drivers/xen/balloon.c

   1 /******************************************************************************
   2  * balloon.c
   3  *
   4  * Xen balloon driver - enables returning/claiming memory to/from Xen.
   5  *
   6  * Copyright (c) 2003, B Dragovic
   7  * Copyright (c) 2003-2004, M Williamson, K Fraser
   8  * Copyright (c) 2005 Dan M. Smith, IBM Corporation
   9  *
  10  * This program is free software; you can redistribute it and/or
  11  * modify it under the terms of the GNU General Public License version 2
  12  * as published by the Free Software Foundation; or, when distributed
  13  * separately from the Linux kernel or incorporated into other
  14  * software packages, subject to the following license:
  15  *
  16  * Permission is hereby granted, free of charge, to any person obtaining a copy
  17  * of this source file (the "Software"), to deal in the Software without
  18  * restriction, including without limitation the rights to use, copy, modify,
  19  * merge, publish, distribute, sublicense, and/or sell copies of the Software,
  20  * and to permit persons to whom the Software is furnished to do so, subject to
  21  * the following conditions:
  22  *
  23  * The above copyright notice and this permission notice shall be included in
  24  * all copies or substantial portions of the Software.
  25  *
  26  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  27  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  28  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  29  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  30  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  31  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  32  * IN THE SOFTWARE.
  33  */
  34
  35 #include <linux/kernel.h>
  36 #include <linux/module.h>
  37 #include <linux/sched.h>
  38 #include <linux/errno.h>
  39 #include <linux/mm.h>
  40 #include <linux/bootmem.h>
  41 #include <linux/pagemap.h>
  42 #include <linux/highmem.h>
  43 #include <linux/mutex.h>
  44 #include <linux/list.h>
  45 #include <linux/sysdev.h>
  46
  47 #include <asm/page.h>
  48 #include <asm/pgalloc.h>
  49 #include <asm/pgtable.h>
  50 #include <asm/uaccess.h>
  51 #include <asm/tlb.h>
  52
  53 #include <asm/xen/hypervisor.h>
  54 #include <asm/xen/hypercall.h>
  55 #include <xen/interface/xen.h>
  56 #include <xen/interface/memory.h>
  57 #include <xen/xenbus.h>
  58 #include <xen/features.h>
  59 #include <xen/page.h>
  60
  61 #define PAGES2KB(_p) ((_p)<<(PAGE_SHIFT-10))
  62
  63 #define BALLOON_CLASS_NAME "xen_memory"
  64
  65 struct balloon_stats {
  66         /* We aim for 'current allocation' == 'target allocation'. */
  67         unsigned long current_pages;
  68         unsigned long target_pages;
  69         /* We may hit the hard limit in Xen. If we do then we remember it. */
  70         unsigned long hard_limit;
  71         /*
  72          * Drivers may alter the memory reservation independently, but they
  73          * must inform the balloon driver so we avoid hitting the hard limit.
  74          */
  75         unsigned long driver_pages;
  76         /* Number of pages in high- and low-memory balloons. */
  77         unsigned long balloon_low;
  78         unsigned long balloon_high;
  79 };
  80
  81 static DEFINE_MUTEX(balloon_mutex);
  82
  83 static struct sys_device balloon_sysdev;
  84
  85 static int register_balloon(struct sys_device *sysdev);
  86
  87 /*
  88  * Protects atomic reservation decrease/increase against concurrent increases.
  89  * Also protects non-atomic updates of current_pages and driver_pages, and
  90  * balloon lists.
