2 * This file is part of the Palacios Virtual Machine Monitor developed
3 * by the V3VEE Project with funding from the United States National
4 * Science Foundation and the Department of Energy.
6 * The V3VEE Project is a joint project between Northwestern University
7 * and the University of New Mexico. You can find out more at
10 * Copyright (c) 2008, Jack Lange <jarusl@cs.northwestern.edu>
11 * Copyright (c) 2008, The V3VEE Project <http://www.v3vee.org>
12 * All rights reserved.
14 * Author: Jack Lange <jarusl@cs.northwestern.edu>
16 * This is free software. You are permitted to use,
17 * redistribute, and modify it as specified in the file "V3VEE_LICENSE".
20 #include <palacios/vmm_config.h>
21 #include <palacios/vmm.h>
22 #include <palacios/vmm_debug.h>
23 #include <palacios/vmm_msr.h>
24 #include <palacios/vmm_decoder.h>
25 #include <palacios/vmm_telemetry.h>
26 #include <palacios/vmm_mem.h>
27 #include <palacios/vmm_hypercall.h>
28 #include <palacios/vmm_dev_mgr.h>
29 #include <palacios/vmm_cpuid.h>
30 #include <palacios/vmm_xml.h>
31 #include <palacios/vmm_io.h>
32 #include <palacios/vmm_msr.h>
33 #include <palacios/vmm_sprintf.h>
36 #ifdef V3_CONFIG_SWAPPING
37 #include <palacios/vmm_swapping.h>
40 #include <palacios/vmm_host_events.h>
41 #include <palacios/vmm_perftune.h>
43 #include "vmm_config_class.h"
46 /* The Palacios cookie encodes "v3vee" followed by a
47 3 byte version code. There are currently two versions:
49 \0\0\0 => original (no checksum)
53 #define COOKIE_V0 "v3vee\0\0\0"
54 #define COOKIE_V1 "v3vee\0\0\1"
56 // This is used to access the configuration file index table
71 struct file_idx_table_v0 {
73 struct file_hdr_v0 hdrs[0];
76 struct file_idx_table_v1 {
78 struct file_hdr_v1 hdrs[0];
84 static int setup_memory_map(struct v3_vm_info * vm, v3_cfg_tree_t * cfg);
85 static int setup_extensions(struct v3_vm_info * vm, v3_cfg_tree_t * cfg);
86 static int setup_devices(struct v3_vm_info * vm, v3_cfg_tree_t * cfg);
90 char * v3_cfg_val(v3_cfg_tree_t * tree, char * tag) {
91 char * attrib = (char *)v3_xml_attr(tree, tag);
92 v3_cfg_tree_t * child_entry = v3_xml_child(tree, tag);
95 if ((child_entry != NULL) && (attrib != NULL)) {
96 PrintError(VM_NONE, VCORE_NONE, "Duplicate Configuration parameters present for %s\n", tag);
100 if (attrib == NULL) {
101 val = v3_xml_txt(child_entry);
112 v3_cfg_tree_t * v3_cfg_subtree(v3_cfg_tree_t * tree, char * tag) {
113 return v3_xml_child(tree, tag);
116 v3_cfg_tree_t * v3_cfg_next_branch(v3_cfg_tree_t * tree) {
117 return v3_xml_next(tree);
122 struct v3_cfg_file * v3_cfg_get_file(struct v3_vm_info * vm, char * tag) {
123 struct v3_cfg_file * file = NULL;
125 file = (struct v3_cfg_file *)v3_htable_search(vm->cfg_data->file_table, (addr_t)tag);
131 static uint_t file_hash_fn(addr_t key) {
132 char * name = (char *)key;
133 return v3_hash_buffer((uchar_t *)name, strlen(name));
136 static int file_eq_fn(addr_t key1, addr_t key2) {
137 char * name1 = (char *)key1;
138 char * name2 = (char *)key2;
140 return (strcmp(name1, name2) == 0);
143 static struct v3_config * parse_config(void * cfg_blob) {
144 struct v3_config * cfg = NULL;
147 struct file_idx_table_v0 * files_v0 = NULL;
148 struct file_idx_table_v1 * files_v1 = NULL;
149 v3_cfg_tree_t * file_tree = NULL;
152 V3_Print(VM_NONE, VCORE_NONE, "cfg data at %p\n", cfg_blob);
154 if (memcmp(cfg_blob, COOKIE_V0, COOKIE_LEN) == 0) {
156 } else if (memcmp(cfg_blob, COOKIE_V1, COOKIE_LEN) == 0) {
