drivers/xen/balloon.c

   1 /******************************************************************************
   2  * Xen balloon driver - enables returning/claiming memory to/from Xen.
   3  *
   4  * Copyright (c) 2003, B Dragovic
   5  * Copyright (c) 2003-2004, M Williamson, K Fraser
   6  * Copyright (c) 2005 Dan M. Smith, IBM Corporation
   7  * Copyright (c) 2010 Daniel Kiper
   8  *
   9  * Memory hotplug support was written by Daniel Kiper. Work on
  10  * it was sponsored by Google under Google Summer of Code 2010
  11  * program. Jeremy Fitzhardinge from Citrix was the mentor for
  12  * this project.
  13  *
  14  * This program is free software; you can redistribute it and/or
  15  * modify it under the terms of the GNU General Public License version 2
  16  * as published by the Free Software Foundation; or, when distributed
  17  * separately from the Linux kernel or incorporated into other
  18  * software packages, subject to the following license:
  19  *
  20  * Permission is hereby granted, free of charge, to any person obtaining a copy
  21  * of this source file (the "Software"), to deal in the Software without
  22  * restriction, including without limitation the rights to use, copy, modify,
  23  * merge, publish, distribute, sublicense, and/or sell copies of the Software,
  24  * and to permit persons to whom the Software is furnished to do so, subject to
  25  * the following conditions:
  26  *
  27  * The above copyright notice and this permission notice shall be included in
  28  * all copies or substantial portions of the Software.
  29  *
  30  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
  31  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
  32  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
  33  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
  34  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
  35  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
  36  * IN THE SOFTWARE.
  37  */
  38
  39 #define pr_fmt(fmt) "xen:" KBUILD_MODNAME ": " fmt
  40
  41 #include <linux/cpu.h>
  42 #include <linux/kernel.h>
  43 #include <linux/sched.h>
  44 #include <linux/errno.h>
  45 #include <linux/mm.h>
  46 #include <linux/bootmem.h>
  47 #include <linux/pagemap.h>
  48 #include <linux/highmem.h>
  49 #include <linux/mutex.h>
  50 #include <linux/list.h>
  51 #include <linux/gfp.h>
  52 #include <linux/notifier.h>
  53 #include <linux/memory.h>
  54 #include <linux/memory_hotplug.h>
  55 #include <linux/percpu-defs.h>
  56 #include <linux/slab.h>
  57 #include <linux/sysctl.h>
  58
  59 #include <asm/page.h>
  60 #include <asm/pgalloc.h>
  61 #include <asm/pgtable.h>
  62 #include <asm/tlb.h>
  63
  64 #include <asm/xen/hypervisor.h>
  65 #include <asm/xen/hypercall.h>
  66
  67 #include <xen/xen.h>
  68 #include <xen/interface/xen.h>
  69 #include <xen/interface/memory.h>
  70 #include <xen/balloon.h>
  71 #include <xen/features.h>
  72 #include <xen/page.h>
  73
  74 static int xen_hotplug_unpopulated;
  75
  76 #ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG
  77
  78 static int zero;
  79 static int one = 1;
  80
  81 static struct ctl_table balloon_table[] = {
  82         {
  83                 .procname       = "hotplug_unpopulated",
  84                 .data           = &xen_hotplug_unpopulated,
  85                 .maxlen         = sizeof(int),
  86                 .mode           = 0644,
  87                 .proc_handler   = proc_dointvec_minmax,
  88                 .extra1         = &zero,
  89                 .extra2         = &one,
  90         },
  91         { }
  92 };
  93
  94 static struct ctl_table balloon_root[] = {
  95         {
  96                 .procname       = "balloon",
  97                 .mode           = 0555,
  98                 .child          = balloon_table,
  99         },
 100         { }
 101 };
 102
 103 static struct ctl_table xen_root[] = {
 104         {
 105                 .procname       = "xen",
 106                 .mode           = 0555,
 107                 .child          = balloon_root,
 108         },
 109         { }
 110 };
 111
 112 #endif
 113
 114 /*
 115  * Use one extent per PAGE_SIZE to avoid to break down the page into
 116  * multiple frame.
