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.


Minor changes to debugging, change to Makefile to support building TAGS files, added...
[palacios.git] / palacios / src / palacios / vmx.c
index 2526f3d..7cea8a1 100644 (file)
 #include <palacios/vmx_io.h>
 #include <palacios/vmx_msr.h>
 
+
+#ifndef CONFIG_DEBUG_VMX
+#undef PrintDebug
+#define PrintDebug(fmt, args...)
+#endif
+
+
 static addr_t host_vmcs_ptrs[CONFIG_MAX_CPUS] = { [0 ... CONFIG_MAX_CPUS - 1] = 0};
 
 
@@ -199,8 +206,7 @@ static int init_vmcs_bios(struct guest_info * info, struct vmx_data * vmx_state)
 
     /********** Setup and VMX Control Fields from MSR ***********/
     /* Setup IO map */
-    v3_init_vmx_io_map(info);
-    v3_init_vmx_msr_map(info);
+
 
     struct v3_msr tmp_msr;
 
@@ -219,12 +225,17 @@ static int init_vmcs_bios(struct guest_info * info, struct vmx_data * vmx_state)
     vmx_state->pri_proc_ctrls.invlpg_exit = 1;
     vmx_state->pri_proc_ctrls.use_msr_bitmap = 1;
     vmx_state->pri_proc_ctrls.pause_exit = 1;
+    vmx_state->pri_proc_ctrls.tsc_offset = 1;
+#ifdef CONFIG_TIME_VIRTUALIZE_TSC
+    vmx_state->pri_proc_ctrls.rdtsc_exit = 1;
+#endif
 
-    vmx_ret |= check_vmcs_write(VMCS_IO_BITMAP_A_ADDR, (addr_t)V3_PAddr(info->io_map.arch_data));
+    vmx_ret |= check_vmcs_write(VMCS_IO_BITMAP_A_ADDR, (addr_t)V3_PAddr(info->vm_info->io_map.arch_data));
     vmx_ret |= check_vmcs_write(VMCS_IO_BITMAP_B_ADDR, 
-            (addr_t)V3_PAddr(info->io_map.arch_data) + PAGE_SIZE_4KB); 
+            (addr_t)V3_PAddr(info->vm_info->io_map.arch_data) + PAGE_SIZE_4KB);
 
-    vmx_ret |= check_vmcs_write(VMCS_MSR_BITMAP, (addr_t)V3_PAddr(info->msr_map.arch_data));
+
+    vmx_ret |= check_vmcs_write(VMCS_MSR_BITMAP, (addr_t)V3_PAddr(info->vm_info->msr_map.arch_data));
 
     v3_get_msr(VMX_EXIT_CTLS_MSR, &(tmp_msr.hi), &(tmp_msr.lo));
     vmx_state->exit_ctrls.value = tmp_msr.lo;
@@ -341,7 +352,7 @@ static int init_vmcs_bios(struct guest_info * info, struct vmx_data * vmx_state)
 #define VMXASSIST_GDT   0x10000
        addr_t vmxassist_gdt = 0;
 
-       if (guest_pa_to_host_va(info, VMXASSIST_GDT, &vmxassist_gdt) == -1) {
+       if (v3_gpa_to_hva(info, VMXASSIST_GDT, &vmxassist_gdt) == -1) {
            PrintError("Could not find VMXASSIST GDT destination\n");
            return -1;
        }
@@ -375,7 +386,7 @@ static int init_vmcs_bios(struct guest_info * info, struct vmx_data * vmx_state)
        extern uint8_t v3_vmxassist_end[];
        addr_t vmxassist_dst = 0;
 
-       if (guest_pa_to_host_va(info, VMXASSIST_START, &vmxassist_dst) == -1) {
+       if (v3_gpa_to_hva(info, VMXASSIST_START, &vmxassist_dst) == -1) {
            PrintError("Could not find VMXASSIST destination\n");
            return -1;
        }
@@ -453,12 +464,12 @@ static int update_irq_exit_state(struct guest_info * info) {
 
     check_vmcs_read(VMCS_IDT_VECTOR_INFO, &(idt_vec_info.value));
 
-    if ((info->intr_state.irq_started == 1) && (idt_vec_info.valid == 0)) {
+    if ((info->intr_core_state.irq_started == 1) && (idt_vec_info.valid == 0)) {
 #ifdef CONFIG_DEBUG_INTERRUPTS
         PrintDebug("Calling v3_injecting_intr\n");
 #endif
-        info->intr_state.irq_started = 0;
-        v3_injecting_intr(info, info->intr_state.irq_vector, V3_EXTERNAL_IRQ);
+        info->intr_core_state.irq_started = 0;
+        v3_injecting_intr(info, info->intr_core_state.irq_vector, V3_EXTERNAL_IRQ);
     }
 
     return 0;
@@ -466,11 +477,11 @@ static int update_irq_exit_state(struct guest_info * info) {
 
 static int update_irq_entry_state(struct guest_info * info) {
     struct vmx_exit_idt_vec_info idt_vec_info;
-    struct vmcs_interrupt_state intr_state;
+    struct vmcs_interrupt_state intr_core_state;
     struct vmx_data * vmx_info = (struct vmx_data *)(info->vmm_data);
 
     check_vmcs_read(VMCS_IDT_VECTOR_INFO, &(idt_vec_info.value));
-    check_vmcs_read(VMCS_GUEST_INT_STATE, &(intr_state));
+    check_vmcs_read(VMCS_GUEST_INT_STATE, &(intr_core_state));
 
