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>
39 #include <palacios/vmm_host_events.h>
40 #include <palacios/vmm_perftune.h>
42 #include "vmm_config_class.h"
45 /* The Palacios cookie encodes "v3vee" followed by a
46 3 byte version code. There are currently two versions:
48 \0\0\0 => original (no checksum)
52 #define COOKIE_V0 "v3vee\0\0\0"
53 #define COOKIE_V1 "v3vee\0\0\1"
58 // This is used to access the configuration file index table
73 struct file_idx_table_v0 {
75 struct file_hdr_v0 hdrs[0];
78 struct file_idx_table_v1 {
80 struct file_hdr_v1 hdrs[0];
86 static int setup_memory_map(struct v3_vm_info * vm, v3_cfg_tree_t * cfg);
87 static int setup_extensions(struct v3_vm_info * vm, v3_cfg_tree_t * cfg);
88 static int setup_devices(struct v3_vm_info * vm, v3_cfg_tree_t * cfg);
92 char * v3_cfg_val(v3_cfg_tree_t * tree, char * tag) {
93 char * attrib = (char *)v3_xml_attr(tree, tag);
94 v3_cfg_tree_t * child_entry = v3_xml_child(tree, tag);
97 if ((child_entry != NULL) && (attrib != NULL)) {
98 PrintError(VM_NONE, VCORE_NONE, "Duplicate Configuration parameters present for %s\n", tag);
102 if (attrib == NULL) {
103 val = v3_xml_txt(child_entry);
114 v3_cfg_tree_t * v3_cfg_subtree(v3_cfg_tree_t * tree, char * tag) {
115 return v3_xml_child(tree, tag);
118 v3_cfg_tree_t * v3_cfg_next_branch(v3_cfg_tree_t * tree) {
119 return v3_xml_next(tree);
124 struct v3_cfg_file * v3_cfg_get_file(struct v3_vm_info * vm, char * tag) {
125 struct v3_cfg_file * file = NULL;
127 file = (struct v3_cfg_file *)v3_htable_search(vm->cfg_data->file_table, (addr_t)tag);
133 static uint_t file_hash_fn(addr_t key) {
134 char * name = (char *)key;
135 return v3_hash_buffer((uchar_t *)name, strlen(name));
138 static int file_eq_fn(addr_t key1, addr_t key2) {
139 char * name1 = (char *)key1;
140 char * name2 = (char *)key2;
142 return (strcmp(name1, name2) == 0);
145 static struct v3_config * parse_config(void * cfg_blob) {
146 struct v3_config * cfg = NULL;
149 struct file_idx_table_v0 * files_v0 = NULL;
150 struct file_idx_table_v1 * files_v1 = NULL;
151 v3_cfg_tree_t * file_tree = NULL;
154 V3_Print(VM_NONE, VCORE_NONE, "cfg data at %p\n", cfg_blob);
156 if (memcmp(cfg_blob, COOKIE_V0, COOKIE_LEN) == 0) {
158 } else if (memcmp(cfg_blob, COOKIE_V1, COOKIE_LEN) == 0) {
161 PrintError(VM_NONE, VCORE_NONE, "Invalid Configuration Header Or Unknown Version\n");
165 V3_Print(VM_NONE, VCORE_NONE, "Handling Palacios Image Format, Version 0x%x\n",version);
167 offset += COOKIE_LEN;
169 cfg = (struct v3_config *)V3_Malloc(sizeof(struct v3_config));
172 PrintError(VM_NONE, VCORE_NONE, "Unable to allocate while parsing\n");
176 memset(cfg, 0, sizeof(struct v3_config));
178 cfg->blob = cfg_blob;
179 INIT_LIST_HEAD(&(cfg->file_list));
180 cfg->file_table = v3_create_htable(0, file_hash_fn, file_eq_fn);
182 if (!(cfg->file_table)) {
183 PrintError(VM_NONE, VCORE_NONE, "Unable to allocate hash table while parsing\n");
188 xml_len = *(uint32_t *)(cfg_blob + offset);
191 cfg->cfg = (v3_cfg_tree_t *)v3_xml_parse((uint8_t *)(cfg_blob + offset));
196 // This is hideous, but the file formats are still very close
