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 MODULE_LICENSE("GPL");
40 int cpu_list[NR_CPUS] = {};
42 module_param_array(cpu_list, int, &cpu_list_len, 0644);
43 MODULE_PARM_DESC(cpu_list, "Comma-delimited list of CPUs that Palacios will run on");
45 static int allow_devmem = 0;
46 module_param(allow_devmem, int, 0);
47 MODULE_PARM_DESC(allow_devmem, "Allow general user-space /dev/mem access even if kernel is strict");
49 // Palacios options parameter
51 module_param(options, charp, 0);
52 MODULE_PARM_DESC(options, "Generic options to internal Palacios modules");
58 static int v3_major_num = 0;
60 static struct v3_guest * guest_map[MAX_VMS] = {[0 ... MAX_VMS - 1] = 0};
61 static struct proc_dir_entry * palacios_proc_dir = NULL;
63 struct class * v3_class = NULL;
64 static struct cdev ctrl_dev;
66 static int register_vm(struct v3_guest * guest) {
69 for (i = 0; i < MAX_VMS; i++) {
70 if (guest_map[i] == NULL) {
81 static long v3_dev_ioctl(struct file * filp,
82 unsigned int ioctl, unsigned long arg) {
83 void __user * argp = (void __user *)arg;
84 DEBUG("V3 IOCTL %d\n", ioctl);
88 case V3_CREATE_GUEST:{
90 struct v3_guest_img user_image;
91 struct v3_guest * guest = palacios_alloc(sizeof(struct v3_guest));
94 ERROR("Palacios: Error allocating Kernel guest_image\n");
98 memset(guest, 0, sizeof(struct v3_guest));
100 INFO("Palacios: Creating V3 Guest...\n");
102 vm_minor = register_vm(guest);
104 if (vm_minor == -1) {
105 ERROR("Palacios Error: Too many VMs are currently running\n");
109 guest->vm_dev = MKDEV(v3_major_num, vm_minor);
111 if (copy_from_user(&user_image, argp, sizeof(struct v3_guest_img))) {
112 ERROR("Palacios Error: copy from user error getting guest image...\n");
116 guest->img_size = user_image.size;
118 DEBUG("Palacios: Allocating kernel memory for guest image (%llu bytes)\n", user_image.size);
119 guest->img = palacios_valloc(guest->img_size);
122 ERROR("Palacios Error: Could not allocate space for guest image\n");
126 if (copy_from_user(guest->img, user_image.guest_data, guest->img_size)) {
127 ERROR("Palacios: Error loading guest data\n");
131 strncpy(guest->name, user_image.name, 127);
133 INIT_LIST_HEAD(&(guest->exts));
135 if (create_palacios_vm(guest) == -1) {
136 ERROR("Palacios: Error creating guest\n");
144 palacios_vfree(guest->img);
146 guest_map[vm_minor] = NULL;
148 palacios_free(guest);
154 case V3_FREE_GUEST: {
155 unsigned long vm_idx = arg;
156 struct v3_guest * guest;
158 if (vm_idx > MAX_VMS) {
159 ERROR("Invalid VM index: %ld\n", vm_idx);
163 guest = guest_map[vm_idx];
166 ERROR("No VM at index %ld\n",vm_idx);
170 INFO("Freeing VM (%s) (%p)\n", guest->name, guest);
172 if (free_palacios_vm(guest)<0) {
173 ERROR("Cannot free guest at index %ld\n",vm_idx);
177 guest_map[vm_idx] = NULL;
180 case V3_ADD_MEMORY: {
181 struct v3_mem_region mem;
183 memset(&mem, 0, sizeof(struct v3_mem_region));
185 if (copy_from_user(&mem, argp, sizeof(struct v3_mem_region))) {
186 ERROR("copy from user error getting mem_region...\n");
190 DEBUG("Adding %llu pages to Palacios memory\n", mem.num_pages);
192 if (add_palacios_memory(&mem) == -1) {
