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/module.h> 46 #include <linux/mm.h> 47 #include <linux/bootmem.h> 48 #include <linux/pagemap.h> 49 #include <linux/highmem.h> 50 #include <linux/mutex.h> 51 #include <linux/list.h> 52 #include <linux/gfp.h> 53 #include <linux/notifier.h> 54 #include <linux/memory.h> 55 #include <linux/memory_hotplug.h> 56 #include <linux/percpu-defs.h> 57 58 #include <asm/page.h> 59 #include <asm/pgalloc.h> 60 #include <asm/pgtable.h> 61 #include <asm/tlb.h> 62 63 #include <asm/xen/hypervisor.h> 64 #include <asm/xen/hypercall.h> 65 66 #include <xen/xen.h> 67 #include <xen/interface/xen.h> 68 #include <xen/interface/memory.h> 69 #include <xen/balloon.h> 70 #include <xen/features.h> 71 #include <xen/page.h> 72 73 /* 74 * balloon_process() state: 75 * 76 * BP_DONE: done or nothing to do, 77 * BP_EAGAIN: error, go to sleep, 78 * BP_ECANCELED: error, balloon operation canceled. 79 */ 80 81 enum bp_state { 82 BP_DONE, 83 BP_EAGAIN, 84 BP_ECANCELED 85 }; 86 87 88 static DEFINE_MUTEX(balloon_mutex); 89 90 struct balloon_stats balloon_stats; 91 EXPORT_SYMBOL_GPL(balloon_stats); 92 93 /* We increase/decrease in batches which fit in a page */ 94 static xen_pfn_t frame_list[PAGE_SIZE / sizeof(unsigned long)]; 95 96 97 /* List of ballooned pages, threaded through the mem_map array. */ 98 static LIST_HEAD(ballooned_pages); 99 100 /* Main work function, always executed in process context. */ 101 static void balloon_process(struct work_struct *work); 102 static DECLARE_DELAYED_WORK(balloon_worker, balloon_process); 103 104 /* When ballooning out (allocating memory to return to Xen) we don't really 105 want the kernel to try too hard since that can trigger the oom killer. */ 106 #define GFP_BALLOON \ 107 (GFP_HIGHUSER | __GFP_NOWARN | __GFP_NORETRY | __GFP_NOMEMALLOC) 108 109 static void scrub_page(struct page *page) 110 { 111 #ifdef CONFIG_XEN_SCRUB_PAGES 112 clear_highpage(page); 113 #endif 114 } 115 116 /* balloon_append: add the given page to the balloon. */ 117 static void __balloon_append(struct page *page) 118 { 119 /* Lowmem is re-populated first, so highmem pages go at list tail. */ 120 if (PageHighMem(page)) { 121 list_add_tail(&page->lru, &ballooned_pages); 122 balloon_stats.balloon_high++; 123 } else { 124 list_add(&page->lru, &ballooned_pages); 125 balloon_stats.balloon_low++; 126 } 127 } 128 129 static void balloon_append(struct page *page) 130 { 131 __balloon_append(page); 132 adjust_managed_page_count(page, -1); 133 } 134 135 /* balloon_retrieve: rescue a page from the balloon, if it is not empty. */ 136 static struct page *balloon_retrieve(bool prefer_highmem) 137 { 138 struct page *page; 139 140 if (list_empty(&ballooned_pages)) 141 return NULL; 142 143 if (prefer_highmem) 144 page = list_entry(ballooned_pages.prev, struct page, lru); 145 else 146 page = list_entry(ballooned_pages.next, struct page, lru); 147 list_del(&page->lru); 148 149 if (PageHighMem(page)) 150 balloon_stats.balloon_high--; 151 else 152 balloon_stats.balloon_low--; 153 154 adjust_managed_page_count(page, 1); 155 156 return page; 157 } 158 159 static struct page *balloon_next_page(struct page *page) 160 { 161 struct list_head *next = page->lru.next; 162 if (next == &ballooned_pages) 163 return NULL; 164 return list_entry(next, struct page, lru); 165 } 166 167 static enum bp_state