ia64/xen-unstable

annotate xen/common/memory.c @ 18972:bd33ff263e2c

PoD memory 4/9: Decrease reservation

Handle balloon driver's calls to decrease_reservation properly.
* Replace PoD entries with p2m_none
* Steal memory for the cache instead of freeing, if need be

Signed-off-by: George Dunlap <george.dunlap@eu.citrix.com>
author Keir Fraser <keir.fraser@citrix.com>
date Mon Jan 05 10:43:50 2009 +0000 (2009-01-05)
parents 9a7b46546e05
children 2090917489c5
rev   line source
kaf24@6486 1 /******************************************************************************
kaf24@6486 2 * memory.c
kaf24@6486 3 *
kaf24@6486 4 * Code to handle memory-related requests.
kaf24@6486 5 *
kaf24@6486 6 * Copyright (c) 2003-2004, B Dragovic
kaf24@6486 7 * Copyright (c) 2003-2005, K A Fraser
kaf24@6486 8 */
kaf24@6486 9
kaf24@6486 10 #include <xen/config.h>
kaf24@6486 11 #include <xen/types.h>
kaf24@6486 12 #include <xen/lib.h>
kaf24@6486 13 #include <xen/mm.h>
kaf24@6486 14 #include <xen/perfc.h>
kaf24@6486 15 #include <xen/sched.h>
kaf24@6486 16 #include <xen/event.h>
Tim@15635 17 #include <xen/paging.h>
kaf24@8468 18 #include <xen/iocap.h>
kaf24@9068 19 #include <xen/guest_access.h>
ack@13295 20 #include <xen/hypercall.h>
kaf24@11219 21 #include <xen/errno.h>
kaf24@6486 22 #include <asm/current.h>
kaf24@6486 23 #include <asm/hardirq.h>
keir@17385 24 #include <xen/numa.h>
kaf24@6486 25 #include <public/memory.h>
kfraser@15815 26 #include <xsm/xsm.h>
kaf24@6486 27
kfraser@12374 28 struct memop_args {
kfraser@12374 29 /* INPUT */
kfraser@12374 30 struct domain *domain; /* Domain to be affected. */
kfraser@12374 31 XEN_GUEST_HANDLE(xen_pfn_t) extent_list; /* List of extent base addrs. */
kfraser@12374 32 unsigned int nr_extents; /* Number of extents to allocate or free. */
kfraser@12374 33 unsigned int extent_order; /* Size of each extent. */
kfraser@12374 34 unsigned int memflags; /* Allocation flags. */
kfraser@12374 35
kfraser@12374 36 /* INPUT/OUTPUT */
kfraser@12374 37 unsigned int nr_done; /* Number of extents processed so far. */
kfraser@12374 38 int preempted; /* Was the hypercall preempted? */
kfraser@12374 39 };
kfraser@12374 40
kfraser@12374 41 static void increase_reservation(struct memop_args *a)
kaf24@6486 42 {
kaf24@8726 43 struct page_info *page;
kaf24@10314 44 unsigned long i;
kaf24@10314 45 xen_pfn_t mfn;
kfraser@12374 46 struct domain *d = a->domain;
kaf24@6486 47
kfraser@12374 48 if ( !guest_handle_is_null(a->extent_list) &&
keir@17860 49 !guest_handle_subrange_okay(a->extent_list, a->nr_done,
keir@17860 50 a->nr_extents-1) )
kfraser@12374 51 return;
kaf24@6486 52
kfraser@12374 53 if ( (a->extent_order != 0) &&
kaf24@8468 54 !multipage_allocation_permitted(current->domain) )
kfraser@12374 55 return;
kaf24@6486 56
kfraser@12374 57 for ( i = a->nr_done; i < a->nr_extents; i++ )
kaf24@6486 58 {
kaf24@6486 59 if ( hypercall_preempt_check() )
kaf24@6607 60 {
kfraser@12374 61 a->preempted = 1;
kfraser@12374 62 goto out;
kaf24@6607 63 }
kaf24@6486 64
keir@17986 65 page = alloc_domheap_pages(d, a->extent_order, a->memflags);
kfraser@12374 66 if ( unlikely(page == NULL) )
kaf24@6486 67 {
kaf24@12038 68 gdprintk(XENLOG_INFO, "Could not allocate order=%d extent: "
