2 Palacios main control interface
6 #include <linux/kernel.h>
7 #include <linux/module.h>
8 #include <linux/moduleparam.h>
9 #include <linux/errno.h>
10 #include <linux/percpu.h>
12 #include <linux/uaccess.h>
13 #include <linux/device.h>
14 #include <linux/cdev.h>
18 #include <linux/file.h>
19 #include <linux/spinlock.h>
20 #include <linux/kthread.h>
22 #include <linux/proc_fs.h>
23 #include <linux/seq_file.h>
25 #include <palacios/vmm.h>
31 #include "allow_devmem.h"
33 #include "lockcheck.h"
35 #include "linux-exts.h"
37 #include "util-hashtable.h"
39 MODULE_LICENSE("GPL");
42 int cpu_list[NR_CPUS] = {};
44 module_param_array(cpu_list, int, &cpu_list_len, 0644);
45 MODULE_PARM_DESC(cpu_list, "Comma-delimited list of CPUs that Palacios will run on");
47 static int allow_devmem = 0;
48 module_param(allow_devmem, int, 0);
49 MODULE_PARM_DESC(allow_devmem, "Allow general user-space /dev/mem access even if kernel is strict");
51 // Palacios options parameter
53 module_param(options, charp, 0);
54 MODULE_PARM_DESC(options, "Generic options to internal Palacios modules");
60 static int v3_major_num = 0;
62 static struct v3_guest * guest_map[MAX_VMS] = {[0 ... MAX_VMS - 1] = 0};
63 static struct proc_dir_entry * palacios_proc_dir = NULL;
65 struct class * v3_class = NULL;
66 static struct cdev ctrl_dev;
69 // mapping from thread ids to their resource control blocks
70 struct hashtable *v3_thread_resource_map=0;
72 static int register_vm(struct v3_guest * guest) {
75 for (i = 0; i < MAX_VMS; i++) {
76 if (guest_map[i] == NULL) {
87 static long v3_dev_ioctl(struct file * filp,
88 unsigned int ioctl, unsigned long arg) {
89 void __user * argp = (void __user *)arg;
90 DEBUG("V3 IOCTL %d\n", ioctl);
94 case V3_CREATE_GUEST:{
96 struct v3_guest_img user_image;
97 struct v3_guest * guest = palacios_alloc(sizeof(struct v3_guest));
100 ERROR("Palacios: Error allocating Kernel guest_image\n");
104 memset(guest, 0, sizeof(struct v3_guest));
106 INFO("Palacios: Creating V3 Guest...\n");
108 vm_minor = register_vm(guest);
110 if (vm_minor == -1) {
111 ERROR("Palacios Error: Too many VMs are currently running\n");
115 guest->vm_dev = MKDEV(v3_major_num, vm_minor);
117 if (copy_from_user(&user_image, argp, sizeof(struct v3_guest_img))) {
118 ERROR("Palacios Error: copy from user error getting guest image...\n");
122 guest->img_size = user_image.size;
124 DEBUG("Palacios: Allocating kernel memory for guest image (%llu bytes)\n", user_image.size);
125 guest->img = palacios_valloc(guest->img_size);
128 ERROR("Palacios Error: Could not allocate space for guest image\n");
132 if (copy_from_user(guest->img, user_image.guest_data, guest->img_size)) {
133 ERROR("Palacios: Error loading guest data\n");
137 strncpy(guest->name, user_image.name, 127);
139 INIT_LIST_HEAD(&(guest->exts));
141 if (create_palacios_vm(guest) == -1) {
142 ERROR("Palacios: Error creating guest\n");
150 palacios_vfree(guest->img);
152 guest_map[vm_minor] = NULL;
154 palacios_free(guest);
160 case V3_FREE_GUEST: {
161 unsigned long vm_idx = arg;
162 struct v3_guest * guest;
164 if (vm_idx > MAX_VMS) {
165 ERROR("Invalid VM index: %ld\n", vm_idx);
169 guest = guest_map[vm_idx];
172 ERROR("No VM at index %ld\n",vm_idx);
176 INFO("Freeing VM (%s) (%p)\n", guest->name, guest);
178 if (free_palacios_vm(guest)<0) {
179 ERROR("Cannot free guest at index %ld\n",vm_idx);
183 guest_map[vm_idx] = NULL;
186 case V3_ADD_MEMORY: {