  91  */
  92 static DEFINE_SPINLOCK(balloon_lock);
  93
  94 static struct balloon_stats balloon_stats;
  95
  96 /* We increase/decrease in batches which fit in a page */
  97 static unsigned long frame_list[PAGE_SIZE / sizeof(unsigned long)];
  98
  99 /* VM /proc information for memory */
 100 extern unsigned long totalram_pages;
 101
 102 #ifdef CONFIG_HIGHMEM
 103 extern unsigned long totalhigh_pages;
 104 #define inc_totalhigh_pages() (totalhigh_pages++)
 105 #define dec_totalhigh_pages() (totalhigh_pages--)
 106 #else
 107 #define inc_totalhigh_pages() do {} while(0)
 108 #define dec_totalhigh_pages() do {} while(0)
 109 #endif
 110
 111 /* List of ballooned pages, threaded through the mem_map array. */
 112 static LIST_HEAD(ballooned_pages);
 113
 114 /* Main work function, always executed in process context. */
 115 static void balloon_process(struct work_struct *work);
 116 static DECLARE_WORK(balloon_worker, balloon_process);
 117 static struct timer_list balloon_timer;
 118
 119 /* When ballooning out (allocating memory to return to Xen) we don't really
 120    want the kernel to try too hard since that can trigger the oom killer. */
 121 #define GFP_BALLOON \
 122         (GFP_HIGHUSER | __GFP_NOWARN | __GFP_NORETRY | __GFP_NOMEMALLOC)
 123
 124 static void scrub_page(struct page *page)
 125 {
 126 #ifdef CONFIG_XEN_SCRUB_PAGES
 127         clear_highpage(page);
 128 #endif
 129 }
 130
 131 /* balloon_append: add the given page to the balloon. */
 132 static void balloon_append(struct page *page)
 133 {
 134         /* Lowmem is re-populated first, so highmem pages go at list tail. */
 135         if (PageHighMem(page)) {
 136                 list_add_tail(&page->lru, &ballooned_pages);
 137                 balloon_stats.balloon_high++;
 138                 dec_totalhigh_pages();
 139         } else {
 140                 list_add(&page->lru, &ballooned_pages);
 141                 balloon_stats.balloon_low++;
 142         }
 143 }
 144
 145 /* balloon_retrieve: rescue a page from the balloon, if it is not empty. */
 146 static struct page *balloon_retrieve(void)
 147 {
 148         struct page *page;
 149
 150         if (list_empty(&ballooned_pages))
 151                 return NULL;
 152
 153         page = list_entry(ballooned_pages.next, struct page, lru);
 154         list_del(&page->lru);
 155
 156         if (PageHighMem(page)) {
 157                 balloon_stats.balloon_high--;
 158                 inc_totalhigh_pages();
 159         }
 160         else
 161                 balloon_stats.balloon_low--;
 162
 163         return page;
 164 }
 165
 166 static struct page *balloon_first_page(void)
 167 {
 168         if (list_empty(&ballooned_pages))
 169                 return NULL;
 170         return list_entry(ballooned_pages.next, struct page, lru);
 171 }
 172
 173 static struct page *balloon_next_page(struct page *page)
 174 {
 175         struct list_head *next = page->lru.next;
 176         if (next == &ballooned_pages)
 177                 return NULL;
 178         return list_entry(next, struct page, lru);
 179 }
 180
 181 static void balloon_alarm(unsigned long unused)
 182 {
 183         schedule_work(&balloon_worker);
 184 }
 185
 186 static unsigned long current_target(void)
 187 {
 188         unsigned long target = min(balloon_stats.target_pages, balloon_stats.hard_limit);
 189
 190         target = min(target,
 191                      balloon_stats.current_pages +
 192                      balloon_stats.balloon_low +
 193                      balloon_stats.balloon_high);
 194
 195         return target;
 196 }
 197
 198 static int increase_reservation(unsigned long nr_pages)
 199 {
 200         unsigned long  pfn, i, flags;
 201         struct page   *page;
 202         long           rc;
 203         struct xen_memory_reservation reservation = {
 204                 .address_bits = 0,
 205                 .extent_order = 0,
 206                 .domid        = DOMID_SELF
 207         };
 208
 209         if (nr_pages > ARRAY_SIZE(frame_list))
 210                 nr_pages = ARRAY_SIZE(frame_list);
 211
 212         spin_lock_irqsave(&balloon_lock, flags);
 213
 214         page = balloon_first_page();
 215         for (i = 0; i < nr_pages; i++) {
 216                 BUG_ON(page == NULL);
 217                 frame_list[i] = page_to_pfn(page);;
 218                 page = balloon_next_page(page);
 219         }
 220
 221         set_xen_guest_handle(reservation.extent_start, frame_list);
 222         reservation.nr_extents = nr_pages;
 223         rc = HYPERVISOR_memory_op(XENMEM_populate_physmap, &reservation);