kfraser@10418 69 "id=%d memflags=%x (%ld of %d)\n",
kfraser@12374 70 a->extent_order, d->domain_id, a->memflags,
kfraser@12374 71 i, a->nr_extents);
kfraser@12374 72 goto out;
kaf24@6486 73 }
kaf24@6486 74
kaf24@6486 75 /* Inform the domain of the new page's machine address. */
kfraser@12374 76 if ( !guest_handle_is_null(a->extent_list) )
kaf24@8859 77 {
kaf24@8859 78 mfn = page_to_mfn(page);
kfraser@12374 79 if ( unlikely(__copy_to_guest_offset(a->extent_list, i, &mfn, 1)) )
kfraser@12374 80 goto out;
kaf24@8859 81 }
kaf24@6486 82 }
kaf24@6486 83
kfraser@12374 84 out:
kfraser@12374 85 a->nr_done = i;
kaf24@6486 86 }
sos22@8688 87
kfraser@12374 88 static void populate_physmap(struct memop_args *a)
kaf24@8673 89 {
kaf24@8726 90 struct page_info *page;
kaf24@10314 91 unsigned long i, j;
kfraser@12374 92 xen_pfn_t gpfn, mfn;
kfraser@12374 93 struct domain *d = a->domain;
kaf24@8673 94
keir@17860 95 if ( !guest_handle_subrange_okay(a->extent_list, a->nr_done,
keir@17860 96 a->nr_extents-1) )
kfraser@12374 97 return;
kaf24@8673 98
kfraser@12374 99 if ( (a->extent_order != 0) &&
kaf24@8673 100 !multipage_allocation_permitted(current->domain) )
kfraser@12374 101 return;
kaf24@8673 102
kfraser@12374 103 for ( i = a->nr_done; i < a->nr_extents; i++ )
kaf24@8673 104 {
kaf24@8673 105 if ( hypercall_preempt_check() )
kaf24@8673 106 {
kfraser@12374 107 a->preempted = 1;
sos22@8688 108 goto out;
kaf24@8673 109 }
kaf24@8673 110
kfraser@12374 111 if ( unlikely(__copy_from_guest_offset(&gpfn, a->extent_list, i, 1)) )
kaf24@8859 112 goto out;
kaf24@8859 113
keir@17986 114 page = alloc_domheap_pages(d, a->extent_order, a->memflags);
kfraser@12374 115 if ( unlikely(page == NULL) )
kaf24@8673 116 {
kaf24@12038 117 gdprintk(XENLOG_INFO, "Could not allocate order=%d extent: "
kfraser@12374 118 "id=%d memflags=%x (%ld of %d)\n",
kfraser@12374 119 a->extent_order, d->domain_id, a->memflags,
kfraser@12374 120 i, a->nr_extents);
sos22@8688 121 goto out;
kaf24@8673 122 }
kaf24@8673 123
kaf24@8726 124 mfn = page_to_mfn(page);
keir@17727 125 guest_physmap_add_page(d, gpfn, mfn, a->extent_order);
kaf24@8673 126
keir@17727 127 if ( !paging_mode_translate(d) )
kaf24@8694 128 {
kfraser@12374 129 for ( j = 0; j < (1 << a->extent_order); j++ )
kaf24@8736 130 set_gpfn_from_mfn(mfn + j, gpfn + j);
kaf24@8673 131
sos22@8688 132 /* Inform the domain of the new page's machine address. */
kfraser@12374 133 if ( unlikely(__copy_to_guest_offset(a->extent_list, i, &mfn, 1)) )
sos22@8688 134 goto out;
sos22@8688 135 }
kaf24@8673 136 }
kaf24@8673 137
sos22@8688 138 out:
kfraser@12374 139 a->nr_done = i;
kaf24@8673 140 }
cl349@9211 141
kfraser@12374 142 int guest_remove_page(struct domain *d, unsigned long gmfn)
cl349@9211 143 {
cl349@9211 144 struct page_info *page;
cl349@9211 145 unsigned long mfn;
cl349@9211 146
cl349@9211 147 mfn = gmfn_to_mfn(d, gmfn);
cl349@9211 148 if ( unlikely(!mfn_valid(mfn)) )
cl349@9211 149 {
kaf24@12038 150 gdprintk(XENLOG_INFO, "Domain %u page number %lx invalid\n",
tdeegan@11172 151 d->domain_id, gmfn);
cl349@9211 152 return 0;
cl349@9211 153 }
cl349@9211 154
cl349@9211 155 page = mfn_to_page(mfn);
cl349@9211 156 if ( unlikely(!get_page(page, d)) )
cl349@9211 157 {
kaf24@12038 158 gdprintk(XENLOG_INFO, "Bad page free for domain %u\n", d->domain_id);
