]> err.no Git - linux-2.6/blob - drivers/xen/balloon.c
Merge branch 'generic-ipi' into generic-ipi-for-linus
[linux-2.6] / 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/highmem.h>
45 #include <linux/list.h>
46 #include <linux/sysdev.h>
47
48 #include <asm/xen/hypervisor.h>
49 #include <asm/page.h>
50 #include <asm/pgalloc.h>
51 #include <asm/pgtable.h>
52 #include <asm/uaccess.h>
53 #include <asm/tlb.h>
54
55 #include <xen/interface/memory.h>
56 #include <xen/balloon.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 "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         if (PageHighMem(page)) {
128                 void *v = kmap(page);
129                 clear_page(v);
130                 kunmap(v);
131         } else {
132                 void *v = page_address(page);
133                 clear_page(v);
134         }
135 #endif
136 }
137
138 /* balloon_append: add the given page to the balloon. */
139 static void balloon_append(struct page *page)
140 {
141         /* Lowmem is re-populated first, so highmem pages go at list tail. */
142         if (PageHighMem(page)) {
143                 list_add_tail(&page->lru, &ballooned_pages);
144                 balloon_stats.balloon_high++;
145                 dec_totalhigh_pages();
146         } else {
147                 list_add(&page->lru, &ballooned_pages);
148                 balloon_stats.balloon_low++;
149         }
150 }
151
152 /* balloon_retrieve: rescue a page from the balloon, if it is not empty. */
153 static struct page *balloon_retrieve(void)
154 {
155         struct page *page;
156
157         if (list_empty(&ballooned_pages))
158                 return NULL;
159
160         page = list_entry(ballooned_pages.next, struct page, lru);
161         list_del(&page->lru);
162
163         if (PageHighMem(page)) {
164                 balloon_stats.balloon_high--;
165                 inc_totalhigh_pages();
166         }
167         else
168                 balloon_stats.balloon_low--;
169
170         return page;
171 }
172
173 static struct page *balloon_first_page(void)
174 {
175         if (list_empty(&ballooned_pages))
176                 return NULL;
177         return list_entry(ballooned_pages.next, struct page, lru);
178 }
179
180 static struct page *balloon_next_page(struct page *page)
181 {
182         struct list_head *next = page->lru.next;
183         if (next == &ballooned_pages)
184                 return NULL;
185         return list_entry(next, struct page, lru);
186 }
187
188 static void balloon_alarm(unsigned long unused)
189 {
190         schedule_work(&balloon_worker);
191 }
192
193 static unsigned long current_target(void)
194 {
195         unsigned long target = min(balloon_stats.target_pages, balloon_stats.hard_limit);
196
197         target = min(target,
198                      balloon_stats.current_pages +
199                      balloon_stats.balloon_low +
200                      balloon_stats.balloon_high);
201
202         return target;
203 }
204
205 static int increase_reservation(unsigned long nr_pages)
206 {
207         unsigned long  pfn, i, flags;
208         struct page   *page;
209         long           rc;
210         struct xen_memory_reservation reservation = {
211                 .address_bits = 0,
212                 .extent_order = 0,
213                 .domid        = DOMID_SELF
214         };
215
216         if (nr_pages > ARRAY_SIZE(frame_list))
217                 nr_pages = ARRAY_SIZE(frame_list);
218
219         spin_lock_irqsave(&balloon_lock, flags);
220
221         page = balloon_first_page();
222         for (i = 0; i < nr_pages; i++) {
223                 BUG_ON(page == NULL);
224                 frame_list[i] = page_to_pfn(page);;
225                 page = balloon_next_page(page);
226         }
227
228         set_xen_guest_handle(reservation.extent_start, frame_list);
229         reservation.nr_extents   = nr_pages;
230         rc = HYPERVISOR_memory_op(
231                 XENMEM_populate_physmap, &reservation);
232         if (rc < nr_pages) {
233                 if (rc > 0) {
234                         int ret;
235
236                         /* We hit the Xen hard limit: reprobe. */
237                         reservation.nr_extents = rc;
238                         ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation,
239                                         &reservation);
240                         BUG_ON(ret != rc);
241                 }
242                 if (rc >= 0)
243                         balloon_stats.hard_limit = (balloon_stats.current_pages + rc -
244                                                     balloon_stats.driver_pages);
245                 goto out;
246         }
247
248         for (i = 0; i < nr_pages; i++) {
249                 page = balloon_retrieve();
250                 BUG_ON(page == NULL);
251
252                 pfn = page_to_pfn(page);
253                 BUG_ON(!xen_feature(XENFEAT_auto_translated_physmap) &&
254                        phys_to_machine_mapping_valid(pfn));
255
256                 set_phys_to_machine(pfn, frame_list[i]);
257
258                 /* Link back into the page tables if not highmem. */
259                 if (pfn < max_low_pfn) {