 117  */
 118 #define EXTENT_ORDER (fls(XEN_PFN_PER_PAGE) - 1)
 119
 120 /*
 121  * balloon_process() state:
 122  *
 123  * BP_DONE: done or nothing to do,
 124  * BP_WAIT: wait to be rescheduled,
 125  * BP_EAGAIN: error, go to sleep,
 126  * BP_ECANCELED: error, balloon operation canceled.
 127  */
 128
 129 enum bp_state {
 130         BP_DONE,
 131         BP_WAIT,
 132         BP_EAGAIN,
 133         BP_ECANCELED
 134 };
 135
 136
 137 static DEFINE_MUTEX(balloon_mutex);
 138
 139 struct balloon_stats balloon_stats;
 140 EXPORT_SYMBOL_GPL(balloon_stats);
 141
 142 /* We increase/decrease in batches which fit in a page */
 143 static xen_pfn_t frame_list[PAGE_SIZE / sizeof(xen_pfn_t)];
 144
 145
 146 /* List of ballooned pages, threaded through the mem_map array. */
 147 static LIST_HEAD(ballooned_pages);
 148 static DECLARE_WAIT_QUEUE_HEAD(balloon_wq);
 149
 150 /* Main work function, always executed in process context. */
 151 static void balloon_process(struct work_struct *work);
 152 static DECLARE_DELAYED_WORK(balloon_worker, balloon_process);
 153
 154 /* When ballooning out (allocating memory to return to Xen) we don't really
 155    want the kernel to try too hard since that can trigger the oom killer. */
 156 #define GFP_BALLOON \
 157         (GFP_HIGHUSER | __GFP_NOWARN | __GFP_NORETRY | __GFP_NOMEMALLOC)
 158
 159 static void scrub_page(struct page *page)
 160 {
 161 #ifdef CONFIG_XEN_SCRUB_PAGES
 162         clear_highpage(page);
 163 #endif
 164 }
 165
 166 /* balloon_append: add the given page to the balloon. */
 167 static void __balloon_append(struct page *page)
 168 {
 169         /* Lowmem is re-populated first, so highmem pages go at list tail. */
 170         if (PageHighMem(page)) {
 171                 list_add_tail(&page->lru, &ballooned_pages);
 172                 balloon_stats.balloon_high++;
 173         } else {
 174                 list_add(&page->lru, &ballooned_pages);
 175                 balloon_stats.balloon_low++;
 176         }
 177         wake_up(&balloon_wq);
 178 }
 179
 180 static void balloon_append(struct page *page)
 181 {
 182         __balloon_append(page);
 183 }
 184
 185 /* balloon_retrieve: rescue a page from the balloon, if it is not empty. */
 186 static struct page *balloon_retrieve(bool require_lowmem)
 187 {
 188         struct page *page;
 189
 190         if (list_empty(&ballooned_pages))
 191                 return NULL;
 192
 193         page = list_entry(ballooned_pages.next, struct page, lru);
 194         if (require_lowmem && PageHighMem(page))
 195                 return NULL;
 196         list_del(&page->lru);
 197
 198         if (PageHighMem(page))
 199                 balloon_stats.balloon_high--;
 200         else
 201                 balloon_stats.balloon_low--;
 202
 203         return page;
 204 }
 205
 206 static struct page *balloon_next_page(struct page *page)
 207 {
 208         struct list_head *next = page->lru.next;
 209         if (next == &ballooned_pages)
 210                 return NULL;
 211         return list_entry(next, struct page, lru);
 212 }
 213
 214 static enum bp_state update_schedule(enum bp_state state)
 215 {
 216         if (state == BP_WAIT)
 217                 return BP_WAIT;
 218
 219         if (state == BP_ECANCELED)
 220                 return BP_ECANCELED;
 221
 222         if (state == BP_DONE) {
 223                 balloon_stats.schedule_delay = 1;
 224                 balloon_stats.retry_count = 1;
 225                 return BP_DONE;
 226         }
 227
 228         ++balloon_stats.retry_count;
 229
 230         if (balloon_stats.max_retry_count != RETRY_UNLIMITED &&
 231                         balloon_stats.retry_count > balloon_stats.max_retry_count) {