cl349@9211 159 return 0;
cl349@9211 160 }
cl349@9211 161
cl349@9211 162 if ( test_and_clear_bit(_PGT_pinned, &page->u.inuse.type_info) )
cl349@9211 163 put_page_and_type(page);
cl349@9211 164
cl349@9211 165 if ( test_and_clear_bit(_PGC_allocated, &page->count_info) )
cl349@9211 166 put_page(page);
cl349@9211 167
keir@17664 168 guest_physmap_remove_page(d, gmfn, mfn, 0);
cl349@9211 169
cl349@9211 170 put_page(page);
cl349@9211 171
cl349@9211 172 return 1;
cl349@9211 173 }
cl349@9211 174
kfraser@12374 175 static void decrease_reservation(struct memop_args *a)
kaf24@6486 176 {
kaf24@10314 177 unsigned long i, j;
kaf24@10314 178 xen_pfn_t gmfn;
kaf24@6486 179
keir@17860 180 if ( !guest_handle_subrange_okay(a->extent_list, a->nr_done,
keir@17860 181 a->nr_extents-1) )
kfraser@12374 182 return;
kaf24@6486 183
kfraser@12374 184 for ( i = a->nr_done; i < a->nr_extents; i++ )
kaf24@6486 185 {
kaf24@6486 186 if ( hypercall_preempt_check() )
kaf24@6607 187 {
kfraser@12374 188 a->preempted = 1;
kfraser@12374 189 goto out;
kaf24@6607 190 }
kaf24@6486 191
kfraser@12374 192 if ( unlikely(__copy_from_guest_offset(&gmfn, a->extent_list, i, 1)) )
kfraser@12374 193 goto out;
kaf24@6486 194
keir@18972 195 /* See if populate-on-demand wants to handle this */
keir@18972 196 if ( is_hvm_domain(a->domain)
keir@18972 197 && p2m_pod_decrease_reservation(a->domain, gmfn, a->extent_order) )
keir@18972 198 continue;
keir@18972 199
kfraser@12374 200 for ( j = 0; j < (1 << a->extent_order); j++ )
kfraser@12374 201 if ( !guest_remove_page(a->domain, gmfn + j) )
kfraser@12374 202 goto out;
kaf24@6486 203 }
kaf24@6486 204
kfraser@12374 205 out:
kfraser@12374 206 a->nr_done = i;
kaf24@6486 207 }
kaf24@6486 208
kfraser@12374 209 static long translate_gpfn_list(
kaf24@9873 210 XEN_GUEST_HANDLE(xen_translate_gpfn_list_t) uop, unsigned long *progress)
kaf24@8871 211 {
kaf24@8871 212 struct xen_translate_gpfn_list op;
kaf24@10314 213 unsigned long i;
kaf24@10314 214 xen_pfn_t gpfn;
kaf24@10314 215 xen_pfn_t mfn;
kaf24@8871 216 struct domain *d;
kfraser@15815 217 int rc;
kaf24@6486 218
kaf24@9068 219 if ( copy_from_guest(&op, uop, 1) )
kaf24@8871 220 return -EFAULT;
kaf24@8871 221
kaf24@8871 222 /* Is size too large for us to encode a continuation? */
ack@13295 223 if ( op.nr_gpfns > (ULONG_MAX >> MEMOP_EXTENT_SHIFT) )
kaf24@8871 224 return -EINVAL;
kaf24@8871 225
keir@17860 226 if ( !guest_handle_subrange_okay(op.gpfn_list, *progress, op.nr_gpfns-1) ||
keir@17860 227 !guest_handle_subrange_okay(op.mfn_list, *progress, op.nr_gpfns-1) )
kaf24@8871 228 return -EFAULT;
kaf24@8871 229
keir@18574 230 rc = rcu_lock_target_domain_by_id(op.domid, &d);
keir@18574 231 if ( rc )
keir@18574 232 return rc;
kaf24@8871 233
Tim@15635 234 if ( !paging_mode_translate(d) )
kaf24@8871 235 {
kfraser@14192 236 rcu_unlock_domain(d);
kaf24@8871 237 return -EINVAL;
kaf24@8871 238 }
kaf24@8871 239
kaf24@8871 240 for ( i = *progress; i < op.nr_gpfns; i++ )
kaf24@8871 241 {
kaf24@8871 242 if ( hypercall_preempt_check() )
kaf24@8871 243 {
kfraser@14192 244 rcu_unlock_domain(d);
kaf24@8871 245 *progress = i;
kaf24@8871 246 return -EAGAIN;
kaf24@8871 247 }
kaf24@8871 248