193 ERROR("Error adding memory to Palacios\n");
200 case V3_REMOVE_MEMORY: {
201 struct v3_mem_region mem;
203 memset(&mem, 0, sizeof(struct v3_mem_region));
205 if (copy_from_user(&mem, argp, sizeof(struct v3_mem_region))) {
206 ERROR("copy from user error getting mem_region...\n");
210 DEBUG("Removing memory at address %p\n", (void*)(mem.base_addr));
212 if (remove_palacios_memory(&mem) == -1) {
213 ERROR("Error removing memory from Palacios\n");
222 case V3_RESET_MEMORY: {
223 DEBUG("Resetting memory\n");
224 if (palacios_deinit_mm() == -1) {
225 ERROR("Error deiniting the Palacios memory manager\n");
228 if (palacios_init_mm()) {
229 ERROR("Error initing the Palacios memory manager\n");
236 struct global_ctrl * ctrl = get_global_ctrl(ioctl);
239 return ctrl->handler(ioctl, arg);
242 WARNING("\tUnhandled global ctrl cmd: %d\n", ioctl);
253 static struct file_operations v3_ctrl_fops = {
254 .owner = THIS_MODULE,
255 .unlocked_ioctl = v3_dev_ioctl,
256 .compat_ioctl = v3_dev_ioctl,
261 struct proc_dir_entry *palacios_get_procdir(void)
263 // INFO("Returning procdir=%p\n",palacios_proc_dir);
264 return palacios_proc_dir;
268 #define MAX_VCORES 256
269 #define MAX_REGIONS 1024
273 static int read_guests_details(struct seq_file *s, void *v)
277 uint64_t num_vcores, num_regions;
278 struct v3_vm_base_state *base=0;
279 struct v3_vm_core_state *core=0;
280 struct v3_vm_mem_state *mem=0;
282 base = palacios_alloc(sizeof(struct v3_vm_base_state));
285 ERROR("No space for base state structure\n");
289 for(i = 0; i < MAX_VMS; i++) {
291 if (guest_map[i] != NULL) {
293 v3_get_state_sizes_vm(guest_map[i]->v3_ctx,&num_vcores,&num_regions);
295 core = palacios_alloc(sizeof(struct v3_vm_core_state) + num_vcores*sizeof(struct v3_vm_vcore_state));
298 ERROR("No space for core state structure\n");
302 mem = palacios_alloc(sizeof(struct v3_vm_mem_state) + num_regions*sizeof(struct v3_vm_mem_region));
305 ERROR("No space for memory state structure\n");
310 "---------------------------------------------------------------------------------------\n");
314 "Device: /dev/v3-vm%d\n",
315 i,guest_map[i]->name, i);
318 core->num_vcores=num_vcores;
319 mem->num_regions=num_regions;
321 if (v3_get_state_vm(guest_map[i]->v3_ctx, base, core, mem)) {
322 ERROR("Cannot get VM info\n");
323 seq_printf(s, "<unable to get data for this VM>\n");
330 "Memsize: %llu (%llu ROS)\n\n",
331 base->vm_type==V3_VM_GENERAL ? "general" :
332 base->vm_type==V3_VM_HVM ? "HVM" : "UNKNOWN",
333 base->state==V3_VM_INVALID ? "INVALID" :
334 base->state==V3_VM_RUNNING ? "running" :
335 base->state==V3_VM_STOPPED ? "stopped" :
336 base->state==V3_VM_PAUSED ? "paused" :
337 base->state==V3_VM_ERROR ? "ERROR" :
338 base->state==V3_VM_SIMULATING ? "simulating" :
339 base->state==V3_VM_RESETTING ? "resetting" : "UNKNOWN",
345 seq_printf(s, "Core States\n");
347 for (j=0;j<core->num_vcores;j++) {
349 " vcore %u %s on pcore %lu %llu exits rip=0x%p %s %s %s %s\n",
351 core->vcore[j].state==V3_VCORE_INVALID ? "INVALID" :
352 core->vcore[j].state==V3_VCORE_RUNNING ? "running" :