 224         if (rc < nr_pages) {
 225                 if (rc > 0) {
 226                         int ret;
 227
 228                         /* We hit the Xen hard limit: reprobe. */
 229                         reservation.nr_extents = rc;
 230                         ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation,
 231                                                    &reservation);
 232                         BUG_ON(ret != rc);
 233                 }
 234                 if (rc >= 0)
 235                         balloon_stats.hard_limit = (balloon_stats.current_pages + rc -
 236                                                     balloon_stats.driver_pages);
 237                 goto out;
 238         }
 239
 240         for (i = 0; i < nr_pages; i++) {
 241                 page = balloon_retrieve();
 242                 BUG_ON(page == NULL);
 243
 244                 pfn = page_to_pfn(page);
 245                 BUG_ON(!xen_feature(XENFEAT_auto_translated_physmap) &&
 246                        phys_to_machine_mapping_valid(pfn));
 247
 248                 set_phys_to_machine(pfn, frame_list[i]);
 249
 250                 /* Link back into the page tables if not highmem. */
 251                 if (pfn < max_low_pfn) {
 252                         int ret;
 253                         ret = HYPERVISOR_update_va_mapping(
 254                                 (unsigned long)__va(pfn << PAGE_SHIFT),
 255                                 mfn_pte(frame_list[i], PAGE_KERNEL),
 256                                 0);
 257                         BUG_ON(ret);
 258                 }
 259
 260                 /* Relinquish the page back to the allocator. */
 261                 ClearPageReserved(page);
 262                 init_page_count(page);
 263                 __free_page(page);
 264         }
 265
 266         balloon_stats.current_pages += nr_pages;
 267         totalram_pages = balloon_stats.current_pages;
 268
 269  out:
 270         spin_unlock_irqrestore(&balloon_lock, flags);
 271
 272         return 0;
 273 }
 274
 275 static int decrease_reservation(unsigned long nr_pages)
 276 {
 277         unsigned long  pfn, i, flags;
 278         struct page   *page;
 279         int            need_sleep = 0;
 280         int ret;
 281         struct xen_memory_reservation reservation = {
 282                 .address_bits = 0,
 283                 .extent_order = 0,
 284                 .domid        = DOMID_SELF
 285         };
 286
 287         if (nr_pages > ARRAY_SIZE(frame_list))
 288                 nr_pages = ARRAY_SIZE(frame_list);
 289
 290         for (i = 0; i < nr_pages; i++) {
 291                 if ((page = alloc_page(GFP_BALLOON)) == NULL) {
 292                         nr_pages = i;
 293                         need_sleep = 1;
 294                         break;
 295                 }
 296
 297                 pfn = page_to_pfn(page);
 298                 frame_list[i] = pfn_to_mfn(pfn);
 299
 300                 scrub_page(page);
 301
 302                 if (!PageHighMem(page)) {
 303                         ret = HYPERVISOR_update_va_mapping(
 304                                 (unsigned long)__va(pfn << PAGE_SHIFT),
 305                                 __pte_ma(0), 0);
 306                         BUG_ON(ret);
 307                 }
 308
 309         }
 310
 311         /* Ensure that ballooned highmem pages don't have kmaps. */
 312         kmap_flush_unused();
 313         flush_tlb_all();
 314
 315         spin_lock_irqsave(&balloon_lock, flags);
 316
 317         /* No more mappings: invalidate P2M and add to balloon. */
 318         for (i = 0; i < nr_pages; i++) {
 319                 pfn = mfn_to_pfn(frame_list[i]);
 320                 set_phys_to_machine(pfn, INVALID_P2M_ENTRY);
 321                 balloon_append(pfn_to_page(pfn));
 322         }
 323
 324         set_xen_guest_handle(reservation.extent_start, frame_list);
 325         reservation.nr_extents   = nr_pages;
 326         ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation);
 327         BUG_ON(ret != nr_pages);
 328
 329         balloon_stats.current_pages -= nr_pages;
 330         totalram_pages = balloon_stats.current_pages;
 331
 332         spin_unlock_irqrestore(&balloon_lock, flags);
 333
 334         return need_sleep;
 335 }
 336
 337 /*
 338  * We avoid multiple worker processes conflicting via the balloon mutex.
 339  * We may of course race updates of the target counts (which are protected
 340  * by the balloon lock), or with changes to the Xen hard limit, but we will
 341  * recover from these in time.