     /* Check for pending exceptions to inject */
     if (v3_excp_pending(info)) {
@@ -494,16 +505,16 @@ static int update_irq_entry_state(struct guest_info * info) {
 
         int_info.valid = 1;
 #ifdef CONFIG_DEBUG_INTERRUPTS
-        PrintDebug("Injecting exception %d (EIP=%p)\n", int_info.vector, (void *)info->rip);
+        PrintDebug("Injecting exception %d (EIP=%p)\n", int_info.vector, (void *)(addr_t)info->rip);
 #endif
         check_vmcs_write(VMCS_ENTRY_INT_INFO, int_info.value);
 
         v3_injecting_excp(info, int_info.vector);
 
     } else if ((((struct rflags *)&(info->ctrl_regs.rflags))->intr == 1) && 
-              (intr_state.val == 0)) {
+              (intr_core_state.val == 0)) {
        
-        if ((info->intr_state.irq_started == 1) && (idt_vec_info.valid == 1)) {
+        if ((info->intr_core_state.irq_started == 1) && (idt_vec_info.valid == 1)) {
 
 #ifdef CONFIG_DEBUG_INTERRUPTS
             PrintDebug("IRQ pending from previous injection\n");
@@ -526,21 +537,21 @@ static int update_irq_entry_state(struct guest_info * info) {
 
             switch (v3_intr_pending(info)) {
                 case V3_EXTERNAL_IRQ: {
-                    info->intr_state.irq_vector = v3_get_intr(info); 
-                    ent_int.vector = info->intr_state.irq_vector;
+                    info->intr_core_state.irq_vector = v3_get_intr(info); 
+                    ent_int.vector = info->intr_core_state.irq_vector;
                     ent_int.type = 0;
                     ent_int.error_code = 0;
                     ent_int.valid = 1;
 
 #ifdef CONFIG_DEBUG_INTERRUPTS
                     PrintDebug("Injecting Interrupt %d at exit %u(EIP=%p)\n", 
-                              info->intr_state.irq_vector, 
+                              info->intr_core_state.irq_vector, 
                               (uint32_t)info->num_exits, 
-                              (void *)info->rip);
+                              (void *)(addr_t)info->rip);
 #endif
 
                     check_vmcs_write(VMCS_ENTRY_INT_INFO, ent_int.value);
-                    info->intr_state.irq_started = 1;
+                    info->intr_core_state.irq_started = 1;
 
                     break;
                 }
@@ -629,13 +640,12 @@ static void print_exit_log(struct guest_info * info) {
  */
 int v3_vmx_enter(struct guest_info * info) {
     int ret = 0;
-    uint64_t tmp_tsc = 0;
+    uint32_t tsc_offset_low, tsc_offset_high;
     struct vmx_exit_info exit_info;
 
     // Conditionally yield the CPU if the timeslice has expired
     v3_yield_cond(info);
 
-
     // v3_print_guest_state(info);
 
     // disable global interrupts for vm state transition
@@ -644,19 +654,33 @@ int v3_vmx_enter(struct guest_info * info) {
     v3_vmx_restore_vmcs(info);
 
 
-#ifdef CONFIG_SYMBIOTIC
-    if (info->sym_state.sym_call_active == 0) {
+#ifdef CONFIG_SYMCALL
+    if (info->sym_core_state.symcall_state.sym_call_active == 0) {
        update_irq_entry_state(info);
     }
 #else 
     update_irq_entry_state(info);
 #endif
 
+    {
+       addr_t guest_cr3;
+       vmcs_read(VMCS_GUEST_CR3, &guest_cr3);
+       vmcs_write(VMCS_GUEST_CR3, guest_cr3);
+    }
+
+    v3_update_timers(info);
+
+    /* If this guest is frequency-lagged behind host time, wait 
+     * for the appropriate host time before resuming the guest. */
+    v3_adjust_time(info);
 
-    rdtscll(info->time_state.cached_host_tsc);
+    tsc_offset_high = (uint32_t)((v3_tsc_host_offset(&info->time_state) >> 32) & 0xffffffff);
+    tsc_offset_low = (uint32_t)(v3_tsc_host_offset(&info->time_state) & 0xffffffff);
+    check_vmcs_write(VMCS_TSC_OFFSET_HIGH, tsc_offset_high);
+    check_vmcs_write(VMCS_TSC_OFFSET, tsc_offset_low);
 