159 PrintError(VM_NONE, VCORE_NONE, "Invalid Configuration Header Or Unknown Version\n");
163 V3_Print(VM_NONE, VCORE_NONE, "Handling Palacios Image Format, Version 0x%x\n",version);
165 offset += COOKIE_LEN;
167 cfg = (struct v3_config *)V3_Malloc(sizeof(struct v3_config));
170 PrintError(VM_NONE, VCORE_NONE, "Unable to allocate while parsing\n");
174 memset(cfg, 0, sizeof(struct v3_config));
176 cfg->blob = cfg_blob;
177 INIT_LIST_HEAD(&(cfg->file_list));
178 cfg->file_table = v3_create_htable(0, file_hash_fn, file_eq_fn);
180 if (!(cfg->file_table)) {
181 PrintError(VM_NONE, VCORE_NONE, "Unable to allocate hash table while parsing\n");
186 xml_len = *(uint32_t *)(cfg_blob + offset);
189 cfg->cfg = (v3_cfg_tree_t *)v3_xml_parse((uint8_t *)(cfg_blob + offset));
194 // This is hideous, but the file formats are still very close
196 files_v0 = (struct file_idx_table_v0 *)(cfg_blob + offset);
197 V3_Print(VM_NONE, VCORE_NONE, "Number of files in cfg: %d\n", (uint32_t)(files_v0->num_files));
199 files_v1 = (struct file_idx_table_v1 *)(cfg_blob + offset);
200 V3_Print(VM_NONE, VCORE_NONE, "Number of files in cfg: %d\n", (uint32_t)(files_v1->num_files));
204 file_tree = v3_cfg_subtree(v3_cfg_subtree(cfg->cfg, "files"), "file");
207 char * id = v3_cfg_val(file_tree, "id");
208 char * index = v3_cfg_val(file_tree, "index");
209 int idx = atoi(index);
210 struct v3_cfg_file * file = NULL;
212 file = (struct v3_cfg_file *)V3_Malloc(sizeof(struct v3_cfg_file));
215 PrintError(VM_NONE, VCORE_NONE, "Could not allocate file structure\n");
216 v3_free_htable(cfg->file_table,0,0);
221 V3_Print(VM_NONE, VCORE_NONE, "File index=%d id=%s\n", idx, id);
223 strncpy(file->tag, id, V3_MAX_TAG_LEN);
226 struct file_hdr_v0 * hdr = &(files_v0->hdrs[idx]);
228 file->size = hdr->size;
229 file->data = cfg_blob + hdr->offset;
232 V3_Print(VM_NONE, VCORE_NONE, "Storing file data offset = %d, size=%d\n", (uint32_t)hdr->offset, hdr->size);
233 V3_Print(VM_NONE, VCORE_NONE, "file data at %p\n", file->data);
235 } else if (version==1) {
236 struct file_hdr_v1 * hdr = &(files_v1->hdrs[idx]);
239 file->size = hdr->size;
240 file->data = cfg_blob + hdr->offset;
241 file->hash = hdr->hash;
243 V3_Print(VM_NONE, VCORE_NONE, "Storing file data offset = %d, size=%d\n", (uint32_t)hdr->offset, hdr->size);
244 V3_Print(VM_NONE, VCORE_NONE, "file data at %p\n", file->data);
245 V3_Print(VM_NONE, VCORE_NONE, "Checking file data integrity...\n");
246 if ((hash = v3_hash_buffer(file->data, file->size)) != file->hash) {
247 PrintError(VM_NONE, VCORE_NONE, "File data corrupted! (orig hash=0x%lx, new=0x%lx\n",
251 V3_Print(VM_NONE, VCORE_NONE, "File data OK\n");
256 list_add( &(file->file_node), &(cfg->file_list));
258 V3_Print(VM_NONE, VCORE_NONE, "Keying file to name\n");
259 v3_htable_insert(cfg->file_table, (addr_t)(file->tag), (addr_t)(file));
261 V3_Print(VM_NONE, VCORE_NONE, "Iterating to next file\n");
263 file_tree = v3_cfg_next_branch(file_tree);
266 V3_Print(VM_NONE, VCORE_NONE, "Configuration parsed successfully\n");
272 static inline uint32_t get_alignment(char * align_str) {
273 // default is 4KB alignment
274 uint32_t alignment = PAGE_SIZE_4KB;
276 if (align_str != NULL) {
277 if (strcasecmp(align_str, "2MB") == 0) {
278 alignment = PAGE_SIZE_2MB;
279 } else if (strcasecmp(align_str, "4MB") == 0) {
280 alignment = PAGE_SIZE_4MB;
284 #ifndef V3_CONFIG_ALIGNED_PG_ALLOC
285 if (alignment != PAGE_SIZE_4KB) {
286 PrintError(VM_NONE, VCORE_NONE, "Aligned page allocations are not supported in this host (requested alignment=%d)\n", alignment);