kaf24@9068 249 if ( unlikely(__copy_from_guest_offset(&gpfn, op.gpfn_list, i, 1)) )
kaf24@8871 250 {
kfraser@14192 251 rcu_unlock_domain(d);
kaf24@8871 252 return -EFAULT;
kaf24@8871 253 }
kaf24@8871 254
kaf24@8871 255 mfn = gmfn_to_mfn(d, gpfn);
kaf24@8871 256
kfraser@15815 257 rc = xsm_translate_gpfn_list(current->domain, mfn);
kfraser@15815 258 if ( rc )
kfraser@15815 259 {
kfraser@15815 260 rcu_unlock_domain(d);
kfraser@15815 261 return rc;
kfraser@15815 262 }
kfraser@15815 263
kaf24@9068 264 if ( unlikely(__copy_to_guest_offset(op.mfn_list, i, &mfn, 1)) )
kaf24@8871 265 {
kfraser@14192 266 rcu_unlock_domain(d);
kaf24@8871 267 return -EFAULT;
kaf24@8871 268 }
kaf24@8871 269 }
kaf24@8871 270
kfraser@14192 271 rcu_unlock_domain(d);
kaf24@8871 272 return 0;
kaf24@8871 273 }
kaf24@8871 274
kfraser@12374 275 static long memory_exchange(XEN_GUEST_HANDLE(xen_memory_exchange_t) arg)
kfraser@10418 276 {
kfraser@10418 277 struct xen_memory_exchange exch;
kfraser@10418 278 LIST_HEAD(in_chunk_list);
kfraser@10418 279 LIST_HEAD(out_chunk_list);
kfraser@10418 280 unsigned long in_chunk_order, out_chunk_order;
kaf24@10459 281 xen_pfn_t gpfn, gmfn, mfn;
kfraser@10418 282 unsigned long i, j, k;
keir@17988 283 unsigned int node, memflags = 0;
kfraser@10418 284 long rc = 0;
kfraser@10418 285 struct domain *d;
kfraser@10418 286 struct page_info *page;
kfraser@10418 287
kfraser@10418 288 if ( copy_from_guest(&exch, arg, 1) )
kfraser@10418 289 return -EFAULT;
kfraser@10418 290
kfraser@10418 291 /* Various sanity checks. */
kfraser@10418 292 if ( (exch.nr_exchanged > exch.in.nr_extents) ||
kfraser@10418 293 /* Input and output domain identifiers match? */
kfraser@10418 294 (exch.in.domid != exch.out.domid) ||
kfraser@10418 295 /* Sizes of input and output lists do not overflow a long? */
kfraser@10418 296 ((~0UL >> exch.in.extent_order) < exch.in.nr_extents) ||
kfraser@10418 297 ((~0UL >> exch.out.extent_order) < exch.out.nr_extents) ||
kfraser@10418 298 /* Sizes of input and output lists match? */
kfraser@10418 299 ((exch.in.nr_extents << exch.in.extent_order) !=
kfraser@10418 300 (exch.out.nr_extents << exch.out.extent_order)) )
kfraser@10418 301 {
kfraser@10418 302 rc = -EINVAL;
kfraser@10418 303 goto fail_early;
kfraser@10418 304 }
kfraser@10418 305
kfraser@10418 306 /* Only privileged guests can allocate multi-page contiguous extents. */
kfraser@10418 307 if ( ((exch.in.extent_order != 0) || (exch.out.extent_order != 0)) &&
kfraser@10418 308 !multipage_allocation_permitted(current->domain) )
kfraser@10418 309 {
kfraser@10418 310 rc = -EPERM;
kfraser@10418 311 goto fail_early;
kfraser@10418 312 }
kfraser@10418 313
kfraser@10418 314 if ( exch.in.extent_order <= exch.out.extent_order )
kfraser@10418 315 {
kfraser@10418 316 in_chunk_order = exch.out.extent_order - exch.in.extent_order;
kfraser@10418 317 out_chunk_order = 0;
kfraser@10418 318 }
kfraser@10418 319 else
kfraser@10418 320 {
kfraser@10418 321 in_chunk_order = 0;
kfraser@10418 322 out_chunk_order = exch.in.extent_order - exch.out.extent_order;
kfraser@10418 323 }
kfraser@10418 324
kfraser@10418 325 /*
kfraser@10418 326 * Only support exchange on calling domain right now. Otherwise there are
kfraser@14642 327 * tricky corner cases to consider (e.g., dying domain).