 342  */
 343 static void balloon_process(struct work_struct *work)
 344 {
 345         int need_sleep = 0;
 346         long credit;
 347
 348         mutex_lock(&balloon_mutex);
 349
 350         do {
 351                 credit = current_target() - balloon_stats.current_pages;
 352                 if (credit > 0)
 353                         need_sleep = (increase_reservation(credit) != 0);
 354                 if (credit < 0)
 355                         need_sleep = (decrease_reservation(-credit) != 0);
 356
 357 #ifndef CONFIG_PREEMPT
 358                 if (need_resched())
 359                         schedule();
 360 #endif
 361         } while ((credit != 0) && !need_sleep);
 362
 363         /* Schedule more work if there is some still to be done. */
 364         if (current_target() != balloon_stats.current_pages)
 365                 mod_timer(&balloon_timer, jiffies + HZ);
 366
 367         mutex_unlock(&balloon_mutex);
 368 }
 369
 370 /* Resets the Xen limit, sets new target, and kicks off processing. */
 371 static void balloon_set_new_target(unsigned long target)
 372 {
 373         /* No need for lock. Not read-modify-write updates. */
 374         balloon_stats.hard_limit   = ~0UL;
 375         balloon_stats.target_pages = target;
 376         schedule_work(&balloon_worker);
 377 }
 378
 379 static struct xenbus_watch target_watch =
 380 {
 381         .node = "memory/target"
 382 };
 383
 384 /* React to a change in the target key */
 385 static void watch_target(struct xenbus_watch *watch,
 386                          const char **vec, unsigned int len)
 387 {
 388         unsigned long long new_target;
 389         int err;
 390
 391         err = xenbus_scanf(XBT_NIL, "memory", "target", "%llu", &new_target);
 392         if (err != 1) {
 393                 /* This is ok (for domain0 at least) - so just return */
 394                 return;
 395         }
 396
 397         /* The given memory/target value is in KiB, so it needs converting to
 398          * pages. PAGE_SHIFT converts bytes to pages, hence PAGE_SHIFT - 10.
 399          */
 400         balloon_set_new_target(new_target >> (PAGE_SHIFT - 10));
 401 }
 402
 403 static int balloon_init_watcher(struct notifier_block *notifier,
 404                                 unsigned long event,
 405                                 void *data)
 406 {
 407         int err;
 408
 409         err = register_xenbus_watch(&target_watch);
 410         if (err)
 411                 printk(KERN_ERR "Failed to set balloon watcher\n");
 412
 413         return NOTIFY_DONE;
 414 }
 415
 416 static struct notifier_block xenstore_notifier;
 417
 418 static int __init balloon_init(void)
 419 {
 420         unsigned long pfn;
 421         struct page *page;
 422
 423         if (!xen_pv_domain())
 424                 return -ENODEV;
 425
 426         pr_info("xen_balloon: Initialising balloon driver.\n");
 427
 428         balloon_stats.current_pages = min(xen_start_info->nr_pages, max_pfn);
 429         totalram_pages   = balloon_stats.current_pages;
 430         balloon_stats.target_pages  = balloon_stats.current_pages;
 431         balloon_stats.balloon_low   = 0;
 432         balloon_stats.balloon_high  = 0;
 433         balloon_stats.driver_pages  = 0UL;
 434         balloon_stats.hard_limit    = ~0UL;
 435
 436         init_timer(&balloon_timer);
 437         balloon_timer.data = 0;
 438         balloon_timer.function = balloon_alarm;
 439
 440         register_balloon(&balloon_sysdev);
 441
 442         /* Initialise the balloon with excess memory space. */
 443         for (pfn = xen_start_info->nr_pages; pfn < max_pfn; pfn++) {
 444                 page = pfn_to_page(pfn);
 445                 if (!PageReserved(page))
 446                         balloon_append(page);
 447         }
 448
 449         target_watch.callback = watch_target;
 450         xenstore_notifier.notifier_call = balloon_init_watcher;
 451
 452         register_xenstore_notifier(&xenstore_notifier);
 453
 454         return 0;
 455 }
 456
 457 subsys_initcall(balloon_init);
 458
 459 static void balloon_exit(void)
 460 {
 461     /* XXX - release balloon here */
 462     return;
 463 }
 464
 465 module_exit(balloon_exit);
 466
 467 #define BALLOON_SHOW(name, format, args...)                             \
 468         static ssize_t show_##name(struct sys_device *dev,              \
 469                                    struct sysdev_attribute *attr,       \
 470                                    char *buf)                           \