 232                 balloon_stats.schedule_delay = 1;
 233                 balloon_stats.retry_count = 1;
 234                 return BP_ECANCELED;
 235         }
 236
 237         balloon_stats.schedule_delay <<= 1;
 238
 239         if (balloon_stats.schedule_delay > balloon_stats.max_schedule_delay)
 240                 balloon_stats.schedule_delay = balloon_stats.max_schedule_delay;
 241
 242         return BP_EAGAIN;
 243 }
 244
 245 #ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG
 246 static void release_memory_resource(struct resource *resource)
 247 {
 248         if (!resource)
 249                 return;
 250
 251         /*
 252          * No need to reset region to identity mapped since we now
 253          * know that no I/O can be in this region
 254          */
 255         release_resource(resource);
 256         kfree(resource);
 257 }
 258
 259 static struct resource *additional_memory_resource(phys_addr_t size)
 260 {
 261         struct resource *res;
 262         int ret;
 263
 264         res = kzalloc(sizeof(*res), GFP_KERNEL);
 265         if (!res)
 266                 return NULL;
 267
 268         res->name = "System RAM";
 269         res->flags = IORESOURCE_SYSTEM_RAM | IORESOURCE_BUSY;
 270
 271         ret = allocate_resource(&iomem_resource, res,
 272                                 size, 0, -1,
 273                                 PAGES_PER_SECTION * PAGE_SIZE, NULL, NULL);
 274         if (ret < 0) {
 275                 pr_err("Cannot allocate new System RAM resource\n");
 276                 kfree(res);
 277                 return NULL;
 278         }
 279
 280 #ifdef CONFIG_SPARSEMEM
 281         {
 282                 unsigned long limit = 1UL << (MAX_PHYSMEM_BITS - PAGE_SHIFT);
 283                 unsigned long pfn = res->start >> PAGE_SHIFT;
 284
 285                 if (pfn > limit) {
 286                         pr_err("New System RAM resource outside addressable RAM (%lu > %lu)\n",
 287                                pfn, limit);
 288                         release_memory_resource(res);
 289                         return NULL;
 290                 }
 291         }
 292 #endif
 293
 294         return res;
 295 }
 296
 297 static enum bp_state reserve_additional_memory(void)
 298 {
 299         long credit;
 300         struct resource *resource;
 301         int nid, rc;
 302         unsigned long balloon_hotplug;
 303
 304         credit = balloon_stats.target_pages + balloon_stats.target_unpopulated
 305                 - balloon_stats.total_pages;
 306
 307         /*
 308          * Already hotplugged enough pages?  Wait for them to be
 309          * onlined.
 310          */
 311         if (credit <= 0)
 312                 return BP_WAIT;
 313
 314         balloon_hotplug = round_up(credit, PAGES_PER_SECTION);
 315
 316         resource = additional_memory_resource(balloon_hotplug * PAGE_SIZE);
 317         if (!resource)
 318                 goto err;
 319
 320         nid = memory_add_physaddr_to_nid(resource->start);
 321
 322 #ifdef CONFIG_XEN_HAVE_PVMMU
 323         /*
 324          * We don't support PV MMU when Linux and Xen is using
 325          * different page granularity.
 326          */
 327         BUILD_BUG_ON(XEN_PAGE_SIZE != PAGE_SIZE);
 328
 329         /*
 330          * add_memory() will build page tables for the new memory so
 331          * the p2m must contain invalid entries so the correct
 332          * non-present PTEs will be written.
 333          *
 334          * If a failure occurs, the original (identity) p2m entries
 335          * are not restored since this region is now known not to
 336          * conflict with any devices.