287 PrintError(VM_NONE, VCORE_NONE, "Ignoring alignment request\n");
296 static int pre_config_vm(struct v3_vm_info * vm, v3_cfg_tree_t * vm_cfg) {
297 char * memory_str = v3_cfg_val(vm_cfg, "memory");
298 char * schedule_hz_str = v3_cfg_val(vm_cfg, "schedule_hz");
299 char * vm_class = v3_cfg_val(vm_cfg, "class");
300 char * align_str = v3_cfg_val(v3_cfg_subtree(vm_cfg, "memory"), "alignment");
301 uint32_t sched_hz = 100; // set the schedule frequency to 100 HZ
305 PrintError(VM_NONE, VCORE_NONE, "Memory is a required configuration parameter\n");
309 PrintDebug(VM_NONE, VCORE_NONE, "Memory=%s\n", memory_str);
311 PrintDebug(VM_NONE, VCORE_NONE, "Alignment=%s\n", align_str);
313 PrintDebug(VM_NONE, VCORE_NONE, "Alignment defaulted to 4KB.\n");
316 // Amount of ram the Guest will have, always in MB
317 vm->mem_size = (addr_t)atoi(memory_str) * 1024 * 1024;
318 vm->mem_align = get_alignment(align_str);
320 #ifdef V3_CONFIG_SWAPPING
321 if (v3_init_swapping_vm(vm,vm_cfg)) {
322 PrintError(vm,VCORE_NONE,"Unable to initialize swapping correctly\n");
325 if (vm->swap_state.enable_swapping) {
326 PrintDebug(vm,VCORE_NONE,"Swapping enabled\n");
328 PrintDebug(vm,VCORE_NONE,"Swapping disabled\n");
332 PrintDebug(VM_NONE, VCORE_NONE, "Alignment for %lu bytes of memory computed as 0x%x\n", vm->mem_size, vm->mem_align);
334 if (strcasecmp(vm_class, "PC") == 0) {
335 vm->vm_class = V3_PC_VM;
337 PrintError(VM_NONE, VCORE_NONE, "Invalid VM class\n");
341 #ifdef V3_CONFIG_TELEMETRY
343 char * telemetry = v3_cfg_val(vm_cfg, "telemetry");
345 // This should go first, because other subsystems will depend on the guest_info flag
346 if ((telemetry) && (strcasecmp(telemetry, "enable") == 0)) {
347 vm->enable_telemetry = 1;
349 vm->enable_telemetry = 0;
354 if (v3_init_vm(vm) == -1) {
355 PrintError(VM_NONE, VCORE_NONE, "Failed to initialize VM\n");
361 if (schedule_hz_str) {
362 sched_hz = atoi(schedule_hz_str);
365 PrintDebug(VM_NONE, VCORE_NONE, "CPU_KHZ = %d, schedule_freq=%p\n", V3_CPU_KHZ(),
366 (void *)(addr_t)sched_hz);
368 vm->yield_cycle_period = (V3_CPU_KHZ() * 1000) / sched_hz;
374 static int determine_paging_mode(struct guest_info * info, v3_cfg_tree_t * core_cfg) {
375 extern v3_cpu_arch_t v3_mach_type;
377 v3_cfg_tree_t * vm_tree = info->vm_info->cfg_data->cfg;
378 v3_cfg_tree_t * pg_tree = v3_cfg_subtree(vm_tree, "paging");
379 char * pg_mode = v3_cfg_val(pg_tree, "mode");
381 PrintDebug(info->vm_info, info, "Paging mode specified as %s\n", pg_mode);
384 if ((strcasecmp(pg_mode, "nested") == 0)) {
385 // we assume symmetric cores, so if core 0 has nested paging they all do
386 if ((v3_mach_type == V3_SVM_REV3_CPU) ||
387 (v3_mach_type == V3_VMX_EPT_CPU) ||
388 (v3_mach_type == V3_VMX_EPT_UG_CPU)) {
390 V3_Print(info->vm_info, info, "Setting paging mode to NESTED\n");
391 info->shdw_pg_mode = NESTED_PAGING;
393 PrintError(info->vm_info, info, "Nested paging not supported on this hardware. Defaulting to shadow paging\n");
394 info->shdw_pg_mode = SHADOW_PAGING;
396 } else if ((strcasecmp(pg_mode, "shadow") == 0)) {
397 V3_Print(info->vm_info, info, "Setting paging mode to SHADOW\n");
398 info->shdw_pg_mode = SHADOW_PAGING;
400 PrintError(info->vm_info, info, "Invalid paging mode (%s) specified in configuration. Defaulting to shadow paging\n", pg_mode);
401 info->shdw_pg_mode = SHADOW_PAGING;
404 V3_Print(info->vm_info, info, "No paging type specified in configuration. Defaulting to shadow paging\n");
405 info->shdw_pg_mode = SHADOW_PAGING;