198 files_v0 = (struct file_idx_table_v0 *)(cfg_blob + offset);
199 V3_Print(VM_NONE, VCORE_NONE, "Number of files in cfg: %d\n", (uint32_t)(files_v0->num_files));
201 files_v1 = (struct file_idx_table_v1 *)(cfg_blob + offset);
202 V3_Print(VM_NONE, VCORE_NONE, "Number of files in cfg: %d\n", (uint32_t)(files_v1->num_files));
206 file_tree = v3_cfg_subtree(v3_cfg_subtree(cfg->cfg, "files"), "file");
209 char * id = v3_cfg_val(file_tree, "id");
210 char * index = v3_cfg_val(file_tree, "index");
211 int idx = atoi(index);
212 struct v3_cfg_file * file = NULL;
214 file = (struct v3_cfg_file *)V3_Malloc(sizeof(struct v3_cfg_file));
217 PrintError(VM_NONE, VCORE_NONE, "Could not allocate file structure\n");
218 v3_free_htable(cfg->file_table,0,0);
223 V3_Print(VM_NONE, VCORE_NONE, "File index=%d id=%s\n", idx, id);
225 strncpy(file->tag, id, V3_MAX_TAG_LEN);
228 struct file_hdr_v0 * hdr = &(files_v0->hdrs[idx]);
230 file->size = hdr->size;
231 file->data = cfg_blob + hdr->offset;
234 V3_Print(VM_NONE, VCORE_NONE, "Storing file data offset = %d, size=%d\n", (uint32_t)hdr->offset, hdr->size);
235 V3_Print(VM_NONE, VCORE_NONE, "file data at %p\n", file->data);
237 } else if (version==1) {
238 struct file_hdr_v1 * hdr = &(files_v1->hdrs[idx]);
241 file->size = hdr->size;
242 file->data = cfg_blob + hdr->offset;
243 file->hash = hdr->hash;
245 V3_Print(VM_NONE, VCORE_NONE, "Storing file data offset = %d, size=%d\n", (uint32_t)hdr->offset, hdr->size);
246 V3_Print(VM_NONE, VCORE_NONE, "file data at %p\n", file->data);
247 V3_Print(VM_NONE, VCORE_NONE, "Checking file data integrity...\n");
248 if ((hash = v3_hash_buffer(file->data, file->size)) != file->hash) {
249 PrintError(VM_NONE, VCORE_NONE, "File data corrupted! (orig hash=0x%lx, new=0x%lx\n",
253 V3_Print(VM_NONE, VCORE_NONE, "File data OK\n");
258 list_add( &(file->file_node), &(cfg->file_list));
260 V3_Print(VM_NONE, VCORE_NONE, "Keying file to name\n");
261 v3_htable_insert(cfg->file_table, (addr_t)(file->tag), (addr_t)(file));
263 V3_Print(VM_NONE, VCORE_NONE, "Iterating to next file\n");
265 file_tree = v3_cfg_next_branch(file_tree);
268 V3_Print(VM_NONE, VCORE_NONE, "Configuration parsed successfully\n");
274 static inline uint32_t get_alignment(char * align_str) {
275 // default is 4KB alignment
276 uint32_t alignment = PAGE_SIZE_4KB;
278 if (align_str != NULL) {
279 if (strcasecmp(align_str, "2MB") == 0) {
280 alignment = PAGE_SIZE_2MB;
281 } else if (strcasecmp(align_str, "4MB") == 0) {
282 alignment = PAGE_SIZE_4MB;
286 #ifndef V3_CONFIG_ALIGNED_PG_ALLOC
287 if (alignment != PAGE_SIZE_4KB) {
288 PrintError(VM_NONE, VCORE_NONE, "Aligned page allocations are not supported in this host (requested alignment=%d)\n", alignment);
289 PrintError(VM_NONE, VCORE_NONE, "Ignoring alignment request\n");
297 static int pre_config_vm(struct v3_vm_info * vm, v3_cfg_tree_t * vm_cfg) {
298 char * memory_str = v3_cfg_val(vm_cfg, "memory");
299 char * schedule_hz_str = v3_cfg_val(vm_cfg, "schedule_hz");
300 char * vm_class = v3_cfg_val(vm_cfg, "class");
301 char * align_str = v3_cfg_val(v3_cfg_subtree(vm_cfg, "memory"), "alignment");