 337          */
 338         if (!xen_feature(XENFEAT_auto_translated_physmap)) {
 339                 unsigned long pfn, i;
 340
 341                 pfn = PFN_DOWN(resource->start);
 342                 for (i = 0; i < balloon_hotplug; i++) {
 343                         if (!set_phys_to_machine(pfn + i, INVALID_P2M_ENTRY)) {
 344                                 pr_warn("set_phys_to_machine() failed, no memory added\n");
 345                                 goto err;
 346                         }
 347                 }
 348         }
 349 #endif
 350
 351         /*
 352          * add_memory_resource() will call online_pages() which in its turn
 353          * will call xen_online_page() callback causing deadlock if we don't
 354          * release balloon_mutex here. Unlocking here is safe because the
 355          * callers drop the mutex before trying again.
 356          */
 357         mutex_unlock(&balloon_mutex);
 358         rc = add_memory_resource(nid, resource, memhp_auto_online);
 359         mutex_lock(&balloon_mutex);
 360
 361         if (rc) {
 362                 pr_warn("Cannot add additional memory (%i)\n", rc);
 363                 goto err;
 364         }
 365
 366         balloon_stats.total_pages += balloon_hotplug;
 367
 368         return BP_WAIT;
 369   err:
 370         release_memory_resource(resource);
 371         return BP_ECANCELED;
 372 }
 373
 374 static void xen_online_page(struct page *page)
 375 {
 376         __online_page_set_limits(page);
 377
 378         mutex_lock(&balloon_mutex);
 379
 380         __balloon_append(page);
 381
 382         mutex_unlock(&balloon_mutex);
 383 }
 384
 385 static int xen_memory_notifier(struct notifier_block *nb, unsigned long val, void *v)
 386 {
 387         if (val == MEM_ONLINE)
 388                 schedule_delayed_work(&balloon_worker, 0);
 389
 390         return NOTIFY_OK;
 391 }
 392
 393 static struct notifier_block xen_memory_nb = {
 394         .notifier_call = xen_memory_notifier,
 395         .priority = 0
 396 };
 397 #else
 398 static enum bp_state reserve_additional_memory(void)
 399 {
 400         balloon_stats.target_pages = balloon_stats.current_pages;
 401         return BP_ECANCELED;
 402 }
 403 #endif /* CONFIG_XEN_BALLOON_MEMORY_HOTPLUG */
 404
 405 static long current_credit(void)
 406 {
 407         return balloon_stats.target_pages - balloon_stats.current_pages;
 408 }
 409
 410 static bool balloon_is_inflated(void)
 411 {
 412         return balloon_stats.balloon_low || balloon_stats.balloon_high;
 413 }
 414
 415 static enum bp_state increase_reservation(unsigned long nr_pages)
 416 {
 417         int rc;
 418         unsigned long i;
 419         struct page   *page;
 420         struct xen_memory_reservation reservation = {
 421                 .address_bits = 0,
 422                 .extent_order = EXTENT_ORDER,
 423                 .domid        = DOMID_SELF
 424         };
 425
 426         if (nr_pages > ARRAY_SIZE(frame_list))
 427                 nr_pages = ARRAY_SIZE(frame_list);
 428
 429         page = list_first_entry_or_null(&ballooned_pages, struct page, lru);
 430         for (i = 0; i < nr_pages; i++) {
 431                 if (!page) {
 432                         nr_pages = i;
 433                         break;
 434                 }
 435
 436                 /* XENMEM_populate_physmap requires a PFN based on Xen
 437                  * granularity.