260                         int ret;
261                         ret = HYPERVISOR_update_va_mapping(
262                                 (unsigned long)__va(pfn << PAGE_SHIFT),
263                                 mfn_pte(frame_list[i], PAGE_KERNEL),
264                                 0);
265                         BUG_ON(ret);
266                 }
267
268                 /* Relinquish the page back to the allocator. */
269                 ClearPageReserved(page);
270                 init_page_count(page);
271                 __free_page(page);
272         }
273
274         balloon_stats.current_pages += nr_pages;
275         totalram_pages = balloon_stats.current_pages;
276
277  out:
278         spin_unlock_irqrestore(&balloon_lock, flags);
279
280         return 0;
281 }
282
283 static int decrease_reservation(unsigned long nr_pages)
284 {
285         unsigned long  pfn, i, flags;
286         struct page   *page;
287         int            need_sleep = 0;
288         int ret;
289         struct xen_memory_reservation reservation = {
290                 .address_bits = 0,
291                 .extent_order = 0,
292                 .domid        = DOMID_SELF
293         };
294
295         if (nr_pages > ARRAY_SIZE(frame_list))
296                 nr_pages = ARRAY_SIZE(frame_list);
297
298         for (i = 0; i < nr_pages; i++) {
299                 if ((page = alloc_page(GFP_BALLOON)) == NULL) {
300                         nr_pages = i;
301                         need_sleep = 1;
302                         break;
303                 }
304
305                 pfn = page_to_pfn(page);
306                 frame_list[i] = pfn_to_mfn(pfn);
307
308                 scrub_page(page);
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 (!is_running_on_xen())
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 static void balloon_update_driver_allowance(long delta)
468 {
469         unsigned long flags;
470
471         spin_lock_irqsave(&balloon_lock, flags);
472         balloon_stats.driver_pages += delta;
473         spin_unlock_irqrestore(&balloon_lock, flags);
474 }
475
476 static int dealloc_pte_fn(
477         pte_t *pte, struct page *pmd_page, unsigned long addr, void *data)
478 {
479         unsigned long mfn = pte_mfn(*pte);
480         int ret;
481         struct xen_memory_reservation reservation = {
482                 .nr_extents   = 1,
483                 .extent_order = 0,
484                 .domid        = DOMID_SELF
485         };
486         set_xen_guest_handle(reservation.extent_start, &mfn);
487         set_pte_at(&init_mm, addr, pte, __pte_ma(0ull));
488         set_phys_to_machine(__pa(addr) >> PAGE_SHIFT, INVALID_P2M_ENTRY);
489         ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation, &reservation);
490         BUG_ON(ret != 1);
491         return 0;
492 }
493
494 static struct page **alloc_empty_pages_and_pagevec(int nr_pages)
495 {
496         unsigned long vaddr, flags;
497         struct page *page, **pagevec;
498         int i, ret;
499
500         pagevec = kmalloc(sizeof(page) * nr_pages, GFP_KERNEL);
501         if (pagevec == NULL)
502                 return NULL;
503
504         for (i = 0; i < nr_pages; i++) {
505                 page = pagevec[i] = alloc_page(GFP_KERNEL);
506                 if (page == NULL)
507                         goto err;
508
509                 vaddr = (unsigned long)page_address(page);
510
511                 scrub_page(page);
512
513                 spin_lock_irqsave(&balloon_lock, flags);
514
515                 if (xen_feature(XENFEAT_auto_translated_physmap)) {
516                         unsigned long gmfn = page_to_pfn(page);
517                         struct xen_memory_reservation reservation = {
518                                 .nr_extents   = 1,
519                                 .extent_order = 0,
520                                 .domid        = DOMID_SELF
521                         };
522                         set_xen_guest_handle(reservation.extent_start, &gmfn);
523                         ret = HYPERVISOR_memory_op(XENMEM_decrease_reservation,
524                                                    &reservation);
525                         if (ret == 1)
526                                 ret = 0; /* success */
527                 } else {
528                         ret = apply_to_page_range(&init_mm, vaddr, PAGE_SIZE,
529                                                   dealloc_pte_fn, NULL);
530                 }
531
532                 if (ret != 0) {
533                         spin_unlock_irqrestore(&balloon_lock, flags);
534                         __free_page(page);
535                         goto err;
536                 }
537
538                 totalram_pages = --balloon_stats.current_pages;
539
540                 spin_unlock_irqrestore(&balloon_lock, flags);
541         }
542
543  out:
544         schedule_work(&balloon_worker);
545         flush_tlb_all();
546         return pagevec;
547
548  err:
549         spin_lock_irqsave(&balloon_lock, flags);
550         while (--i >= 0)
551                 balloon_append(pagevec[i]);
552         spin_unlock_irqrestore(&balloon_lock, flags);