187 struct v3_mem_region mem;
189 memset(&mem, 0, sizeof(struct v3_mem_region));
191 if (copy_from_user(&mem, argp, sizeof(struct v3_mem_region))) {
192 ERROR("copy from user error getting mem_region...\n");
196 DEBUG("Adding %llu pages to Palacios memory\n", mem.num_pages);
198 if (add_palacios_memory(&mem) == -1) {
199 ERROR("Error adding memory to Palacios\n");
206 case V3_REMOVE_MEMORY: {
207 struct v3_mem_region mem;
209 memset(&mem, 0, sizeof(struct v3_mem_region));
211 if (copy_from_user(&mem, argp, sizeof(struct v3_mem_region))) {
212 ERROR("copy from user error getting mem_region...\n");
216 DEBUG("Removing memory at address %p\n", (void*)(mem.base_addr));
218 if (remove_palacios_memory(&mem) == -1) {
219 ERROR("Error removing memory from Palacios\n");
228 case V3_RESET_MEMORY: {
229 DEBUG("Resetting memory\n");
230 if (palacios_deinit_mm() == -1) {
231 ERROR("Error deiniting the Palacios memory manager\n");
234 if (palacios_init_mm()) {
235 ERROR("Error initing the Palacios memory manager\n");
242 struct global_ctrl * ctrl = get_global_ctrl(ioctl);
245 return ctrl->handler(ioctl, arg);
248 WARNING("\tUnhandled global ctrl cmd: %d\n", ioctl);
259 static struct file_operations v3_ctrl_fops = {
260 .owner = THIS_MODULE,
261 .unlocked_ioctl = v3_dev_ioctl,
262 .compat_ioctl = v3_dev_ioctl,
267 struct proc_dir_entry *palacios_get_procdir(void)
269 // INFO("Returning procdir=%p\n",palacios_proc_dir);
270 return palacios_proc_dir;
274 #define MAX_CORES 1024
275 #define MAX_REGIONS 1024
276 #define MIN(x,y) ((x)<(y) ? (x) : (y))
278 static int read_guests_details(struct seq_file *s, void *v)
282 uint64_t num_vcores, num_regions;
283 uint64_t alloc_num_vcores, alloc_num_regions;
284 struct v3_vm_base_state *base=0;
285 struct v3_vm_core_state *core=0;
286 struct v3_vm_mem_state *mem=0;
289 base = palacios_valloc(sizeof(struct v3_vm_base_state));
293 ERROR("No space for base state structure\n");
298 for(i = 0; i < MAX_VMS; i++) {
300 if (guest_map[i] != NULL) {
302 v3_get_state_sizes_vm(guest_map[i]->v3_ctx,&num_vcores,&num_regions);
304 alloc_num_vcores = MIN(num_vcores,MAX_CORES);
305 alloc_num_regions = MIN(num_regions,MAX_REGIONS);
307 core = palacios_valloc(sizeof(struct v3_vm_core_state) + alloc_num_vcores*sizeof(struct v3_vm_vcore_state));
310 ERROR("No space for core state structure\n");
314 mem = palacios_valloc(sizeof(struct v3_vm_mem_state) + alloc_num_regions*sizeof(struct v3_vm_mem_region));
317 ERROR("No space for memory state structure\n");
322 "---------------------------------------------------------------------------------------\n");
326 "Device: /dev/v3-vm%d\n",
327 i,guest_map[i]->name, i);
330 core->num_vcores=alloc_num_vcores;
331 mem->num_regions=alloc_num_regions;
333 if (v3_get_state_vm(guest_map[i]->v3_ctx, base, core, mem)) {
334 ERROR("Cannot get VM info\n");
335 seq_printf(s, "<unable to get data for this VM>\n");
340 "Cores: %llu (%llu shown)\n"
341 "Regions: %llu (%llu shown)\n"
342 "Memsize: %llu (%llu ROS)\n\n",
343 base->vm_type==V3_VM_GENERAL ? "general" :
344 base->vm_type==V3_VM_HVM ? "HVM" : "UNKNOWN",
345 base->state==V3_VM_INVALID ? "INVALID" :
346 base->state==V3_VM_RUNNING ? "running" :
347 base->state==V3_VM_STOPPED ? "stopped" :
348 base->state==V3_VM_PAUSED ? "paused" :
349 base->state==V3_VM_ERROR ? "ERROR" :
350 base->state==V3_VM_SIMULATING ? "simulating" :
351 base->state==V3_VM_RESETTING ? "resetting" : "UNKNOWN",