353 core->vcore[j].state==V3_VCORE_STOPPED ? "stopped" :
354 core->vcore[j].state==V3_VCORE_RESETTING ? "resetting" : "UNKNOWN",
355 core->vcore[j].pcore,
356 core->vcore[j].num_exits,
357 core->vcore[j].last_rip,
358 core->vcore[j].cpu_mode==V3_VCORE_CPU_REAL ? "real" :
359 core->vcore[j].cpu_mode==V3_VCORE_CPU_PROTECTED ? "protected" :
360 core->vcore[j].cpu_mode==V3_VCORE_CPU_PROTECTED_PAE ? "protectedpae" :
361 core->vcore[j].cpu_mode==V3_VCORE_CPU_LONG ? "long" :
362 core->vcore[j].cpu_mode==V3_VCORE_CPU_LONG_32_COMPAT ? "long32" :
363 core->vcore[j].cpu_mode==V3_VCORE_CPU_LONG_16_COMPAT ? "long16" : "UNKNOWN",
364 core->vcore[j].mem_mode==V3_VCORE_MEM_MODE_PHYSICAL ? "physical" :
365 core->vcore[j].mem_mode==V3_VCORE_MEM_MODE_VIRTUAL ? "virtual" : "UNKNOWN",
366 core->vcore[j].mem_state==V3_VCORE_MEM_STATE_SHADOW ? "shadow" :
367 core->vcore[j].mem_state==V3_VCORE_MEM_STATE_NESTED ? "nested" : "UNKNOWN",
368 core->vcore[j].vcore_type==V3_VCORE_GENERAL ? "" :
369 core->vcore[j].vcore_type==V3_VCORE_ROS ? "ros" :
370 core->vcore[j].vcore_type==V3_VCORE_HRT ? "hrt" : "UNKNOWN");
373 seq_printf(s, "\nMemory Regions\n");
374 for (j=0;j<mem->num_regions;j++) {
375 seq_printf(s," region %u has HPAs 0x%016llx-0x%016llx (node %d) GPA 0x%016llx %s %s\n",
376 j, (uint64_t)mem->region[j].host_paddr, (uint64_t)mem->region[j].host_paddr+mem->region[j].size,
377 numa_addr_to_node((uintptr_t)(mem->region[j].host_paddr)),
378 (uint64_t)mem->region[j].guest_paddr,
379 mem->region[j].swapped ? "swapped" : "",
380 mem->region[j].pinned ? "pinned" : "");
384 "---------------------------------------------------------------------------------------\n");
386 palacios_free(mem); mem=0;
387 palacios_free(core); core=0;
395 if (mem) { palacios_free(mem); }
396 if (core) { palacios_free(core); }
397 if (base) { palacios_free(base); }
402 static int read_guests(struct seq_file *s, void *v)
405 struct v3_vm_base_state *base=0;
406 struct v3_vm_core_state *core=0;
407 struct v3_vm_mem_state *mem=0;
409 base = palacios_alloc(sizeof(struct v3_vm_base_state));
412 ERROR("No space for base state structure\n");
416 core = palacios_alloc(sizeof(struct v3_vm_core_state) + MAX_VCORES*sizeof(struct v3_vm_vcore_state));
419 ERROR("No space for core state structure\n");
423 mem = palacios_alloc(sizeof(struct v3_vm_mem_state) + MAX_REGIONS*sizeof(struct v3_vm_mem_region));
426 ERROR("No space for memory state structure\n");
430 for(i = 0; i < MAX_VMS; i++) {
431 if (guest_map[i] != NULL) {
432 seq_printf(s,"%s\t/dev/v3-vm%d", guest_map[i]->name, i);
434 core->num_vcores=MAX_VCORES; // max we can handle
435 mem->num_regions=MAX_REGIONS; // max we can handle
437 if (v3_get_state_vm(guest_map[i]->v3_ctx, base, core, mem)) {
438 ERROR("Cannot get VM info\n");
439 seq_printf(s, "\t<unable to get data for this VM>\n");
441 seq_printf(s,"\t%s\t%llu vcores\t%llu regions\t%llu mem\t%s\n",
442 base->state==V3_VM_INVALID ? "INVALID" :
443 base->state==V3_VM_RUNNING ? "running" :
444 base->state==V3_VM_STOPPED ? "stopped" :
445 base->state==V3_VM_PAUSED ? "paused" :
446 base->state==V3_VM_ERROR ? "ERROR" :