 471         {                                                               \
 472                 return sprintf(buf, format, ##args);                    \
 473         }                                                               \
 474         static SYSDEV_ATTR(name, S_IRUGO, show_##name, NULL)
 475
 476 BALLOON_SHOW(current_kb, "%lu\n", PAGES2KB(balloon_stats.current_pages));
 477 BALLOON_SHOW(low_kb, "%lu\n", PAGES2KB(balloon_stats.balloon_low));
 478 BALLOON_SHOW(high_kb, "%lu\n", PAGES2KB(balloon_stats.balloon_high));
 479 BALLOON_SHOW(hard_limit_kb,
 480              (balloon_stats.hard_limit!=~0UL) ? "%lu\n" : "???\n",
 481              (balloon_stats.hard_limit!=~0UL) ? PAGES2KB(balloon_stats.hard_limit) : 0);
 482 BALLOON_SHOW(driver_kb, "%lu\n", PAGES2KB(balloon_stats.driver_pages));
 483
 484 static ssize_t show_target_kb(struct sys_device *dev, struct sysdev_attribute *attr,
 485                               char *buf)
 486 {
 487         return sprintf(buf, "%lu\n", PAGES2KB(balloon_stats.target_pages));
 488 }
 489
 490 static ssize_t store_target_kb(struct sys_device *dev,
 491                                struct sysdev_attribute *attr,
 492                                const char *buf,
 493                                size_t count)
 494 {
 495         char *endchar;
 496         unsigned long long target_bytes;
 497
 498         if (!capable(CAP_SYS_ADMIN))
 499                 return -EPERM;
 500
 501         target_bytes = simple_strtoull(buf, &endchar, 0) * 1024;
 502
 503         balloon_set_new_target(target_bytes >> PAGE_SHIFT);
 504
 505         return count;
 506 }
 507
 508 static SYSDEV_ATTR(target_kb, S_IRUGO | S_IWUSR,
 509                    show_target_kb, store_target_kb);
 510
 511
 512 static ssize_t show_target(struct sys_device *dev, struct sysdev_attribute *attr,
 513                               char *buf)
 514 {
 515         return sprintf(buf, "%llu\n",
 516                        (u64)balloon_stats.target_pages << PAGE_SHIFT);
 517 }
 518
 519 static ssize_t store_target(struct sys_device *dev,
 520                             struct sysdev_attribute *attr,
 521                             const char *buf,
 522                             size_t count)
 523 {
 524         char *endchar;
 525         unsigned long long target_bytes;
 526
 527         if (!capable(CAP_SYS_ADMIN))
 528                 return -EPERM;
 529
 530         target_bytes = memparse(buf, &endchar);
 531
 532         balloon_set_new_target(target_bytes >> PAGE_SHIFT);
 533
 534         return count;
 535 }
 536
 537 static SYSDEV_ATTR(target, S_IRUGO | S_IWUSR,
 538                    show_target, store_target);
 539
 540
 541 static struct sysdev_attribute *balloon_attrs[] = {
 542         &attr_target_kb,
 543         &attr_target,
 544 };
 545
 546 static struct attribute *balloon_info_attrs[] = {
 547         &attr_current_kb.attr,
 548         &attr_low_kb.attr,
 549         &attr_high_kb.attr,
 550         &attr_hard_limit_kb.attr,
 551         &attr_driver_kb.attr,
 552         NULL
 553 };
 554
 555 static struct attribute_group balloon_info_group = {
 556         .name = "info",
 557         .attrs = balloon_info_attrs,
 558 };
 559
 560 static struct sysdev_class balloon_sysdev_class = {
 561         .name = BALLOON_CLASS_NAME,
 562 };
 563
 564 static int register_balloon(struct sys_device *sysdev)
 565 {
 566         int i, error;
 567
 568         error = sysdev_class_register(&balloon_sysdev_class);
 569         if (error)
 570                 return error;
 571
 572         sysdev->id = 0;
 573         sysdev->cls = &balloon_sysdev_class;
 574
 575         error = sysdev_register(sysdev);
 576         if (error) {
 577                 sysdev_class_unregister(&balloon_sysdev_class);
 578                 return error;
 579         }
 580
 581         for (i = 0; i < ARRAY_SIZE(balloon_attrs); i++) {
 582                 error = sysdev_create_file(sysdev, balloon_attrs[i]);
 583                 if (error)
 584                         goto fail;
 585         }
 586
 587         error = sysfs_create_group(&sysdev->kobj, &balloon_info_group);
 588         if (error)
 589                 goto fail;
 590
 591         return 0;
 592
 593  fail:
 594         while (--i >= 0)
 595                 sysdev_remove_file(sysdev, balloon_attrs[i]);
 596         sysdev_unregister(sysdev);
 597         sysdev_class_unregister(&balloon_sysdev_class);
 598         return error;
 599 }
 600
 601 MODULE_LICENSE("GPL");