302 uint32_t sched_hz = 100; // set the schedule frequency to 100 HZ
306 PrintError(VM_NONE, VCORE_NONE, "Memory is a required configuration parameter\n");
310 PrintDebug(VM_NONE, VCORE_NONE, "Memory=%s\n", memory_str);
312 PrintDebug(VM_NONE, VCORE_NONE, "Alignment=%s\n", align_str);
314 PrintDebug(VM_NONE, VCORE_NONE, "Alignment defaulted to 4KB.\n");
317 // Amount of ram the Guest will have, always in MB
318 vm->mem_size = (addr_t)atoi(memory_str) * 1024 * 1024;
319 vm->mem_align = get_alignment(align_str);
322 PrintDebug(VM_NONE, VCORE_NONE, "Alignment for %lu bytes of memory computed as 0x%x\n", vm->mem_size, vm->mem_align);
324 if (strcasecmp(vm_class, "PC") == 0) {
325 vm->vm_class = V3_PC_VM;
327 PrintError(VM_NONE, VCORE_NONE, "Invalid VM class\n");
331 #ifdef V3_CONFIG_TELEMETRY
333 char * telemetry = v3_cfg_val(vm_cfg, "telemetry");
335 // This should go first, because other subsystems will depend on the guest_info flag
336 if ((telemetry) && (strcasecmp(telemetry, "enable") == 0)) {
337 vm->enable_telemetry = 1;
339 vm->enable_telemetry = 0;
344 if (v3_init_vm(vm) == -1) {
345 PrintError(VM_NONE, VCORE_NONE, "Failed to initialize VM\n");
351 if (schedule_hz_str) {
352 sched_hz = atoi(schedule_hz_str);
355 PrintDebug(VM_NONE, VCORE_NONE, "CPU_KHZ = %d, schedule_freq=%p\n", V3_CPU_KHZ(),
356 (void *)(addr_t)sched_hz);
358 vm->yield_cycle_period = (V3_CPU_KHZ() * 1000) / sched_hz;
364 static int determine_paging_mode(struct guest_info * info, v3_cfg_tree_t * core_cfg) {
365 extern v3_cpu_arch_t v3_mach_type;
367 v3_cfg_tree_t * vm_tree = info->vm_info->cfg_data->cfg;
368 v3_cfg_tree_t * pg_tree = v3_cfg_subtree(vm_tree, "paging");
369 char * pg_mode = v3_cfg_val(pg_tree, "mode");
371 PrintDebug(info->vm_info, info, "Paging mode specified as %s\n", pg_mode);
374 if ((strcasecmp(pg_mode, "nested") == 0)) {
375 // we assume symmetric cores, so if core 0 has nested paging they all do
376 if ((v3_mach_type == V3_SVM_REV3_CPU) ||
377 (v3_mach_type == V3_VMX_EPT_CPU) ||
378 (v3_mach_type == V3_VMX_EPT_UG_CPU)) {
380 V3_Print(info->vm_info, info, "Setting paging mode to NESTED\n");
381 info->shdw_pg_mode = NESTED_PAGING;
383 PrintError(info->vm_info, info, "Nested paging not supported on this hardware. Defaulting to shadow paging\n");
384 info->shdw_pg_mode = SHADOW_PAGING;
386 } else if ((strcasecmp(pg_mode, "shadow") == 0)) {
387 V3_Print(info->vm_info, info, "Setting paging mode to SHADOW\n");
388 info->shdw_pg_mode = SHADOW_PAGING;
390 PrintError(info->vm_info, info, "Invalid paging mode (%s) specified in configuration. Defaulting to shadow paging\n", pg_mode);
391 info->shdw_pg_mode = SHADOW_PAGING;
394 V3_Print(info->vm_info, info, "No paging type specified in configuration. Defaulting to shadow paging\n");
395 info->shdw_pg_mode = SHADOW_PAGING;
399 if (v3_cfg_val(pg_tree, "large_pages") != NULL) {
400 if (strcasecmp(v3_cfg_val(pg_tree, "large_pages"), "true") == 0) {
401 info->use_large_pages = 1;
402 PrintDebug(info->vm_info, info, "Use of large pages in memory virtualization enabled.\n");
408 static int pre_config_core(struct guest_info * info, v3_cfg_tree_t * core_cfg) {