 438                  */
 439                 frame_list[i] = page_to_xen_pfn(page);
 440                 page = balloon_next_page(page);
 441         }
 442
 443         set_xen_guest_handle(reservation.extent_start, frame_list);
 444         reservation.nr_extents = nr_pages;
 445         rc = HYPERVISOR_memory_op(XENMEM_populate_physmap, &reservation);
 446         if (rc <= 0)
 447                 return BP_EAGAIN;
 448
 449         for (i = 0; i < rc; i++) {
 450                 page = balloon_retrieve(false);
 451                 BUG_ON(page == NULL);
 452
 453 #ifdef CONFIG_XEN_HAVE_PVMMU
 454                 /*
 455                  * We don't support PV MMU when Linux and Xen is using
 456                  * different page granularity.
 457                  */
 458                 BUILD_BUG_ON(XEN_PAGE_SIZE != PAGE_SIZE);
 459
 460                 if (!xen_feature(XENFEAT_auto_translated_physmap)) {
 461                         unsigned long pfn = page_to_pfn(page);
 462
 463                         set_phys_to_machine(pfn, frame_list[i]);
 464
 465                         /* Link back into the page tables if not highmem. */
 466                         if (!PageHighMem(page)) {
 467                                 int ret;
 468                                 ret = HYPERVISOR_update_va_mapping(
 469                                                 (unsigned long)__va(pfn << PAGE_SHIFT),
 470                                                 mfn_pte(frame_list[i], PAGE_KERNEL),
 471                                                 0);
 472                                 BUG_ON(ret);
 473                         }
 474                 }
 475 #endif
 476
 477                 /* Relinquish the page back to the allocator. */
 478                 free_reserved_page(page);
 479         }
 480
 481         balloon_stats.current_pages += rc;
 482
 483         return BP_DONE;
 484 }
 485
 486 static enum bp_state decrease_reservation(unsigned long nr_pages, gfp_t gfp)
 487 {
 488         enum bp_state state = BP_DONE;
 489         unsigned long i;
 490         struct page *page, *tmp;
 491         int ret;
 492         struct xen_memory_reservation reservation = {
 493                 .address_bits = 0,
 494                 .extent_order = EXTENT_ORDER,
 495                 .domid        = DOMID_SELF
 496         };
 497         LIST_HEAD(pages);
 498
 499         if (nr_pages > ARRAY_SIZE(frame_list))
 500                 nr_pages = ARRAY_SIZE(frame_list);
 501
 502         for (i = 0; i < nr_pages; i++) {
 503                 page = alloc_page(gfp);
 504                 if (page == NULL) {
 505                         nr_pages = i;
 506                         state = BP_EAGAIN;
 507                         break;
 508                 }
 509                 adjust_managed_page_count(page, -1);
 510                 scrub_page(page);
 511                 list_add(&page->lru, &pages);
 512         }
 513
 514         /*
 515          * Ensure that ballooned highmem pages don't have kmaps.
 516          *
 517          * Do this before changing the p2m as kmap_flush_unused()
 518          * reads PTEs to obtain pages (and hence needs the original
 519          * p2m entry).
 520          */
 521         kmap_flush_unused();
 522
 523         /*
 524          * Setup the frame, update direct mapping, invalidate P2M,
 525          * and add to balloon.
 526          */
 527         i = 0;
 528         list_for_each_entry_safe(page, tmp, &pages, lru) {
 529                 /* XENMEM_decrease_reservation requires a GFN */
 530                 frame_list[i++] = xen_page_to_gfn(page);
 531
 532 #ifdef CONFIG_XEN_HAVE_PVMMU
 533                 /*
 534                  * We don't support PV MMU when Linux and Xen is using
 535                  * different page granularity.