update_schedule(enum bp_state state) 168 { 169 if (state == BP_ECANCELED) 170 return BP_ECANCELED; 171 172 if (state == BP_DONE) { 173 balloon_stats.schedule_delay = 1; 174 balloon_stats.retry_count = 1; 175 return BP_DONE; 176 } 177 178 ++balloon_stats.retry_count; 179 180 if (balloon_stats.max_retry_count != RETRY_UNLIMITED && 181 balloon_stats.retry_count > balloon_stats.max_retry_count) { 182 balloon_stats.schedule_delay = 1; 183 balloon_stats.retry_count = 1; 184 return BP_ECANCELED; 185 } 186 187 balloon_stats.schedule_delay <<= 1; 188 189 if (balloon_stats.schedule_delay > balloon_stats.max_schedule_delay) 190 balloon_stats.schedule_delay = balloon_stats.max_schedule_delay; 191 192 return BP_EAGAIN; 193 } 194 195 #ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG 196 static long current_credit(void) 197 { 198 return balloon_stats.target_pages - balloon_stats.current_pages - 199 balloon_stats.hotplug_pages; 200 } 201 202 static bool balloon_is_inflated(void) 203 { 204 if (balloon_stats.balloon_low || balloon_stats.balloon_high || 205 balloon_stats.balloon_hotplug) 206 return true; 207 else 208 return false; 209 } 210 211 /* 212 * reserve_additional_memory() adds memory region of size >= credit above 213 * max_pfn. New region is section aligned and size is modified to be multiple 214 * of section size. Those features allow optimal use of address space and 215 * establish proper alignment when this function is called first time after 216 * boot (last section not fully populated at boot time contains unused memory 217 * pages with PG_reserved bit not set; online_pages_range() does not allow page 218 * onlining in whole range if first onlined page does not have PG_reserved 219 * bit set). Real size of added memory is established at page onlining stage. 220 */ 221 222 static enum bp_state reserve_additional_memory(long credit) 223 { 224 int nid, rc; 225 u64 hotplug_start_paddr; 226 unsigned long balloon_hotplug = credit; 227 228 hotplug_start_paddr = PFN_PHYS(SECTION_ALIGN_UP(max_pfn)); 229 balloon_hotplug = round_up(balloon_hotplug, PAGES_PER_SECTION); 230 nid = memory_add_physaddr_to_nid(hotplug_start_paddr); 231 232 #ifdef CONFIG_XEN_HAVE_PVMMU 233 /* 234 * add_memory() will build page tables for the new memory so 235 * the p2m must contain invalid entries so the correct 236 * non-present PTEs will be written. 237 * 238 * If a failure occurs, the original (identity) p2m entries 239 * are not restored since this region is now known not to 240 * conflict with any devices. 241 */ 242 if (!xen_feature(XENFEAT_auto_translated_physmap)) { 243 unsigned long pfn, i; 244 245 pfn = PFN_DOWN(hotplug_start_paddr); 246 for (i = 0; i < balloon_hotplug; i++) { 247 if (!set_phys_to_machine(pfn + i, INVALID_P2M_ENTRY)) { 248 pr_warn("set_phys_to_machine() failed, no memory added\n"); 249 return BP_ECANCELED; 250 } 251 } 252 } 253 #endif 254 255 rc = add_memory(nid, hotplug_start_paddr, balloon_hotplug << PAGE_SHIFT); 256 257 if (rc) { 258 pr_warn("Cannot add additional memory (%i)\n", rc); 259 return BP_ECANCELED; 260 } 261 262 balloon_hotplug -= credit; 263 264 balloon_stats.hotplug_pages += credit; 265 balloon_stats.balloon_hotplug = balloon_hotplug; 266 267 return BP_DONE; 268 } 269 270 static void xen_online_page(struct page *page) 271 { 272 __online_page_set_limits(page); 273 274 