359 seq_printf(s, "Core States\n");
361 for (j=0;j<core->num_vcores;j++) {
363 " vcore %u %s on pcore %lu %llu exits rip=0x%p %s %s %s %s\n",
365 core->vcore[j].state==V3_VCORE_INVALID ? "INVALID" :
366 core->vcore[j].state==V3_VCORE_RUNNING ? "running" :
367 core->vcore[j].state==V3_VCORE_STOPPED ? "stopped" :
368 core->vcore[j].state==V3_VCORE_RESETTING ? "resetting" : "UNKNOWN",
369 core->vcore[j].pcore,
370 core->vcore[j].num_exits,
371 core->vcore[j].last_rip,
372 core->vcore[j].cpu_mode==V3_VCORE_CPU_REAL ? "real" :
373 core->vcore[j].cpu_mode==V3_VCORE_CPU_PROTECTED ? "protected" :
374 core->vcore[j].cpu_mode==V3_VCORE_CPU_PROTECTED_PAE ? "protectedpae" :
375 core->vcore[j].cpu_mode==V3_VCORE_CPU_LONG ? "long" :
376 core->vcore[j].cpu_mode==V3_VCORE_CPU_LONG_32_COMPAT ? "long32" :
377 core->vcore[j].cpu_mode==V3_VCORE_CPU_LONG_16_COMPAT ? "long16" : "UNKNOWN",
378 core->vcore[j].mem_mode==V3_VCORE_MEM_MODE_PHYSICAL ? "physical" :
379 core->vcore[j].mem_mode==V3_VCORE_MEM_MODE_VIRTUAL ? "virtual" : "UNKNOWN",
380 core->vcore[j].mem_state==V3_VCORE_MEM_STATE_SHADOW ? "shadow" :
381 core->vcore[j].mem_state==V3_VCORE_MEM_STATE_NESTED ? "nested" : "UNKNOWN",
382 core->vcore[j].vcore_type==V3_VCORE_GENERAL ? "" :
383 core->vcore[j].vcore_type==V3_VCORE_ROS ? "ros" :
384 core->vcore[j].vcore_type==V3_VCORE_HRT ? "hrt" : "UNKNOWN");
388 seq_printf(s, "\nMemory Regions\n");
389 for (j=0;j<mem->num_regions;j++) {
390 seq_printf(s," region %u has HPAs 0x%016llx-0x%016llx (node %d) GPA 0x%016llx %s %s\n",
391 j, (uint64_t)mem->region[j].host_paddr, (uint64_t)mem->region[j].host_paddr+mem->region[j].size,
392 numa_addr_to_node((uintptr_t)(mem->region[j].host_paddr)),
393 (uint64_t)mem->region[j].guest_paddr,
394 mem->region[j].swapped ? "swapped" : "",
395 mem->region[j].pinned ? "pinned" : "");
400 "---------------------------------------------------------------------------------------\n");
402 palacios_vfree(mem); mem=0;
403 palacios_vfree(core); core=0;
411 if (mem) { palacios_vfree(mem); }
412 if (core) { palacios_vfree(core); }
413 if (base) { palacios_vfree(base); }
418 static int read_guests(struct seq_file *s, void *v)
421 struct v3_vm_base_state *base=0;
422 struct v3_vm_core_state *core=0;
423 struct v3_vm_mem_state *mem=0;
424 uint64_t num_vcores, num_regions;
427 INFO("READ GUEST\n");
429 base = palacios_valloc(sizeof(struct v3_vm_base_state));
432 ERROR("No space for base state structure\n");
436 core = palacios_valloc(sizeof(struct v3_vm_core_state));
439 ERROR("No space for core state structure\n");
443 mem = palacios_valloc(sizeof(struct v3_vm_mem_state));
446 ERROR("No space for memory state structure\n");
451 for(i = 0; i < MAX_VMS; i++) {
452 if (guest_map[i] != NULL) {
454 v3_get_state_sizes_vm(guest_map[i]->v3_ctx,&num_vcores,&num_regions);
456 seq_printf(s,"%s\t/dev/v3-vm%d", guest_map[i]->name, i);
458 // Skip getting per core and per-region
462 if (v3_get_state_vm(guest_map[i]->v3_ctx, base, core, mem)) {
463 ERROR("Cannot get VM info\n");
464 seq_printf(s, "\t<unable to get data for this VM>\n");
466 seq_printf(s,"\t%s\t%llu vcores\t%llu regions\t%llu mem\t%s\n",
467 base->state==V3_VM_INVALID ? "INVALID" :
468 base->state==V3_VM_RUNNING ? "running" :
469 base->state==V3_VM_STOPPED ? "stopped" :
470 base->state==V3_VM_PAUSED ? "paused" :
471 base->state==V3_VM_ERROR ? "ERROR" :
472 base->state==V3_VM_SIMULATING ? "simulating" : "UNKNOWN",