kfraser@10418 328 */
kfraser@10418 329 if ( unlikely(exch.in.domid != DOMID_SELF) )
kfraser@10418 330 {
kfraser@10418 331 rc = IS_PRIV(current->domain) ? -EINVAL : -EPERM;
kfraser@10418 332 goto fail_early;
kfraser@10418 333 }
kfraser@10418 334 d = current->domain;
kfraser@10418 335
keir@16548 336 memflags |= MEMF_bits(domain_clamp_alloc_bitsize(
keir@17986 337 d,
keir@17986 338 XENMEMF_get_address_bits(exch.out.mem_flags) ? :
keir@17986 339 (BITS_PER_LONG+PAGE_SHIFT)));
keir@17988 340 node = XENMEMF_get_node(exch.out.mem_flags);
keir@17988 341 if ( node == NUMA_NO_NODE )
keir@17988 342 node = domain_to_node(d);
keir@17988 343 memflags |= MEMF_node(node);
kfraser@11973 344
kfraser@12374 345 for ( i = (exch.nr_exchanged >> in_chunk_order);
kfraser@12374 346 i < (exch.in.nr_extents >> in_chunk_order);
kfraser@12374 347 i++ )
kfraser@10418 348 {
kfraser@10418 349 if ( hypercall_preempt_check() )
kfraser@10418 350 {
kfraser@12374 351 exch.nr_exchanged = i << in_chunk_order;
kfraser@10418 352 if ( copy_field_to_guest(arg, &exch, nr_exchanged) )
kfraser@10418 353 return -EFAULT;
kfraser@10418 354 return hypercall_create_continuation(
kfraser@10418 355 __HYPERVISOR_memory_op, "lh", XENMEM_exchange, arg);
kfraser@10418 356 }
kfraser@10418 357
kfraser@10418 358 /* Steal a chunk's worth of input pages from the domain. */
kfraser@10418 359 for ( j = 0; j < (1UL << in_chunk_order); j++ )
kfraser@10418 360 {
kfraser@10418 361 if ( unlikely(__copy_from_guest_offset(
kfraser@10418 362 &gmfn, exch.in.extent_start, (i<<in_chunk_order)+j, 1)) )
kfraser@10418 363 {
kfraser@10418 364 rc = -EFAULT;
kfraser@10418 365 goto fail;
kfraser@10418 366 }
kfraser@10418 367
kfraser@10418 368 for ( k = 0; k < (1UL << exch.in.extent_order); k++ )
kfraser@10418 369 {
kfraser@10418 370 mfn = gmfn_to_mfn(d, gmfn + k);
kfraser@10418 371 if ( unlikely(!mfn_valid(mfn)) )
kfraser@10418 372 {
kfraser@10418 373 rc = -EINVAL;
kfraser@10418 374 goto fail;
kfraser@10418 375 }
kfraser@10418 376
kfraser@10418 377 page = mfn_to_page(mfn);
kfraser@10418 378
kfraser@10418 379 if ( unlikely(steal_page(d, page, MEMF_no_refcount)) )
kfraser@10418 380 {
kfraser@10418 381 rc = -EINVAL;
kfraser@10418 382 goto fail;
kfraser@10418 383 }
kfraser@10418 384
kfraser@10418 385 list_add(&page->list, &in_chunk_list);
kfraser@10418 386 }
kfraser@10418 387 }
kfraser@10418 388
kfraser@10418 389 /* Allocate a chunk's worth of anonymous output pages. */
kfraser@10418 390 for ( j = 0; j < (1UL << out_chunk_order); j++ )
kfraser@10418 391 {
keir@17385 392 page = alloc_domheap_pages(NULL, exch.out.extent_order, memflags);
kfraser@10418 393 if ( unlikely(page == NULL) )
kfraser@10418 394 {
kfraser@10418 395 rc = -ENOMEM;
kfraser@10418 396 goto fail;
kfraser@10418 397 }
kfraser@10418 398
kfraser@10418 399 list_add(&page->list, &out_chunk_list);
kfraser@10418 400 }
kfraser@10418 401
kfraser@10418 402 /*
kfraser@10418 403 * Success! Beyond this point we cannot fail for this chunk.