409 if (v3_cfg_val(pg_tree, "large_pages") != NULL) {
410 if (strcasecmp(v3_cfg_val(pg_tree, "large_pages"), "true") == 0) {
411 info->use_large_pages = 1;
412 PrintDebug(info->vm_info, info, "Use of large pages in memory virtualization enabled.\n");
418 static int pre_config_core(struct guest_info * info, v3_cfg_tree_t * core_cfg) {
419 if (determine_paging_mode(info, core_cfg) != 0) {
423 if (v3_init_core(info) == -1) {
424 PrintError(info->vm_info, info, "Error Initializing Core\n");
428 if (info->vm_info->vm_class == V3_PC_VM) {
429 if (pre_config_pc_core(info, core_cfg) == -1) {
430 PrintError(info->vm_info, info, "PC Post configuration failure\n");
434 PrintError(info->vm_info, info, "Invalid VM Class\n");
443 static int post_config_vm(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
447 // Configure the memory map for the guest
448 if (setup_memory_map(vm, cfg) == -1) {
449 PrintError(vm, VCORE_NONE,"Setting up guest memory map failed...\n");
454 if (vm->vm_class == V3_PC_VM) {
455 if (post_config_pc(vm, cfg) == -1) {
456 PrintError(vm, VCORE_NONE,"PC Post configuration failure\n");
460 PrintError(vm, VCORE_NONE,"Invalid VM Class\n");
466 // Initialize fw_cfg state for VMM<->VM SEABIOS communication
467 if (v3_fw_cfg_init(vm) == -1) {
468 PrintError(vm, VCORE_NONE, "Error initializing Firmware Config (fw_cfg) state\n");
473 * Initialize configured devices
475 if (setup_devices(vm, cfg) == -1) {
476 PrintError(vm, VCORE_NONE,"Failed to setup devices\n");
481 // v3_print_io_map(info);
482 v3_print_msr_map(vm);
488 * Initialize configured extensions
490 if (setup_extensions(vm, cfg) == -1) {
491 PrintError(vm, VCORE_NONE,"Failed to setup extensions\n");
495 if (v3_setup_performance_tuning(vm, cfg) == -1) {
496 PrintError(vm, VCORE_NONE,"Failed to configure performance tuning parameters\n");
501 vm->run_state = VM_STOPPED;
508 static int post_config_core(struct guest_info * info, v3_cfg_tree_t * cfg) {
511 if (v3_init_core_extensions(info) == -1) {
512 PrintError(info->vm_info, info, "Error intializing extension core states\n");
516 if (info->vm_info->vm_class == V3_PC_VM) {
517 if (post_config_pc_core(info, cfg) == -1) {
518 PrintError(info->vm_info, info, "PC Post configuration failure\n");
522 PrintError(info->vm_info, info, "Invalid VM Class\n");
532 static struct v3_vm_info * allocate_guest(int num_cores) {
533 int guest_state_size = sizeof(struct v3_vm_info) + (sizeof(struct guest_info) * num_cores);
534 struct v3_vm_info * vm = V3_Malloc(guest_state_size);
537 PrintError(VM_NONE, VCORE_NONE, "Unable to allocate space for guest data structures\n");
543 memset(vm, 0, guest_state_size);
545 vm->num_cores = num_cores;
547 for (i = 0; i < num_cores; i++) {
548 vm->cores[i].core_run_state = CORE_INVALID;
551 vm->run_state = VM_INVALID;
558 struct v3_vm_info * v3_config_guest(void * cfg_blob, void * priv_data) {
559 extern v3_cpu_arch_t v3_mach_type;
560 struct v3_config * cfg_data = NULL;
561 struct v3_vm_info * vm = NULL;
564 v3_cfg_tree_t * cores_cfg = NULL;
565 v3_cfg_tree_t * per_core_cfg = NULL;
567 if (v3_mach_type == V3_INVALID_CPU) {
568 PrintError(VM_NONE, VCORE_NONE, "Configuring guest on invalid CPU\n");
572 cfg_data = parse_config(cfg_blob);
575 PrintError(VM_NONE, VCORE_NONE, "Could not parse configuration\n");
579 cores_cfg = v3_cfg_subtree(cfg_data->cfg, "cores");
582 PrintError(VM_NONE, VCORE_NONE, "Could not find core configuration (new config format required)\n");
586 num_cores = atoi(v3_cfg_val(cores_cfg, "count"));