476 base->vm_type == V3_VM_GENERAL ? "general" :
477 base->vm_type == V3_VM_HVM ? "hvm" : "UNKNOWN");
484 if (mem) { palacios_vfree(mem); }
485 if (core) { palacios_vfree(core); }
486 if (base) { palacios_vfree(base); }
492 static int guests_short_proc_open(struct inode * inode, struct file * filp)
494 struct proc_dir_entry * proc_entry = PDE(inode);
495 return single_open(filp, read_guests, proc_entry->data);
498 static int guests_full_proc_open(struct inode * inode, struct file * filp)
500 struct proc_dir_entry * proc_entry = PDE(inode);
501 return single_open(filp, read_guests_details, proc_entry->data);
507 static struct file_operations guest_full_proc_ops = {
508 .owner = THIS_MODULE,
509 .open = guests_full_proc_open,
512 .release = single_release,
515 static struct file_operations guest_short_proc_ops = {
516 .owner = THIS_MODULE,
517 .open = guests_short_proc_open,
520 .release = single_release,
523 // Supply basic information that the user-space tools need
524 // to manipulate Palacios. The current use case here is to
525 // convey memory information
526 static int read_info(struct seq_file *s, void *v)
528 uint64_t mem_block_size;
531 seq_printf(s,"kernel MAX_ORDER:\t%d\n",MAX_ORDER);
532 seq_printf(s,"number of nodes:\t%d\n", numa_num_nodes());
533 seq_printf(s,"number of cpus: \t%d\n", num_online_cpus());
534 seq_printf(s,"\npalacios compiled mem_block_size:\t%d\n", V3_CONFIG_MEM_BLOCK_SIZE);
535 if (!v3_lookup_option("mem_block_size")) {
536 mem_block_size = V3_CONFIG_MEM_BLOCK_SIZE;
538 if (strict_strtoull(v3_lookup_option("mem_block_size"), 0, &mem_block_size)) {
543 seq_printf(s,"palacios run-time mem_block_size:\t%llu\n", mem_block_size);
545 seq_printf(s,"\nCPU to node mappings\n");
546 for (i=0;i<num_online_cpus();i++) {
547 seq_printf(s,"cpu %d -> node %d\n", i, numa_cpu_to_node(i));
548 if (numa_cpu_to_node(i)>max_node) {
549 max_node=numa_cpu_to_node(i);
552 seq_printf(s,"\nNode to node distances\n");
553 for (j=0;j<=max_node;j++) {
554 seq_printf(s," \t%2d", j);
557 for (i=0;i<=max_node;i++) {
558 seq_printf(s,"%2d ",i);
559 for (j=0;j<=max_node;j++) {
560 seq_printf(s,"\t%2d", numa_get_distance(i,j));
564 seq_printf(s,"\nCPU to CPU distances\n");
565 for (j=0;j<num_online_cpus();j++) {
566 seq_printf(s," \t%2d", j);
569 for (i=0;i<num_online_cpus();i++) {
570 seq_printf(s,"%2d ",i);
571 for (j=0;j<num_online_cpus();j++) {
572 seq_printf(s,"\t%2d", numa_get_distance(numa_cpu_to_node(i),numa_cpu_to_node(j)));
579 static int info_proc_open(struct inode * inode, struct file * filp)
581 struct proc_dir_entry * proc_entry = PDE(inode);
582 return single_open(filp, read_info, proc_entry->data);
587 static struct file_operations info_proc_ops = {
588 .owner = THIS_MODULE,
589 .open = info_proc_open,
592 .release = single_release,
596 static inline uint_t thr_hash_func(addr_t key)
598 return palacios_hash_long((long)key,64);
601 static inline int thr_hash_comp(addr_t k1, addr_t k2)
606 static int __init v3_init(void) {
608 dev_t dev = MKDEV(0, 0); // We dynamicallly assign the major number
615 if (!(v3_thread_resource_map = palacios_create_htable(MAX_THREADS,thr_hash_func,thr_hash_comp))) {
616 ERROR("Could not create thread/resource map\n");
621 palacios_proc_dir = proc_mkdir("v3vee", NULL);
622 if (!palacios_proc_dir) {
623 ERROR("Could not create proc entry\n");
628 // this will populate the v3vee tree...