kfraser@10418 404 */
kfraser@10418 405
kfraser@10418 406 /* Destroy final reference to each input page. */
kfraser@10418 407 while ( !list_empty(&in_chunk_list) )
kfraser@10418 408 {
kfraser@10418 409 page = list_entry(in_chunk_list.next, struct page_info, list);
kfraser@10418 410 list_del(&page->list);
kfraser@10418 411 if ( !test_and_clear_bit(_PGC_allocated, &page->count_info) )
kfraser@10418 412 BUG();
kfraser@10418 413 mfn = page_to_mfn(page);
keir@17664 414 guest_physmap_remove_page(d, mfn_to_gmfn(d, mfn), mfn, 0);
kfraser@10418 415 put_page(page);
kfraser@10418 416 }
kfraser@10418 417
kfraser@10418 418 /* Assign each output page to the domain. */
kfraser@10418 419 j = 0;
kfraser@10418 420 while ( !list_empty(&out_chunk_list) )
kfraser@10418 421 {
kfraser@10418 422 page = list_entry(out_chunk_list.next, struct page_info, list);
kfraser@10418 423 list_del(&page->list);
kfraser@10418 424 if ( assign_pages(d, page, exch.out.extent_order,
kfraser@10418 425 MEMF_no_refcount) )
kfraser@10418 426 BUG();
kfraser@10418 427
kfraser@10418 428 /* Note that we ignore errors accessing the output extent list. */
kfraser@10418 429 (void)__copy_from_guest_offset(
kfraser@10418 430 &gpfn, exch.out.extent_start, (i<<out_chunk_order)+j, 1);
kfraser@10418 431
kfraser@10418 432 mfn = page_to_mfn(page);
keir@17727 433 guest_physmap_add_page(d, gpfn, mfn, exch.out.extent_order);
keir@17727 434
keir@17727 435 if ( !paging_mode_translate(d) )
kfraser@10418 436 {
kfraser@10418 437 for ( k = 0; k < (1UL << exch.out.extent_order); k++ )
kfraser@10418 438 set_gpfn_from_mfn(mfn + k, gpfn + k);
kfraser@10418 439 (void)__copy_to_guest_offset(
kfraser@10418 440 exch.out.extent_start, (i<<out_chunk_order)+j, &mfn, 1);
kfraser@10418 441 }
kfraser@10418 442
kfraser@10418 443 j++;
kfraser@10418 444 }
kfraser@10418 445 BUG_ON(j != (1UL << out_chunk_order));
kfraser@10418 446 }
kfraser@10418 447
kfraser@12374 448 exch.nr_exchanged = exch.in.nr_extents;
kfraser@10418 449 if ( copy_field_to_guest(arg, &exch, nr_exchanged) )
kfraser@10418 450 rc = -EFAULT;
kfraser@10418 451 return rc;
kfraser@10418 452
kfraser@10418 453 /*
kfraser@10418 454 * Failed a chunk! Free any partial chunk work. Tell caller how many
kfraser@10418 455 * chunks succeeded.