587 if (num_cores == 0) {
588 PrintError(VM_NONE, VCORE_NONE, "No cores specified in configuration\n");
592 V3_Print(VM_NONE, VCORE_NONE, "Configuring %d cores\n", num_cores);
594 vm = allocate_guest(num_cores);
597 PrintError(VM_NONE, VCORE_NONE, "Could not allocate %d core guest\n", vm->num_cores);
601 vm->host_priv_data = priv_data;
603 vm->cfg_data = cfg_data;
605 V3_Print(vm, VCORE_NONE, "Preconfiguration\n");
607 if (pre_config_vm(vm, vm->cfg_data->cfg) == -1) {
608 PrintError(vm, VCORE_NONE, "Error in preconfiguration, attempting to free\n");
609 vm->run_state=VM_ERROR;
614 V3_Print(vm, VCORE_NONE, "Per core configuration\n");
615 per_core_cfg = v3_cfg_subtree(cores_cfg, "core");
617 // per core configuration
618 for (i = 0; i < vm->num_cores; i++) {
619 struct guest_info * info = &(vm->cores[i]);
623 info->core_cfg_data = per_core_cfg;
625 if (pre_config_core(info, per_core_cfg) == -1) {
626 PrintError(vm, VCORE_NONE, "Error in core %d preconfiguration, attempting to free guest\n", i);
627 vm->run_state=VM_ERROR;
633 per_core_cfg = v3_cfg_next_branch(per_core_cfg);
637 V3_Print(vm, VCORE_NONE, "Post Configuration\n");
639 if (post_config_vm(vm, vm->cfg_data->cfg) == -1) {
640 PrintError(vm, VCORE_NONE, "Error in postconfiguration, attempting to free guest\n");
641 vm->run_state=VM_ERROR;
647 per_core_cfg = v3_cfg_subtree(cores_cfg, "core");
649 // per core configuration
650 for (i = 0; i < vm->num_cores; i++) {
651 struct guest_info * info = &(vm->cores[i]);
653 post_config_core(info, per_core_cfg);
655 per_core_cfg = v3_cfg_next_branch(per_core_cfg);
658 V3_Print(vm, VCORE_NONE, "Configuration successfull\n");
665 int v3_free_config(struct v3_vm_info * vm) {
667 v3_free_htable(vm->cfg_data->file_table, 1, 0);
669 v3_xml_free(vm->cfg_data->cfg);
671 V3_Free(vm->cfg_data);
678 static int setup_memory_map(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
679 v3_cfg_tree_t * mem_region = v3_cfg_subtree(v3_cfg_subtree(cfg, "memmap"), "region");
682 addr_t start_addr = atox(v3_cfg_val(mem_region, "start"));
683 addr_t end_addr = atox(v3_cfg_val(mem_region, "end"));
684 addr_t host_addr = atox(v3_cfg_val(mem_region, "host_addr"));
687 if (v3_add_shadow_mem(vm, V3_MEM_CORE_ANY, start_addr, end_addr, host_addr) == -1) {
688 PrintError(vm, VCORE_NONE,"Could not map memory region: %p-%p => %p\n",
689 (void *)start_addr, (void *)end_addr, (void *)host_addr);
693 mem_region = v3_cfg_next_branch(mem_region);
700 static int setup_extensions(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
701 v3_cfg_tree_t * extension = v3_cfg_subtree(v3_cfg_subtree(cfg, "extensions"), "extension");
704 char * ext_name = v3_cfg_val(extension, "name");
707 PrintError(vm, VCORE_NONE, "Extension has no name\n");
711 V3_Print(vm, VCORE_NONE, "Configuring extension %s\n", ext_name);
713 if (v3_add_extension(vm, ext_name, extension) == -1) {
714 PrintError(vm, VCORE_NONE, "Error adding extension %s\n", ext_name);
718 extension = v3_cfg_next_branch(extension);
725 static int setup_devices(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
726 v3_cfg_tree_t * device = v3_cfg_subtree(v3_cfg_subtree(cfg, "devices"), "device");
730 char * dev_class = v3_cfg_val(device, "class");
732 V3_Print(vm, VCORE_NONE, "configuring device %s\n", dev_class);
734 if (v3_create_device(vm, dev_class, device) == -1) {
735 PrintError(vm, VCORE_NONE, "Error creating device %s\n", dev_class);
739 device = v3_cfg_next_branch(device);
742 v3_print_dev_mgr(vm);