-    if (info->run_state == VM_STOPPED) {
-       info->run_state = VM_RUNNING;
+    if (info->vm_info->run_state == VM_STOPPED) {
+       info->vm_info->run_state = VM_RUNNING;
        ret = v3_vmx_launch(&(info->vm_regs), info, &(info->ctrl_regs));
     } else {
        ret = v3_vmx_resume(&(info->vm_regs), info, &(info->ctrl_regs));
@@ -673,15 +697,13 @@ int v3_vmx_enter(struct guest_info * info) {
        return -1;
     }
 
-    rdtscll(tmp_tsc);
-
     info->num_exits++;
 
-    v3_update_time(info, tmp_tsc - info->time_state.cached_host_tsc);
-
     /* Update guest state */
     v3_vmx_save_vmcs(info);
 
+    // info->cpl = info->segments.cs.selector & 0x3;
+
     info->mem_mode = v3_get_vm_mem_mode(info);
     info->cpu_mode = v3_get_vm_cpu_mode(info);
 
@@ -699,8 +721,8 @@ int v3_vmx_enter(struct guest_info * info) {
     exit_log[info->num_exits % 10] = exit_info;
 
 
-#ifdef CONFIG_SYMBIOTIC
-    if (info->sym_state.sym_call_active == 0) {
+#ifdef CONFIG_SYMCALL
+    if (info->sym_core_state.symcall_state.sym_call_active == 0) {
        update_irq_exit_state(info);
     }
 #else
@@ -724,11 +746,38 @@ int v3_vmx_enter(struct guest_info * info) {
 
 int v3_start_vmx_guest(struct guest_info* info) {
 
+    PrintDebug("Starting VMX core %u\n",info->cpu_id);
+    if (info->cpu_mode==INIT) {
+        PrintDebug("VMX core %u: I am an AP in INIT mode, waiting for that to change\n",info->cpu_id);
+        while (info->cpu_mode==INIT) {
+            v3_yield(info);
+            //PrintDebug("VMX core %u: still waiting for INIT\n",info->cpu_id);
+        }
+        PrintDebug("VMX core %u: I am out of INIT\n",info->cpu_id);
+        if (info->cpu_mode==SIPI) {
+            PrintDebug("VMX core %u: I am waiting on a SIPI to set my starting address\n",info->cpu_id);
+            while (info->cpu_mode==SIPI) {
+                v3_yield(info);
+                //PrintDebug("VMX core %u: still waiting for SIPI\n",info->cpu_id);
+            }
+        }
+        PrintDebug("VMX core %u: I have my SIPI\n", info->cpu_id);
+    }
+
+    if (info->cpu_mode!=REAL) {
+        PrintError("VMX core %u: I am not in REAL mode at launch!  Huh?!\n", info->cpu_id);
+        return -1;
+    }
+
+    PrintDebug("VMX core %u: I am starting at CS=0x%x (base=0x%p, limit=0x%x),  RIP=0x%p\n",
+               info->cpu_id, info->segments.cs.selector, (void*)(info->segments.cs.base),
+               info->segments.cs.limit,(void*)(info->rip));
+
 
-    PrintDebug("Launching VMX guest\n");
 
-    rdtscll(info->time_state.cached_host_tsc);
+    PrintDebug("VMX core %u: Launching VMX VM\n", info->cpu_id);
 
+    v3_start_time(info);
 
     while (1) {
        if (v3_vmx_enter(info) == -1) {
@@ -787,7 +836,7 @@ void v3_init_vmx_cpu(int cpu_id) {
     uint64_t ret = 0;
 
     v3_get_msr(VMX_CR4_FIXED0_MSR,&(tmp_msr.hi),&(tmp_msr.lo));
-    
+#ifdef __V3_64BIT__
     __asm__ __volatile__ (
                          "movq %%cr4, %%rbx;"
                          "orq  $0x00002000, %%rbx;"
@@ -816,6 +865,38 @@ void v3_init_vmx_cpu(int cpu_id) {
                          :
                          : "%rbx"
                          );
+#elif __V3_32BIT__
+    __asm__ __volatile__ (
+                         "movl %%cr4, %%ecx;"
+                         "orl  $0x00002000, %%ecx;"
+                         "movl %%ecx, %0;"
+                         : "=m"(ret) 
+                         :
+                         : "%ecx"
+                         );
+
+    if ((~ret & tmp_msr.value) == 0) {
+        __asm__ __volatile__ (
+                             "movl %0, %%cr4;"
+                             :
+                             : "q"(ret)
+                             );
+    } else {
+        PrintError("Invalid CR4 Settings!\n");
+        return;
+    }
+
+    __asm__ __volatile__ (
+                         "movl %%cr0, %%ecx; "
+                         "orl  $0x00000020,%%ecx; "
+                         "movl %%ecx, %%cr0;"
+                         :
+                         :
+                         : "%ecx"
+                         );
+
+#endif
+
     //
     // Should check and return Error here....