kfraser@10418 456 */
kfraser@10418 457 fail:
kfraser@10418 458 /* Reassign any input pages we managed to steal. */
kfraser@10418 459 while ( !list_empty(&in_chunk_list) )
kfraser@10418 460 {
kfraser@10418 461 page = list_entry(in_chunk_list.next, struct page_info, list);
kfraser@10418 462 list_del(&page->list);
kfraser@10418 463 if ( assign_pages(d, page, 0, MEMF_no_refcount) )
kfraser@10418 464 BUG();
kfraser@10418 465 }
kfraser@10418 466
kfraser@10418 467 /* Free any output pages we managed to allocate. */
kfraser@10418 468 while ( !list_empty(&out_chunk_list) )
kfraser@10418 469 {
kfraser@10418 470 page = list_entry(out_chunk_list.next, struct page_info, list);
kfraser@10418 471 list_del(&page->list);
kfraser@10418 472 free_domheap_pages(page, exch.out.extent_order);
kfraser@10418 473 }
kfraser@10418 474
kfraser@12374 475 exch.nr_exchanged = i << in_chunk_order;
kfraser@10418 476
kfraser@10418 477 fail_early:
kfraser@10418 478 if ( copy_field_to_guest(arg, &exch, nr_exchanged) )
kfraser@10418 479 rc = -EFAULT;
kfraser@10418 480 return rc;
kfraser@10418 481 }
kfraser@10418 482
kaf24@9873 483 long do_memory_op(unsigned long cmd, XEN_GUEST_HANDLE(void) arg)
kaf24@6486 484 {
kaf24@6486 485 struct domain *d;
kfraser@12374 486 int rc, op;
keir@17986 487 unsigned int address_bits;
kaf24@8871 488 unsigned long start_extent, progress;
kaf24@6486 489 struct xen_memory_reservation reservation;
kfraser@12374 490 struct memop_args args;
kaf24@7959 491 domid_t domid;
kaf24@6486 492
ack@13295 493 op = cmd & MEMOP_CMD_MASK;
kaf24@6486 494
kaf24@6486 495 switch ( op )
kaf24@6486 496 {
kaf24@6486 497 case XENMEM_increase_reservation:
kaf24@6486 498 case XENMEM_decrease_reservation:
kaf24@8673 499 case XENMEM_populate_physmap:
ack@13295 500 start_extent = cmd >> MEMOP_EXTENT_SHIFT;
kfraser@10418 501
kaf24@9068 502 if ( copy_from_guest(&reservation, arg, 1) )
kfraser@10418 503 return start_extent;
kaf24@6486 504
kaf24@8871 505 /* Is size too large for us to encode a continuation? */
ack@13295 506 if ( reservation.nr_extents > (ULONG_MAX >> MEMOP_EXTENT_SHIFT) )
kfraser@10418 507 return start_extent;
kaf24@8871 508
kaf24@6486 509 if ( unlikely(start_extent > reservation.nr_extents) )
kfraser@10418 510 return start_extent;
kaf24@9068 511
kfraser@12374 512 args.extent_list = reservation.extent_start;
kfraser@12374 513 args.nr_extents = reservation.nr_extents;
kfraser@12374 514 args.extent_order = reservation.extent_order;
kfraser@12374 515 args.nr_done = start_extent;
kfraser@12374 516 args.preempted = 0;
kfraser@12374 517 args.memflags = 0;
kaf24@6486 518
keir@17986 519 address_bits = XENMEMF_get_address_bits(reservation.mem_flags);
keir@17986 520 if ( (address_bits != 0) &&
keir@17986 521 (address_bits < (get_order_from_pages(max_page) + PAGE_SHIFT)) )
kaf24@6486 522 {
keir@17986 523 if ( address_bits <= PAGE_SHIFT )
kfraser@10418 524 return start_extent;
keir@17986 525 args.memflags = MEMF_bits(address_bits);
kaf24@6486 526 }
kaf24@6486 527
keir@17986 528 args.memflags |= MEMF_node(XENMEMF_get_node(reservation.mem_flags));
keir@17986 529
kaf24@6486 530 if ( likely(reservation.domid == DOMID_SELF) )
keir@17349 531 {
keir@17349 532 d = rcu_lock_current_domain();
keir@17349 533 }
keir@17349 534 else
keir@17349 535 {
keir@17349 536 if ( (d = rcu_lock_domain_by_id(reservation.domid)) == NULL )
keir@16856 537 return start_extent;
keir@17349 538 if ( !IS_PRIV_FOR(current->domain, d) )
keir@17349 539 {
keir@16856 540 rcu_unlock_domain(d);