mutex_lock(&balloon_mutex); 275 276 __balloon_append(page); 277 278 if (balloon_stats.hotplug_pages) 279 --balloon_stats.hotplug_pages; 280 else 281 --balloon_stats.balloon_hotplug; 282 283 mutex_unlock(&balloon_mutex); 284 } 285 286 static int xen_memory_notifier(struct notifier_block *nb, unsigned long val, void *v) 287 { 288 if (val == MEM_ONLINE) 289 schedule_delayed_work(&balloon_worker, 0); 290 291 return NOTIFY_OK; 292 } 293 294 static struct notifier_block xen_memory_nb = { 295 .notifier_call = xen_memory_notifier, 296 .priority = 0 297 }; 298 #else 299 static long current_credit(void) 300 { 301 unsigned long target = balloon_stats.target_pages; 302 303 target = min(target, 304 balloon_stats.current_pages + 305 balloon_stats.balloon_low + 306 balloon_stats.balloon_high); 307 308 return target - balloon_stats.current_pages; 309 } 310 311 static bool balloon_is_inflated(void) 312 { 313 if (balloon_stats.balloon_low || balloon_stats.balloon_high) 314 return true; 315 else 316 return false; 317 } 318 319 static enum bp_state reserve_additional_memory(long credit) 320 { 321 balloon_stats.target_pages = balloon_stats.current_pages; 322 return BP_DONE; 323 } 324 #endif /* CONFIG_XEN_BALLOON_MEMORY_HOTPLUG */ 325 326 static enum bp_state increase_reservation(unsigned long nr_pages) 327 { 328 int rc; 329 unsigned long pfn, i; 330 struct page *page; 331 struct xen_memory_reservation reservation = { 332 .address_bits = 0, 333 .extent_order = 0, 334 .domid = DOMID_SELF 335 }; 336 337 #ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG 338 if (!balloon_stats.balloon_low && !balloon_stats.balloon_high) { 339 nr_pages = min(nr_pages, balloon_stats.balloon_hotplug); 340 balloon_stats.hotplug_pages += nr_pages; 341 balloon_stats.balloon_hotplug -= nr_pages; 342 return BP_DONE; 343 } 344 #endif 345 346 if (nr_pages > ARRAY_SIZE(frame_list)) 347 nr_pages = ARRAY_SIZE(frame_list); 348 349 page = list_first_entry_or_null(&ballooned_pages, struct page, lru); 350 for (i = 0; i < nr_pages; i++) { 351 if (!page) { 352 nr_pages = i; 353 break; 354 } 355 frame_list[i] = page_to_pfn(page); 356 page = balloon_next_page(page); 357 } 358 359 set_xen_guest_handle(reservation.extent_start, frame_list); 360 reservation.nr_extents = nr_pages; 361 rc = HYPERVISOR_memory_op(XENMEM_populate_physmap, &reservation); 362 if (rc <= 0) 363 return BP_EAGAIN; 364 365 for (i = 0; i < rc; i++) { 366 page = balloon_retrieve(false); 367 BUG_ON(page == NULL); 368 369 pfn = page_to_pfn(page); 370 371 #ifdef CONFIG_XEN_HAVE_PVMMU 372 if (!xen_feature(XENFEAT_auto_translated_physmap)) { 373 set_phys_to_machine(pfn, frame_list[i]); 374 375 /* Link back into the page tables if not highmem. */ 376 if (!PageHighMem(page)) { 377 int ret; 378 ret = HYPERVISOR_update_va_mapping( 379 (unsigned long)__va(pfn << PAGE_SHIFT), 380 mfn_pte(frame_list[i], PAGE_KERNEL), 381 0); 382 BUG_ON(ret); 383 } 384 } 385 #endif 386 387 /* Relinquish the page back to the allocator. */ 388 __free_reserved_page(page); 389 } 390 391 balloon_stats.current_pages += rc; 392 393 return BP_DONE; 394 } 395 396 static enum bp_state decrease_reservation(unsigned long nr_pages, gfp_t gfp) 397 { 398 enum bp_state state = BP_DONE; 399 unsigned long pfn, i; 400 struct page *page; 401 int ret; 402 struct xen_memory_reservation reservation = { 403 .address_bits = 0, 404 .extent_order = 