 536                  */
 537                 BUILD_BUG_ON(XEN_PAGE_SIZE != PAGE_SIZE);
 538
 539                 if (!xen_feature(XENFEAT_auto_translated_physmap)) {
 540                         unsigned long pfn = page_to_pfn(page);
 541
 542                         if (!PageHighMem(page)) {
 543                                 ret = HYPERVISOR_update_va_mapping(
 544                                                 (unsigned long)__va(pfn << PAGE_SHIFT),
 545                                                 __pte_ma(0), 0);
 546                                 BUG_ON(ret);
 547                         }
 548                         __set_phys_to_machine(pfn, INVALID_P2M_ENTRY);
 549                 }
 550 #endif
 551                 list_del(&page->lru);
 552
 553                 balloon_append(page);
 554         }
 555
 556         flush_tlb_all();
 557
 558         set_xen_guest_handle(reservation.extent_start, frame_list);
 559         reservation.nr_extents   = nr_pages;
 560         ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation);
 561         BUG_ON(ret != nr_pages);
 562
 563         balloon_stats.current_pages -= nr_pages;
 564
 565         return state;
 566 }
 567
 568 /*
 569  * As this is a work item it is guaranteed to run as a single instance only.
 570  * We may of course race updates of the target counts (which are protected
 571  * by the balloon lock), or with changes to the Xen hard limit, but we will
 572  * recover from these in time.
 573  */
 574 static void balloon_process(struct work_struct *work)
 575 {
 576         enum bp_state state = BP_DONE;
 577         long credit;
 578
 579
 580         do {
 581                 mutex_lock(&balloon_mutex);
 582
 583                 credit = current_credit();
 584
 585                 if (credit > 0) {
 586                         if (balloon_is_inflated())
 587                                 state = increase_reservation(credit);
 588                         else
 589                                 state = reserve_additional_memory();
 590                 }
 591
 592                 if (credit < 0)
 593                         state = decrease_reservation(-credit, GFP_BALLOON);
 594
 595                 state = update_schedule(state);
 596
 597                 mutex_unlock(&balloon_mutex);
 598
 599                 cond_resched();
 600
 601         } while (credit && state == BP_DONE);
 602
 603         /* Schedule more work if there is some still to be done. */
 604         if (state == BP_EAGAIN)
 605                 schedule_delayed_work(&balloon_worker, balloon_stats.schedule_delay * HZ);
 606 }
 607
 608 /* Resets the Xen limit, sets new target, and kicks off processing. */
 609 void balloon_set_new_target(unsigned long target)
 610 {
 611         /* No need for lock. Not read-modify-write updates. */
 612         balloon_stats.target_pages = target;
 613         schedule_delayed_work(&balloon_worker, 0);
 614 }
 615 EXPORT_SYMBOL_GPL(balloon_set_new_target);
 616
 617 static int add_ballooned_pages(int nr_pages)
 618 {
 619         enum bp_state st;
 620
 621         if (xen_hotplug_unpopulated) {
 622                 st = reserve_additional_memory();
 623                 if (st != BP_ECANCELED) {
 624                         mutex_unlock(&balloon_mutex);
 625                         wait_event(balloon_wq,
 626                                    !list_empty(&ballooned_pages));
 627                         mutex_lock(&balloon_mutex);
 628                         return 0;
 629                 }
 630         }
 631
 632         st = decrease_reservation(nr_pages, GFP_USER);
 633         if (st != BP_DONE)
 634                 return -ENOMEM;
 635
 636         return 0;
 637 }
 638
 639 /**
 640  * alloc_xenballooned_pages - get pages that have been ballooned out
 641  * @nr_pages: Number of pages to get
 642  * @pages: pages returned
 643  * @return 0 on success, error otherwise
 644  */
 645 int alloc_xenballooned_pages(int nr_pages, struct page **pages)
 646 {
 647         int pgno = 0;
 648         struct page *page;
 649         int ret;
 650
 651         mutex_lock(&balloon_mutex);
 652
 653         balloon_stats.target_unpopulated += nr_pages;
 654
 655         while (pgno < nr_pages) {
 656                 page = balloon_retrieve(true);
 657                 if (page) {
 658                         pages[pgno++] = page;
 659 #ifdef CONFIG_XEN_HAVE_PVMMU
 660                         /*
 661                          * We don't support PV MMU when Linux and Xen is using
 662                          * different page granularity.