409 if (determine_paging_mode(info, core_cfg) != 0) {
413 if (v3_init_core(info) == -1) {
414 PrintError(info->vm_info, info, "Error Initializing Core\n");
418 if (info->vm_info->vm_class == V3_PC_VM) {
419 if (pre_config_pc_core(info, core_cfg) == -1) {
420 PrintError(info->vm_info, info, "PC Post configuration failure\n");
424 PrintError(info->vm_info, info, "Invalid VM Class\n");
433 static int post_config_vm(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
437 // Configure the memory map for the guest
438 if (setup_memory_map(vm, cfg) == -1) {
439 PrintError(vm, VCORE_NONE,"Setting up guest memory map failed...\n");
444 if (vm->vm_class == V3_PC_VM) {
445 if (post_config_pc(vm, cfg) == -1) {
446 PrintError(vm, VCORE_NONE,"PC Post configuration failure\n");
450 PrintError(vm, VCORE_NONE,"Invalid VM Class\n");
456 * Initialize configured devices
458 if (setup_devices(vm, cfg) == -1) {
459 PrintError(vm, VCORE_NONE,"Failed to setup devices\n");
464 // v3_print_io_map(info);
465 v3_print_msr_map(vm);
471 * Initialize configured extensions
473 if (setup_extensions(vm, cfg) == -1) {
474 PrintError(vm, VCORE_NONE,"Failed to setup extensions\n");
478 if (v3_setup_performance_tuning(vm, cfg) == -1) {
479 PrintError(vm, VCORE_NONE,"Failed to configure performance tuning parameters\n");
484 vm->run_state = VM_STOPPED;
491 static int post_config_core(struct guest_info * info, v3_cfg_tree_t * cfg) {
494 if (v3_init_core_extensions(info) == -1) {
495 PrintError(info->vm_info, info, "Error intializing extension core states\n");
499 if (info->vm_info->vm_class == V3_PC_VM) {
500 if (post_config_pc_core(info, cfg) == -1) {
501 PrintError(info->vm_info, info, "PC Post configuration failure\n");
505 PrintError(info->vm_info, info, "Invalid VM Class\n");
515 static struct v3_vm_info * allocate_guest(int num_cores) {
516 int guest_state_size = sizeof(struct v3_vm_info) + (sizeof(struct guest_info) * num_cores);
517 struct v3_vm_info * vm = V3_Malloc(guest_state_size);
520 PrintError(VM_NONE, VCORE_NONE, "Unable to allocate space for guest data structures\n");
526 memset(vm, 0, guest_state_size);
528 vm->num_cores = num_cores;
530 for (i = 0; i < num_cores; i++) {
531 vm->cores[i].core_run_state = CORE_INVALID;
534 vm->run_state = VM_INVALID;
541 struct v3_vm_info * v3_config_guest(void * cfg_blob, void * priv_data) {
542 extern v3_cpu_arch_t v3_mach_type;
543 struct v3_config * cfg_data = NULL;
544 struct v3_vm_info * vm = NULL;
547 v3_cfg_tree_t * cores_cfg = NULL;
548 v3_cfg_tree_t * per_core_cfg = NULL;
550 if (v3_mach_type == V3_INVALID_CPU) {
551 PrintError(VM_NONE, VCORE_NONE, "Configuring guest on invalid CPU\n");
555 cfg_data = parse_config(cfg_blob);
558 PrintError(VM_NONE, VCORE_NONE, "Could not parse configuration\n");
562 cores_cfg = v3_cfg_subtree(cfg_data->cfg, "cores");
565 PrintError(VM_NONE, VCORE_NONE, "Could not find core configuration (new config format required)\n");
569 num_cores = atoi(v3_cfg_val(cores_cfg, "count"));
570 if (num_cores == 0) {
571 PrintError(VM_NONE, VCORE_NONE, "No cores specified in configuration\n");
575 V3_Print(VM_NONE, VCORE_NONE, "Configuring %d cores\n", num_cores);
