u core %u: rejecting clustered IRQ (mda 0x%x != log_dst 0x%x)\n",
+                  apic->lapic_id.val, core->cpu_id, mda, apic->log_dst.dst_log_id);
+      return 0;
+  }
 }
 
+static int apic_should_deliver(struct guest_info * core, uint8_t mda, void *private_data)
+{
+    struct apic_state * apic = (struct apic_state *)private_data;
+    if (apic->dst_fmt.model == 0xf) {
+       return apic_should_deliver_cluster(core, mda, private_data);
+    } else if (apic->dst_fmt.model == 0x0) {
+       return apic_should_deliver_flat(core, mda, private_data);
+    } else {
+       PrintError("apic %u core %u: invalid destination format register value 0x%x for logical mode delivery.\n", apic->lapic_id.val, core->cpu_id, apic->dst_fmt.model);
+       return 0;
+    }
+}
 
-static struct vm_device_ops dev_ops = {
-  .init = apic_init,
-  .deinit = apic_deinit,
-  .reset = NULL,
-  .start = NULL,
-  .stop = NULL,
+static struct v3_icc_ops icc_ops = {
+    .raise_intr = apic_raise_intr,
+    .should_deliver = apic_should_deliver,
 };
 
+static int apic_init(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
+    PrintDebug("apic: creating an APIC for each core\n");
+    char * dev_id = v3_cfg_val(cfg, "ID");
+    char * icc_bus_id = v3_cfg_val(cfg, "bus");
+    struct vm_device * icc = v3_find_dev(vm, icc_bus_id);
+    int i;
+
+    if (!icc) {
+        PrintError("apic: Cannot find ICC Bus (%s)\n", icc_bus_id);
+        return -1;
+    }
 
-struct vm_device * v3_create_apic() {
-  PrintDebug("Creating APIC\n");
+    // We allocate one apic per core
+    // APICs are accessed via index which correlates with the core's cpu_id 
+    // 0..num_cores-1   at num_cores is the ioapic (one only)
+    struct apic_state * apic = (struct apic_state *)V3_Malloc(sizeof(struct apic_state) * vm->num_cores);
 
-  struct apic_state * apic = (struct apic_state *)V3_Malloc(sizeof(struct apic_state));
+    struct vm_device * dev = v3_allocate_device(dev_id, &dev_ops, apic);
 
-  struct vm_device * device = v3_create_device("APIC", &dev_ops, apic);
-  
-  return device;
+    if (v3_attach_device(vm, dev) == -1) {
+       PrintError("apic: Could not attach device %s\n", dev_id);
+       return -1;
+    }
+
+    
+    for (i = 0; i < vm->num_cores; i++) {
+       struct guest_info * core = &(vm->cores[i]);
+
+       init_apic_state(&(apic[i]),i,icc);
+
+       v3_register_intr_controller(core, &intr_ops, &(apic[i]));
+
+       v3_add_timer(core, &timer_ops, &(apic[i]));
+
+       v3_hook_full_mem(vm, core->cpu_id, apic->base_addr, apic->base_addr + PAGE_SIZE_4KB, apic_read, apic_write, &(apic[i]));
+
+       v3_icc_register_apic(core, icc, i, &icc_ops, &(apic[i]));
+
+       PrintDebug("apic %u: (setup device): done, my id is %u\n", i, apic[i].lapic_id.val);
+
+    }
+
+    for (i = 0; i < vm->num_cores; i++) {
+       PrintDebug("apic: sanity check: apic %u (at %p) has id %u and msr value %llx\n",
+                  i, &(apic[i]), apic[i].lapic_id.val, apic[i].base_addr_msr.value);
+    }
+
+    PrintDebug("apic: priv_data is at %p\n", apic);
+
+    v3_hook_msr(vm, BASE_ADDR_MSR, read_apic_msr, write_apic_msr, dev);
+
+    return 0;
 }
+
+
+
+device_register("LAPIC", apic_init)