629 if (palacios_init_mm()) {
634 palacios_allow_devmem();
637 // numa is now a required interface and we need it
638 // up before primary initiatilization
639 palacios_init_numa();
641 // Initialize Palacios
642 palacios_vmm_init(options);
644 // initialize extensions
645 init_lnx_extensions();
648 v3_class = class_create(THIS_MODULE, "vms");
649 if (!v3_class || IS_ERR(v3_class)) {
650 ERROR("Failed to register V3 VM device class\n");
651 ret = PTR_ERR(v3_class);
655 INFO("intializing V3 Control device\n");
657 ret = alloc_chrdev_region(&dev, 0, MAX_VMS + 1, "v3vee");
660 ERROR("Error registering device region for V3 devices\n");
664 v3_major_num = MAJOR(dev);
666 dev = MKDEV(v3_major_num, MAX_VMS + 1);
669 DEBUG("Creating V3 Control device: Major %d, Minor %d\n", v3_major_num, MINOR(dev));
670 cdev_init(&ctrl_dev, &v3_ctrl_fops);
671 ctrl_dev.owner = THIS_MODULE;
672 ctrl_dev.ops = &v3_ctrl_fops;
673 cdev_add(&ctrl_dev, dev, 1);
675 device_create(v3_class, NULL, dev, NULL, "v3vee");
678 ERROR("Error adding v3 control device\n");
683 struct proc_dir_entry *entry;
685 entry = create_proc_entry("v3-guests", 0444, palacios_proc_dir);
687 entry->proc_fops = &guest_short_proc_ops;
688 INFO("/proc/v3vee/v3-guests successfully created\n");
690 ERROR("Could not create proc entry\n");
693 entry = create_proc_entry("v3-guests-details", 0444, palacios_proc_dir);
695 entry->proc_fops = &guest_full_proc_ops;
696 INFO("/proc/v3vee/v3-guests-details successfully created\n");
698 ERROR("Could not create proc entry\n");
702 entry = create_proc_entry("v3-info", 0444, palacios_proc_dir);
704 entry->proc_fops = &info_proc_ops;
705 INFO("/proc/v3vee/v3-info successfully created\n");
707 ERROR("Could not create proc entry\n");
717 remove_proc_entry("v3-guests-details", palacios_proc_dir);
719 remove_proc_entry("v3-guests", palacios_proc_dir);
721 device_destroy(v3_class, dev);
723 unregister_chrdev_region(MKDEV(v3_major_num, 0), MAX_VMS + 1);
725 class_destroy(v3_class);
728 palacios_restore_devmem();
730 palacios_deinit_mm();
732 remove_proc_entry("v3vee", NULL);
734 palacios_free_htable(v3_thread_resource_map,0,0);
743 static void __exit v3_exit(void) {
744 extern u32 pg_allocs;
751 struct v3_guest * guest;
755 /* Stop and free any running VMs */
756 for (i = 0; i < MAX_VMS; i++) {
757 if (guest_map[i] != NULL) {
758 guest = (struct v3_guest *)(guest_map[i]);
760 if (!guest->v3_ctx) {
761 ERROR("Orphan VM detected and skipped: index=%d name=%s\n", i, guest->name);
765 if (v3_stop_vm(guest->v3_ctx) < 0)
766 ERROR("Couldn't stop VM %d\n", i);
768 free_palacios_vm(guest);
773 dev = MKDEV(v3_major_num, MAX_VMS + 1);
775 INFO("Removing V3 Control device\n");
780 palacios_deinit_numa();
782 DEBUG("Palacios Mallocs = %d, Frees = %d\n", mallocs, frees);
783 DEBUG("Palacios Vmallocs = %d, Vfrees = %d\n", vmallocs, vfrees);
784 DEBUG("Palacios Page Allocs = %d, Page Frees = %d\n", pg_allocs, pg_frees);
786 unregister_chrdev_region(MKDEV(v3_major_num, 0), MAX_VMS + 1);
790 device_destroy(v3_class, dev);
791 class_destroy(v3_class);
794 deinit_lnx_extensions();
797 palacios_restore_devmem();
800 palacios_deinit_mm();
802 remove_proc_entry("v3-info", palacios_proc_dir);
803 remove_proc_entry("v3-guests-details", palacios_proc_dir);
804 remove_proc_entry("v3-guests", palacios_proc_dir);
805 remove_proc_entry("v3vee", NULL);
807 DEBUG("Palacios Module Mallocs = %d, Frees = %d\n", mod_allocs, mod_frees);
809 palacios_free_htable(v3_thread_resource_map,0,0);
817 module_init(v3_init);
818 module_exit(v3_exit);
822 void * trace_malloc(size_t size, gfp_t flags) {
826 addr = palacios_alloc_extended(size, flags, -1);
832 void trace_free(const void * objp) {
834 palacios_free((void*)objp);