keir@16856 541 return start_extent;
keir@16856 542 }
keir@16856 543 }
kfraser@12374 544 args.domain = d;
kaf24@6486 545
kfraser@15815 546 rc = xsm_memory_adjust_reservation(current->domain, d);
kfraser@15815 547 if ( rc )
kfraser@15815 548 {
keir@17349 549 rcu_unlock_domain(d);
kfraser@15815 550 return rc;
kfraser@15815 551 }
kfraser@15815 552
kaf24@8673 553 switch ( op )
kaf24@8673 554 {
kaf24@8673 555 case XENMEM_increase_reservation:
kfraser@12374 556 increase_reservation(&args);
kaf24@8673 557 break;
kaf24@8673 558 case XENMEM_decrease_reservation:
kfraser@12374 559 decrease_reservation(&args);
kaf24@8673 560 break;
kfraser@12374 561 default: /* XENMEM_populate_physmap */
kfraser@12374 562 populate_physmap(&args);
kaf24@8673 563 break;
kaf24@8673 564 }
kaf24@6486 565
keir@17349 566 rcu_unlock_domain(d);
kaf24@6486 567
kfraser@12374 568 rc = args.nr_done;
kaf24@6486 569
kfraser@12374 570 if ( args.preempted )
kaf24@9068 571 return hypercall_create_continuation(
kaf24@9068 572 __HYPERVISOR_memory_op, "lh",
ack@13295 573 op | (rc << MEMOP_EXTENT_SHIFT), arg);
kaf24@6607 574
kaf24@6486 575 break;
kaf24@6486 576
kfraser@10418 577 case XENMEM_exchange:
kfraser@10418 578 rc = memory_exchange(guest_handle_cast(arg, xen_memory_exchange_t));
kfraser@10418 579 break;
kfraser@10418 580
kaf24@6486 581 case XENMEM_maximum_ram_page:
kaf24@7959 582 rc = max_page;
kaf24@7959 583 break;
kaf24@7959 584
kaf24@7959 585 case XENMEM_current_reservation:
kaf24@7959 586 case XENMEM_maximum_reservation:
kfraser@14471 587 case XENMEM_maximum_gpfn:
kaf24@9068 588 if ( copy_from_guest(&domid, arg, 1) )
kaf24@6486 589 return -EFAULT;
kaf24@7959 590
keir@18574 591 rc = rcu_lock_target_domain_by_id(domid, &d);
keir@18574 592 if ( rc )
keir@18574 593 return rc;
kaf24@7959 594
kfraser@15815 595 rc = xsm_memory_stat_reservation(current->domain, d);
kfraser@15815 596 if ( rc )
kfraser@15815 597 {
keir@17349 598 rcu_unlock_domain(d);
kfraser@15815 599 return rc;
kfraser@15815 600 }
kfraser@15815 601
kfraser@14471 602 switch ( op )
kfraser@14471 603 {
kfraser@14471 604 case XENMEM_current_reservation:
kfraser@14471 605 rc = d->tot_pages;
kfraser@14471 606 break;
kfraser@14471 607 case XENMEM_maximum_reservation:
kfraser@14471 608 rc = d->max_pages;
kfraser@14471 609 break;
kfraser@14471 610 default:
kfraser@14471 611 ASSERT(op == XENMEM_maximum_gpfn);
kfraser@14471 612 rc = domain_get_maximum_gpfn(d);
kfraser@14471 613 break;
kfraser@14471 614 }
kaf24@7959 615
keir@17349 616 rcu_unlock_domain(d);
kaf24@7959 617
kaf24@6486 618 break;
kaf24@6486 619
kaf24@8871 620 case XENMEM_translate_gpfn_list:
ack@13295 621 progress = cmd >> MEMOP_EXTENT_SHIFT;
kaf24@9068 622 rc = translate_gpfn_list(
kaf24@9068 623 guest_handle_cast(arg, xen_translate_gpfn_list_t),
kaf24@9068 624 &progress);
kaf24@8871 625 if ( rc == -EAGAIN )
kaf24@9068 626 return hypercall_create_continuation(
kaf24@9068 627 __HYPERVISOR_memory_op, "lh",
ack@13295 628 op | (progress << MEMOP_EXTENT_SHIFT), arg);
kaf24@8871 629 break;
kaf24@8871 630
kaf24@6486 631 default:
kaf24@8059 632 rc = arch_memory_op(op, arg);
kaf24@6486 633 break;
kaf24@6486 634 }
kaf24@6486 635
kaf24@6486 636 return rc;
kaf24@6486 637 }
kaf24@6486 638
kaf24@6486 639 /*
kaf24@6486 640 * Local variables:
kaf24@6486 641 * mode: C
kaf24@6486 642 * c-set-style: "BSD"
kaf24@6486 643 * c-basic-offset: 4
kaf24@6486 644 * tab-width: 4
kaf24@6486 645 * indent-tabs-mode: nil
kaf24@6486 646 * End:
kaf24@6486 647 */