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.


Simplified interface to timekeeping API, added to VMX code
[palacios.git] / palacios / src / palacios / vmx.c
index 15d108b..ef0f614 100644 (file)
 #include <palacios/vmm_lowlevel.h>
 #include <palacios/vmm_ctrl_regs.h>
 #include <palacios/vmm_config.h>
+#include <palacios/vmm_time.h>
 #include <palacios/vm_guest_mem.h>
 #include <palacios/vmm_direct_paging.h>
 #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};
 
 
@@ -218,6 +226,10 @@ 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->vm_info->io_map.arch_data));
     vmx_ret |= check_vmcs_write(VMCS_IO_BITMAP_B_ADDR, 
@@ -341,7 +353,7 @@ 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 +387,7 @@ 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;
        }
@@ -391,8 +403,12 @@ static int init_vmcs_bios(struct guest_info * info, struct vmx_data * vmx_state)
 
     info->dbg_regs.dr7 = 0x400;
 
+#ifdef __V3_64BIT__
     vmx_ret |= check_vmcs_write(VMCS_LINK_PTR, (addr_t)0xffffffffffffffffULL);
-    
+#else
+    vmx_ret |= check_vmcs_write(VMCS_LINK_PTR, (addr_t)0xffffffffUL);
+    vmx_ret |= check_vmcs_write(VMCS_LINK_PTR_HIGH, (addr_t)0xffffffffUL);
+#endif
 
     if (v3_update_vmcs_ctrl_fields(info)) {
         PrintError("Could not write control fields!\n");
@@ -448,6 +464,18 @@ int v3_init_vmx_vmcs(struct guest_info * info, v3_vm_class_t vm_class) {
     return 0;
 }
 
+
+int v3_deinit_vmx_vmcs(struct guest_info * core) {
+    struct vmx_data * vmx_state = core->vmm_data;
+
+    V3_FreePages((void *)(vmx_state->vmcs_ptr_phys), 1);
+
+    V3_Free(vmx_state);
+
+    return 0;
+}
+
+
 static int update_irq_exit_state(struct guest_info * info) {
     struct vmx_exit_idt_vec_info idt_vec_info;
 
@@ -494,7 +522,7 @@ 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);
 
@@ -536,7 +564,7 @@ static int update_irq_entry_state(struct guest_info * info) {
                     PrintDebug("Injecting Interrupt %d at exit %u(EIP=%p)\n", 
                               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);
@@ -629,14 +657,17 @@ 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);
 
+    // Perform any additional yielding needed for time adjustment
+    v3_adjust_time(info);
 
-    // v3_print_guest_state(info);
+    // Update timer devices prior to entering VM.
+    v3_update_timers(info);
 
     // disable global interrupts for vm state transition
     v3_disable_ints();
@@ -644,8 +675,8 @@ int v3_vmx_enter(struct guest_info * info) {
     v3_vmx_restore_vmcs(info);
 
 
-#ifdef CONFIG_SYMBIOTIC
-    if (info->vm_info->sym_state.symcalls[info->cpu_id].sym_call_active == 0) {
+#ifdef CONFIG_SYMCALL
+    if (info->sym_core_state.symcall_state.sym_call_active == 0) {
        update_irq_entry_state(info);
     }
 #else 
@@ -658,7 +689,13 @@ int v3_vmx_enter(struct guest_info * info) {
        vmcs_write(VMCS_GUEST_CR3, guest_cr3);
     }
 
-    rdtscll(info->time_state.cached_host_tsc);
+    // Perform last-minute time bookkeeping prior to entering the VM
+    v3_time_enter_vm(info);
+
+    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->vm_info->run_state == VM_STOPPED) {
        info->vm_info->run_state = VM_RUNNING;
@@ -678,12 +715,11 @@ int v3_vmx_enter(struct guest_info * info) {
        return -1;
     }
 
-    rdtscll(tmp_tsc);
+    // Immediate exit from VM time bookkeeping
+    v3_time_exit_vm(info);
 
     info->num_exits++;
 
-    v3_update_time(info, tmp_tsc - info->time_state.cached_host_tsc);
-
     /* Update guest state */
     v3_vmx_save_vmcs(info);
 
@@ -706,8 +742,8 @@ int v3_vmx_enter(struct guest_info * info) {
     exit_log[info->num_exits % 10] = exit_info;
 
 
-#ifdef CONFIG_SYMBIOTIC
-    if (info->vm_info->sym_state.symcalls[info->cpu_id].sym_call_active == 0) {
+#ifdef CONFIG_SYMCALL
+    if (info->sym_core_state.symcall_state.sym_call_active == 0) {
        update_irq_exit_state(info);
     }
 #else
@@ -729,21 +765,54 @@ int v3_vmx_enter(struct guest_info * info) {
 }
 
 
-int v3_start_vmx_guest(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_id == 0) {
+       info->core_run_state = CORE_RUNNING;
+       info->vm_info->run_state = VM_RUNNING;
+    } else {
+
+        PrintDebug("VMX core %u: Waiting for core initialization\n", info->cpu_id);
+
+        while (info->core_run_state == CORE_STOPPED) {
+            v3_yield(info);
+            //PrintDebug("VMX core %u: still waiting for INIT\n",info->cpu_id);
+        }
+       
+       PrintDebug("VMX core %u initialized\n", info->cpu_id);
+    }
+
 
+    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 (info->vm_info->run_state == VM_STOPPED) {
+           info->core_run_state = CORE_STOPPED;
+           break;
+       }
+
        if (v3_vmx_enter(info) == -1) {
            v3_print_vmcs();
            print_exit_log(info);
            return -1;
        }
 
+
+
+       if (info->vm_info->run_state == VM_STOPPED) {
+           info->core_run_state = CORE_STOPPED;
+           break;
+       }
 /*
        if ((info->num_exits % 5000) == 0) {
            V3_Print("VMX Exit number %d\n", (uint32_t)info->num_exits);
@@ -793,8 +862,9 @@ void v3_init_vmx_cpu(int cpu_id) {
     struct v3_msr tmp_msr;
     uint64_t ret = 0;
 
-    v3_get_msr(VMX_CR4_FIXED0_MSR,&(tmp_msr.hi),&(tmp_msr.lo));
-    
+    v3_get_msr(VMX_CR4_FIXED0_MSR, &(tmp_msr.hi), &(tmp_msr.lo));
+
+#ifdef __V3_64BIT__
     __asm__ __volatile__ (
                          "movq %%cr4, %%rbx;"
                          "orq  $0x00002000, %%rbx;"
@@ -823,6 +893,38 @@ 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.... 
 
@@ -848,3 +950,9 @@ void v3_init_vmx_cpu(int cpu_id) {
 
 }
 
+
+void v3_deinit_vmx_cpu(int cpu_id) {
+    extern v3_cpu_arch_t v3_cpu_types[];
+    v3_cpu_types[cpu_id] = V3_INVALID_CPU;
+    V3_FreePages((void *)host_vmcs_ptrs[cpu_id], 1);
+}