0, 405 .domid = DOMID_SELF 406 }; 407 408 #ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG 409 if (balloon_stats.hotplug_pages) { 410 nr_pages = min(nr_pages, balloon_stats.hotplug_pages); 411 balloon_stats.hotplug_pages -= nr_pages; 412 balloon_stats.balloon_hotplug += nr_pages; 413 return BP_DONE; 414 } 415 #endif 416 417 if (nr_pages > ARRAY_SIZE(frame_list)) 418 nr_pages = ARRAY_SIZE(frame_list); 419 420 for (i = 0; i < nr_pages; i++) { 421 page = alloc_page(gfp); 422 if (page == NULL) { 423 nr_pages = i; 424 state = BP_EAGAIN; 425 break; 426 } 427 scrub_page(page); 428 429 frame_list[i] = page_to_pfn(page); 430 } 431 432 /* 433 * Ensure that ballooned highmem pages don't have kmaps. 434 * 435 * Do this before changing the p2m as kmap_flush_unused() 436 * reads PTEs to obtain pages (and hence needs the original 437 * p2m entry). 438 */ 439 kmap_flush_unused(); 440 441 /* Update direct mapping, invalidate P2M, and add to balloon. */ 442 for (i = 0; i < nr_pages; i++) { 443 pfn = frame_list[i]; 444 frame_list[i] = pfn_to_gfn(pfn); 445 page = pfn_to_page(pfn); 446 447 #ifdef CONFIG_XEN_HAVE_PVMMU 448 if (!xen_feature(XENFEAT_auto_translated_physmap)) { 449 if (!PageHighMem(page)) { 450 ret = HYPERVISOR_update_va_mapping( 451 (unsigned long)__va(pfn << PAGE_SHIFT), 452 __pte_ma(0), 0); 453 BUG_ON(ret); 454 } 455 __set_phys_to_machine(pfn, INVALID_P2M_ENTRY); 456 } 457 #endif 458 459 balloon_append(page); 460 } 461 462 flush_tlb_all(); 463 464 set_xen_guest_handle(reservation.extent_start, frame_list); 465 reservation.nr_extents = nr_pages; 466 ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation); 467 BUG_ON(ret != nr_pages); 468 469 balloon_stats.current_pages -= nr_pages; 470 471 return state; 472 } 473 474 /* 475 * As this is a work item it is guaranteed to run as a single instance only. 476 * We may of course race updates of the target counts (which are protected 477 * by the balloon lock), or with changes to the Xen hard limit, but we will 478 * recover from these in time. 479 */ 480 static void balloon_process(struct work_struct *work) 481 { 482 enum bp_state state = BP_DONE; 483 long credit; 484 485 486 do { 487 mutex_lock(&balloon_mutex); 488 489 credit = current_credit(); 490 491 if (credit > 0) { 492 if (balloon_is_inflated()) 493 state = increase_reservation(credit); 494 else 495 state = reserve_additional_memory(credit); 496 } 497 498 if (credit < 0) 499 state = decrease_reservation(-credit, GFP_BALLOON); 500 501 state = update_schedule(state); 502 503 mutex_unlock(&balloon_mutex); 504 505 cond_resched(); 506 507 } while (credit && state == BP_DONE); 508 509 /* Schedule more work if there is some still to be done. */ 510 if (state == BP_EAGAIN) 511 schedule_delayed_work(&balloon_worker, balloon_stats.schedule_delay * HZ); 512 } 513 514 /* Resets the Xen limit, sets new target, and kicks off processing. */ 515 void balloon_set_new_target(unsigned long target) 516 { 517 /* No need for lock. Not read-modify-write updates. */ 518 balloon_stats.target_pages = target; 519 schedule_delayed_work(&balloon_worker, 0); 520 } 521 EXPORT_SYMBOL_GPL(balloon_set_new_target); 522 523 /** 524 * alloc_xenballooned_pages - get pages that have been ballooned out 525 * @nr_pages: Number of pages to get 526 * @pages: pages returned 527 * @highmem: allow highmem pages 528 * @return 0 on