 663                          */
 664                         BUILD_BUG_ON(XEN_PAGE_SIZE != PAGE_SIZE);
 665
 666                         ret = xen_alloc_p2m_entry(page_to_pfn(page));
 667                         if (ret < 0)
 668                                 goto out_undo;
 669 #endif
 670                 } else {
 671                         ret = add_ballooned_pages(nr_pages - pgno);
 672                         if (ret < 0)
 673                                 goto out_undo;
 674                 }
 675         }
 676         mutex_unlock(&balloon_mutex);
 677         return 0;
 678  out_undo:
 679         mutex_unlock(&balloon_mutex);
 680         free_xenballooned_pages(pgno, pages);
 681         return ret;
 682 }
 683 EXPORT_SYMBOL(alloc_xenballooned_pages);
 684
 685 /**
 686  * free_xenballooned_pages - return pages retrieved with get_ballooned_pages
 687  * @nr_pages: Number of pages
 688  * @pages: pages to return
 689  */
 690 void free_xenballooned_pages(int nr_pages, struct page **pages)
 691 {
 692         int i;
 693
 694         mutex_lock(&balloon_mutex);
 695
 696         for (i = 0; i < nr_pages; i++) {
 697                 if (pages[i])
 698                         balloon_append(pages[i]);
 699         }
 700
 701         balloon_stats.target_unpopulated -= nr_pages;
 702
 703         /* The balloon may be too large now. Shrink it if needed. */
 704         if (current_credit())
 705                 schedule_delayed_work(&balloon_worker, 0);
 706
 707         mutex_unlock(&balloon_mutex);
 708 }
 709 EXPORT_SYMBOL(free_xenballooned_pages);
 710
 711 static void __init balloon_add_region(unsigned long start_pfn,
 712                                       unsigned long pages)
 713 {
 714         unsigned long pfn, extra_pfn_end;
 715         struct page *page;
 716
 717         /*
 718          * If the amount of usable memory has been limited (e.g., with
 719          * the 'mem' command line parameter), don't add pages beyond
 720          * this limit.
 721          */
 722         extra_pfn_end = min(max_pfn, start_pfn + pages);
 723
 724         for (pfn = start_pfn; pfn < extra_pfn_end; pfn++) {
 725                 page = pfn_to_page(pfn);
 726                 /* totalram_pages and totalhigh_pages do not
 727                    include the boot-time balloon extension, so
 728                    don't subtract from it. */
 729                 __balloon_append(page);
 730         }
 731
 732         balloon_stats.total_pages += extra_pfn_end - start_pfn;
 733 }
 734
 735 static int __init balloon_init(void)
 736 {
 737         int i;
 738
 739         if (!xen_domain())
 740                 return -ENODEV;
 741
 742         pr_info("Initialising balloon driver\n");
 743
 744         balloon_stats.current_pages = xen_pv_domain()
 745                 ? min(xen_start_info->nr_pages - xen_released_pages, max_pfn)
 746                 : get_num_physpages();
 747         balloon_stats.target_pages  = balloon_stats.current_pages;
 748         balloon_stats.balloon_low   = 0;
 749         balloon_stats.balloon_high  = 0;
 750         balloon_stats.total_pages   = balloon_stats.current_pages;
 751
 752         balloon_stats.schedule_delay = 1;
 753         balloon_stats.max_schedule_delay = 32;
 754         balloon_stats.retry_count = 1;
 755         balloon_stats.max_retry_count = RETRY_UNLIMITED;
 756
 757 #ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG
 758         set_online_page_callback(&xen_online_page);
 759         register_memory_notifier(&xen_memory_nb);
 760         register_sysctl_table(xen_root);
 761 #endif
 762
 763         /*
 764          * Initialize the balloon with pages from the extra memory
 765          * regions (see arch/x86/xen/setup.c).
 766          */
 767         for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++)
 768                 if (xen_extra_mem[i].n_pfns)
 769                         balloon_add_region(xen_extra_mem[i].start_pfn,
 770                                            xen_extra_mem[i].n_pfns);
 771
 772         return 0;
 773 }
 774 subsys_initcall(balloon_init);