447 base->state==V3_VM_SIMULATING ? "simulating" : "UNKNOWN",
451 base->vm_type == V3_VM_GENERAL ? "general" :
452 base->vm_type == V3_VM_HVM ? "hvm" : "UNKNOWN");
459 if (mem) { palacios_free(mem); }
460 if (core) { palacios_free(core); }
461 if (base) { palacios_free(base); }
467 static int guests_short_proc_open(struct inode * inode, struct file * filp)
469 struct proc_dir_entry * proc_entry = PDE(inode);
470 return single_open(filp, read_guests, proc_entry->data);
473 static int guests_full_proc_open(struct inode * inode, struct file * filp)
475 struct proc_dir_entry * proc_entry = PDE(inode);
476 return single_open(filp, read_guests_details, proc_entry->data);
482 static struct file_operations guest_full_proc_ops = {
483 .owner = THIS_MODULE,
484 .open = guests_full_proc_open,
487 .release = single_release,
490 static struct file_operations guest_short_proc_ops = {
491 .owner = THIS_MODULE,
492 .open = guests_short_proc_open,
495 .release = single_release,
498 // Supply basic information that the user-space tools need
499 // to manipulate Palacios. The current use case here is to
500 // convey memory information
501 static int read_info(struct seq_file *s, void *v)
503 uint64_t mem_block_size;
506 seq_printf(s,"kernel MAX_ORDER:\t%d\n",MAX_ORDER);
507 seq_printf(s,"number of nodes:\t%d\n", numa_num_nodes());
508 seq_printf(s,"number of cpus: \t%d\n", num_online_cpus());
509 seq_printf(s,"\npalacios compiled mem_block_size:\t%d\n", V3_CONFIG_MEM_BLOCK_SIZE);
510 if (!v3_lookup_option("mem_block_size")) {
511 mem_block_size = V3_CONFIG_MEM_BLOCK_SIZE;
513 if (strict_strtoull(v3_lookup_option("mem_block_size"), 0, &mem_block_size)) {
518 seq_printf(s,"palacios run-time mem_block_size:\t%llu\n", mem_block_size);
520 seq_printf(s,"\nCPU to node mappings\n");
521 for (i=0;i<num_online_cpus();i++) {
522 seq_printf(s,"cpu %d -> node %d\n", i, numa_cpu_to_node(i));
523 if (numa_cpu_to_node(i)>max_node) {
524 max_node=numa_cpu_to_node(i);
527 seq_printf(s,"\nNode to node distances\n");
528 for (j=0;j<=max_node;j++) {
529 seq_printf(s," \t%2d", j);
532 for (i=0;i<=max_node;i++) {
533 seq_printf(s,"%2d ",i);
534 for (j=0;j<=max_node;j++) {
535 seq_printf(s,"\t%2d", numa_get_distance(i,j));
539 seq_printf(s,"\nCPU to CPU distances\n");
540 for (j=0;j<num_online_cpus();j++) {
541 seq_printf(s," \t%2d", j);
544 for (i=0;i<num_online_cpus();i++) {
545 seq_printf(s,"%2d ",i);
546 for (j=0;j<num_online_cpus();j++) {
547 seq_printf(s,"\t%2d", numa_get_distance(numa_cpu_to_node(i),numa_cpu_to_node(j)));
554 static int info_proc_open(struct inode * inode, struct file * filp)
556 struct proc_dir_entry * proc_entry = PDE(inode);
557 return single_open(filp, read_info, proc_entry->data);
562 static struct file_operations info_proc_ops = {
563 .owner = THIS_MODULE,
564 .open = info_proc_open,
567 .release = single_release,
571 static int __init v3_init(void) {
573 dev_t dev = MKDEV(0, 0); // We dynamicallly assign the major number
579 palacios_proc_dir = proc_mkdir("v3vee", NULL);
580 if (!palacios_proc_dir) {
581 ERROR("Could not create proc entry\n");
586 // this will populate the v3vee tree...