success, error otherwise 529 */ 530 int alloc_xenballooned_pages(int nr_pages, struct page **pages, bool highmem) 531 { 532 int pgno = 0; 533 struct page *page; 534 mutex_lock(&balloon_mutex); 535 while (pgno < nr_pages) { 536 page = balloon_retrieve(highmem); 537 if (page && (highmem || !PageHighMem(page))) { 538 pages[pgno++] = page; 539 } else { 540 enum bp_state st; 541 if (page) 542 balloon_append(page); 543 st = decrease_reservation(nr_pages - pgno, 544 highmem ? GFP_HIGHUSER : GFP_USER); 545 if (st != BP_DONE) 546 goto out_undo; 547 } 548 } 549 mutex_unlock(&balloon_mutex); 550 return 0; 551 out_undo: 552 while (pgno) 553 balloon_append(pages[--pgno]); 554 /* Free the memory back to the kernel soon */ 555 schedule_delayed_work(&balloon_worker, 0); 556 mutex_unlock(&balloon_mutex); 557 return -ENOMEM; 558 } 559 EXPORT_SYMBOL(alloc_xenballooned_pages); 560 561 /** 562 * free_xenballooned_pages - return pages retrieved with get_ballooned_pages 563 * @nr_pages: Number of pages 564 * @pages: pages to return 565 */ 566 void free_xenballooned_pages(int nr_pages, struct page **pages) 567 { 568 int i; 569 570 mutex_lock(&balloon_mutex); 571 572 for (i = 0; i < nr_pages; i++) { 573 if (pages[i]) 574 balloon_append(pages[i]); 575 } 576 577 /* The balloon may be too large now. Shrink it if needed. */ 578 if (current_credit()) 579 schedule_delayed_work(&balloon_worker, 0); 580 581 mutex_unlock(&balloon_mutex); 582 } 583 EXPORT_SYMBOL(free_xenballooned_pages); 584 585 static void __init balloon_add_region(unsigned long start_pfn, 586 unsigned long pages) 587 { 588 unsigned long pfn, extra_pfn_end; 589 struct page *page; 590 591 /* 592 * If the amount of usable memory has been limited (e.g., with 593 * the 'mem' command line parameter), don't add pages beyond 594 * this limit. 595 */ 596 extra_pfn_end = min(max_pfn, start_pfn + pages); 597 598 for (pfn = start_pfn; pfn < extra_pfn_end; pfn++) { 599 page = pfn_to_page(pfn); 600 /* totalram_pages and totalhigh_pages do not 601 include the boot-time balloon extension, so 602 don't subtract from it. */ 603 __balloon_append(page); 604 } 605 } 606 607 static int __init balloon_init(void) 608 { 609 int i; 610 611 if (!xen_domain()) 612 return -ENODEV; 613 614 pr_info("Initialising balloon driver\n"); 615 616 balloon_stats.current_pages = xen_pv_domain() 617 ? min(xen_start_info->nr_pages - xen_released_pages, max_pfn) 618 : get_num_physpages(); 619 balloon_stats.target_pages = balloon_stats.current_pages; 620 balloon_stats.balloon_low = 0; 621 balloon_stats.balloon_high = 0; 622 623 balloon_stats.schedule_delay = 1; 624 balloon_stats.max_schedule_delay = 32; 625 balloon_stats.retry_count = 1; 626 balloon_stats.max_retry_count = RETRY_UNLIMITED; 627 628 #ifdef CONFIG_XEN_BALLOON_MEMORY_HOTPLUG 629 balloon_stats.hotplug_pages = 0; 630 balloon_stats.balloon_hotplug = 0; 631 632 set_online_page_callback(&xen_online_page); 633 register_memory_notifier(&xen_memory_nb); 634 #endif 635 636 /* 637 * Initialize the balloon with pages from the extra memory 638 * regions (see arch/x86/xen/setup.c). 639 */ 640 for (i = 0; i < XEN_EXTRA_MEM_MAX_REGIONS; i++) 641 if (xen_extra_mem[i].n_pfns) 642 balloon_add_region(xen_extra_mem[i].start_pfn, 643 xen_extra_mem[i].n_pfns); 644 645 return 0; 646 } 647 648 subsys_initcall(balloon_init); 649 650 MODULE_LICENSE("GPL"); 651