577 vm = allocate_guest(num_cores);
580 PrintError(VM_NONE, VCORE_NONE, "Could not allocate %d core guest\n", vm->num_cores);
584 vm->host_priv_data = priv_data;
586 vm->cfg_data = cfg_data;
588 V3_Print(vm, VCORE_NONE, "Preconfiguration\n");
590 if (pre_config_vm(vm, vm->cfg_data->cfg) == -1) {
591 PrintError(vm, VCORE_NONE, "Error in preconfiguration\n");
595 V3_Print(vm, VCORE_NONE, "Per core configuration\n");
596 per_core_cfg = v3_cfg_subtree(cores_cfg, "core");
598 // per core configuration
599 for (i = 0; i < vm->num_cores; i++) {
600 struct guest_info * info = &(vm->cores[i]);
604 info->core_cfg_data = per_core_cfg;
606 if (pre_config_core(info, per_core_cfg) == -1) {
607 PrintError(vm, VCORE_NONE, "Error in core %d preconfiguration\n", i);
612 per_core_cfg = v3_cfg_next_branch(per_core_cfg);
616 V3_Print(vm, VCORE_NONE, "Post Configuration\n");
618 if (post_config_vm(vm, vm->cfg_data->cfg) == -1) {
619 PrintError(vm, VCORE_NONE, "Error in postconfiguration\n");
624 per_core_cfg = v3_cfg_subtree(cores_cfg, "core");
626 // per core configuration
627 for (i = 0; i < vm->num_cores; i++) {
628 struct guest_info * info = &(vm->cores[i]);
630 post_config_core(info, per_core_cfg);
632 per_core_cfg = v3_cfg_next_branch(per_core_cfg);
635 V3_Print(vm, VCORE_NONE, "Configuration successfull\n");
642 int v3_free_config(struct v3_vm_info * vm) {
644 v3_free_htable(vm->cfg_data->file_table, 1, 0);
646 v3_xml_free(vm->cfg_data->cfg);
648 V3_Free(vm->cfg_data);
655 static int setup_memory_map(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
656 v3_cfg_tree_t * mem_region = v3_cfg_subtree(v3_cfg_subtree(cfg, "memmap"), "region");
659 addr_t start_addr = atox(v3_cfg_val(mem_region, "start"));
660 addr_t end_addr = atox(v3_cfg_val(mem_region, "end"));
661 addr_t host_addr = atox(v3_cfg_val(mem_region, "host_addr"));
664 if (v3_add_shadow_mem(vm, V3_MEM_CORE_ANY, start_addr, end_addr, host_addr) == -1) {
665 PrintError(vm, VCORE_NONE,"Could not map memory region: %p-%p => %p\n",
666 (void *)start_addr, (void *)end_addr, (void *)host_addr);
670 mem_region = v3_cfg_next_branch(mem_region);
677 static int setup_extensions(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
678 v3_cfg_tree_t * extension = v3_cfg_subtree(v3_cfg_subtree(cfg, "extensions"), "extension");
681 char * ext_name = v3_cfg_val(extension, "name");
684 PrintError(vm, VCORE_NONE, "Extension has no name\n");
688 V3_Print(vm, VCORE_NONE, "Configuring extension %s\n", ext_name);
690 if (v3_add_extension(vm, ext_name, extension) == -1) {
691 PrintError(vm, VCORE_NONE, "Error adding extension %s\n", ext_name);
695 extension = v3_cfg_next_branch(extension);
702 static int setup_devices(struct v3_vm_info * vm, v3_cfg_tree_t * cfg) {
703 v3_cfg_tree_t * device = v3_cfg_subtree(v3_cfg_subtree(cfg, "devices"), "device");
707 char * dev_class = v3_cfg_val(device, "class");
709 V3_Print(vm, VCORE_NONE, "configuring device %s\n", dev_class);
711 if (v3_create_device(vm, dev_class, device) == -1) {
712 PrintError(vm, VCORE_NONE, "Error creating device %s\n", dev_class);
716 device = v3_cfg_next_branch(device);
719 v3_print_dev_mgr(vm);