587 if (palacios_init_mm()) {
592 palacios_allow_devmem();
595 // numa is now a required interface and we need it
596 // up before primary initiatilization
597 palacios_init_numa();
599 // Initialize Palacios
600 palacios_vmm_init(options);
602 // initialize extensions
603 init_lnx_extensions();
606 v3_class = class_create(THIS_MODULE, "vms");
607 if (!v3_class || IS_ERR(v3_class)) {
608 ERROR("Failed to register V3 VM device class\n");
609 ret = PTR_ERR(v3_class);
613 INFO("intializing V3 Control device\n");
615 ret = alloc_chrdev_region(&dev, 0, MAX_VMS + 1, "v3vee");
618 ERROR("Error registering device region for V3 devices\n");
622 v3_major_num = MAJOR(dev);
624 dev = MKDEV(v3_major_num, MAX_VMS + 1);
627 DEBUG("Creating V3 Control device: Major %d, Minor %d\n", v3_major_num, MINOR(dev));
628 cdev_init(&ctrl_dev, &v3_ctrl_fops);
629 ctrl_dev.owner = THIS_MODULE;
630 ctrl_dev.ops = &v3_ctrl_fops;
631 cdev_add(&ctrl_dev, dev, 1);
633 device_create(v3_class, NULL, dev, NULL, "v3vee");
636 ERROR("Error adding v3 control device\n");
641 struct proc_dir_entry *entry;
643 entry = create_proc_entry("v3-guests", 0444, palacios_proc_dir);
645 entry->proc_fops = &guest_short_proc_ops;
646 INFO("/proc/v3vee/v3-guests successfully created\n");
648 ERROR("Could not create proc entry\n");
651 entry = create_proc_entry("v3-guests-details", 0444, palacios_proc_dir);
653 entry->proc_fops = &guest_full_proc_ops;
654 INFO("/proc/v3vee/v3-guests-details successfully created\n");
656 ERROR("Could not create proc entry\n");
660 entry = create_proc_entry("v3-info", 0444, palacios_proc_dir);
662 entry->proc_fops = &info_proc_ops;
663 INFO("/proc/v3vee/v3-info successfully created\n");
665 ERROR("Could not create proc entry\n");
675 remove_proc_entry("v3-guests-details", palacios_proc_dir);
677 remove_proc_entry("v3-guests", palacios_proc_dir);
679 device_destroy(v3_class, dev);
681 unregister_chrdev_region(MKDEV(v3_major_num, 0), MAX_VMS + 1);
683 class_destroy(v3_class);
686 palacios_restore_devmem();
688 palacios_deinit_mm();
690 remove_proc_entry("v3vee", NULL);
699 static void __exit v3_exit(void) {
700 extern u32 pg_allocs;
707 struct v3_guest * guest;
711 /* Stop and free any running VMs */
712 for (i = 0; i < MAX_VMS; i++) {
713 if (guest_map[i] != NULL) {
714 guest = (struct v3_guest *)(guest_map[i]);
716 if (!guest->v3_ctx) {
717 ERROR("Orphan VM detected and skipped: index=%d name=%s\n", i, guest->name);
721 if (v3_stop_vm(guest->v3_ctx) < 0)
722 ERROR("Couldn't stop VM %d\n", i);
724 free_palacios_vm(guest);
729 dev = MKDEV(v3_major_num, MAX_VMS + 1);
731 INFO("Removing V3 Control device\n");
736 palacios_deinit_numa();
738 DEBUG("Palacios Mallocs = %d, Frees = %d\n", mallocs, frees);
739 DEBUG("Palacios Vmallocs = %d, Vfrees = %d\n", vmallocs, vfrees);
740 DEBUG("Palacios Page Allocs = %d, Page Frees = %d\n", pg_allocs, pg_frees);
742 unregister_chrdev_region(MKDEV(v3_major_num, 0), MAX_VMS + 1);
746 device_destroy(v3_class, dev);
747 class_destroy(v3_class);
750 deinit_lnx_extensions();
753 palacios_restore_devmem();
756 palacios_deinit_mm();
758 remove_proc_entry("v3-info", palacios_proc_dir);
759 remove_proc_entry("v3-guests-details", palacios_proc_dir);
760 remove_proc_entry("v3-guests", palacios_proc_dir);
761 remove_proc_entry("v3vee", NULL);
763 DEBUG("Palacios Module Mallocs = %d, Frees = %d\n", mod_allocs, mod_frees);
771 module_init(v3_init);
772 module_exit(v3_exit);
776 void * trace_malloc(size_t size, gfp_t flags) {
780 addr = palacios_alloc_extended(size, flags, -1);
786 void trace_free(const void * objp) {
788 palacios_free((void*)objp);