553         kfree(pagevec);
554         pagevec = NULL;
555         goto out;
556 }
557
558 static void free_empty_pages_and_pagevec(struct page **pagevec, int nr_pages)
559 {
560         unsigned long flags;
561         int i;
562
563         if (pagevec == NULL)
564                 return;
565
566         spin_lock_irqsave(&balloon_lock, flags);
567         for (i = 0; i < nr_pages; i++) {
568                 BUG_ON(page_count(pagevec[i]) != 1);
569                 balloon_append(pagevec[i]);
570         }
571         spin_unlock_irqrestore(&balloon_lock, flags);
572
573         kfree(pagevec);
574
575         schedule_work(&balloon_worker);
576 }
577
578 static void balloon_release_driver_page(struct page *page)
579 {
580         unsigned long flags;
581
582         spin_lock_irqsave(&balloon_lock, flags);
583         balloon_append(page);
584         balloon_stats.driver_pages--;
585         spin_unlock_irqrestore(&balloon_lock, flags);
586
587         schedule_work(&balloon_worker);
588 }
589
590
591 #define BALLOON_SHOW(name, format, args...)                     \
592         static ssize_t show_##name(struct sys_device *dev,      \
593                                    char *buf)                   \
594         {                                                       \
595                 return sprintf(buf, format, ##args);            \
596         }                                                       \
597         static SYSDEV_ATTR(name, S_IRUGO, show_##name, NULL)
598
599 BALLOON_SHOW(current_kb, "%lu\n", PAGES2KB(balloon_stats.current_pages));
600 BALLOON_SHOW(low_kb, "%lu\n", PAGES2KB(balloon_stats.balloon_low));
601 BALLOON_SHOW(high_kb, "%lu\n", PAGES2KB(balloon_stats.balloon_high));
602 BALLOON_SHOW(hard_limit_kb,
603              (balloon_stats.hard_limit!=~0UL) ? "%lu\n" : "???\n",
604              (balloon_stats.hard_limit!=~0UL) ? PAGES2KB(balloon_stats.hard_limit) : 0);
605 BALLOON_SHOW(driver_kb, "%lu\n", PAGES2KB(balloon_stats.driver_pages));
606
607 static ssize_t show_target_kb(struct sys_device *dev, char *buf)
608 {
609         return sprintf(buf, "%lu\n", PAGES2KB(balloon_stats.target_pages));
610 }
611
612 static ssize_t store_target_kb(struct sys_device *dev,
613                                const char *buf,
614                                size_t count)
615 {
616         char memstring[64], *endchar;
617         unsigned long long target_bytes;
618
619         if (!capable(CAP_SYS_ADMIN))
620                 return -EPERM;
621
622         if (count <= 1)
623                 return -EBADMSG; /* runt */
624         if (count > sizeof(memstring))
625                 return -EFBIG;   /* too long */
626         strcpy(memstring, buf);
627
628         target_bytes = memparse(memstring, &endchar);
629         balloon_set_new_target(target_bytes >> PAGE_SHIFT);
630
631         return count;
632 }
633
634 static SYSDEV_ATTR(target_kb, S_IRUGO | S_IWUSR,
635                    show_target_kb, store_target_kb);
636
637 static struct sysdev_attribute *balloon_attrs[] = {
638         &attr_target_kb,
639 };
640
641 static struct attribute *balloon_info_attrs[] = {
642         &attr_current_kb.attr,
643         &attr_low_kb.attr,
644         &attr_high_kb.attr,
645         &attr_hard_limit_kb.attr,
646         &attr_driver_kb.attr,
647         NULL
648 };
649
650 static struct attribute_group balloon_info_group = {
651         .name = "info",
652         .attrs = balloon_info_attrs,
653 };
654
655 static struct sysdev_class balloon_sysdev_class = {
656         .name = BALLOON_CLASS_NAME,
657 };
658
659 static int register_balloon(struct sys_device *sysdev)
660 {
661         int i, error;
662
663         error = sysdev_class_register(&balloon_sysdev_class);
664         if (error)
665                 return error;
666
667         sysdev->id = 0;
668         sysdev->cls = &balloon_sysdev_class;
669
670         error = sysdev_register(sysdev);
671         if (error) {
672                 sysdev_class_unregister(&balloon_sysdev_class);
673                 return error;
674         }
675
676         for (i = 0; i < ARRAY_SIZE(balloon_attrs); i++) {
677                 error = sysdev_create_file(sysdev, balloon_attrs[i]);
678                 if (error)
679                         goto fail;
680         }
681
682         error = sysfs_create_group(&sysdev->kobj, &balloon_info_group);
683         if (error)
684                 goto fail;
685
686         return 0;
687
688  fail:
689         while (--i >= 0)
690                 sysdev_remove_file(sysdev, balloon_attrs[i]);
691         sysdev_unregister(sysdev);
692         sysdev_class_unregister(&balloon_sysdev_class);
693         return error;
694 }
695
696 static void unregister_balloon(struct sys_device *sysdev)
697 {
698         int i;
699
700         sysfs_remove_group(&sysdev->kobj, &balloon_info_group);
701         for (i = 0; i < ARRAY_SIZE(balloon_attrs); i++)
702                 sysdev_remove_file(sysdev, balloon_attrs[i]);
703         sysdev_unregister(sysdev);
704         sysdev_class_unregister(&balloon_sysdev_class);
705 }
706
707 static void balloon_sysfs_exit(void)
708 {
709         unregister_balloon(&balloon_sysdev);
710 }
711
712 MODULE_LICENSE("GPL");