| 1 | /* |
| 2 | * This program is free software; you can redistribute it and/or modify |
| 3 | * it under the terms of the GNU General Public License, version 2, as |
| 4 | * published by the Free Software Foundation. |
| 5 | * |
| 6 | * This program is distributed in the hope that it will be useful, |
| 7 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 8 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 9 | * GNU General Public License for more details. |
| 10 | * |
| 11 | * You should have received a copy of the GNU General Public License |
| 12 | * along with this program; if not, write to the Free Software |
| 13 | * Foundation, 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA. |
| 14 | * |
| 15 | * Copyright 2010 Paul Mackerras, IBM Corp. <paulus@au1.ibm.com> |
| 16 | * Copyright 2011 David Gibson, IBM Corporation <dwg@au1.ibm.com> |
| 17 | * Copyright 2016 Alexey Kardashevskiy, IBM Corporation <aik@au1.ibm.com> |
| 18 | */ |
| 19 | |
| 20 | #include <linux/types.h> |
| 21 | #include <linux/string.h> |
| 22 | #include <linux/kvm.h> |
| 23 | #include <linux/kvm_host.h> |
| 24 | #include <linux/highmem.h> |
| 25 | #include <linux/gfp.h> |
| 26 | #include <linux/slab.h> |
| 27 | #include <linux/sched/signal.h> |
| 28 | #include <linux/hugetlb.h> |
| 29 | #include <linux/list.h> |
| 30 | #include <linux/anon_inodes.h> |
| 31 | #include <linux/iommu.h> |
| 32 | #include <linux/file.h> |
| 33 | |
| 34 | #include <asm/kvm_ppc.h> |
| 35 | #include <asm/kvm_book3s.h> |
| 36 | #include <asm/book3s/64/mmu-hash.h> |
| 37 | #include <asm/hvcall.h> |
| 38 | #include <asm/synch.h> |
| 39 | #include <asm/ppc-opcode.h> |
| 40 | #include <asm/kvm_host.h> |
| 41 | #include <asm/udbg.h> |
| 42 | #include <asm/iommu.h> |
| 43 | #include <asm/tce.h> |
| 44 | #include <asm/mmu_context.h> |
| 45 | |
| 46 | static unsigned long kvmppc_tce_pages(unsigned long iommu_pages) |
| 47 | { |
| 48 | return ALIGN(iommu_pages * sizeof(u64), PAGE_SIZE) / PAGE_SIZE; |
| 49 | } |
| 50 | |
| 51 | static unsigned long kvmppc_stt_pages(unsigned long tce_pages) |
| 52 | { |
| 53 | unsigned long stt_bytes = sizeof(struct kvmppc_spapr_tce_table) + |
| 54 | (tce_pages * sizeof(struct page *)); |
| 55 | |
| 56 | return tce_pages + ALIGN(stt_bytes, PAGE_SIZE) / PAGE_SIZE; |
| 57 | } |
| 58 | |
| 59 | static long kvmppc_account_memlimit(unsigned long stt_pages, bool inc) |
| 60 | { |
| 61 | long ret = 0; |
| 62 | |
| 63 | if (!current || !current->mm) |
| 64 | return ret; /* process exited */ |
| 65 | |
| 66 | down_write(¤t->mm->mmap_sem); |
| 67 | |
| 68 | if (inc) { |
| 69 | unsigned long locked, lock_limit; |
| 70 | |
| 71 | locked = current->mm->locked_vm + stt_pages; |
| 72 | lock_limit = rlimit(RLIMIT_MEMLOCK) >> PAGE_SHIFT; |
| 73 | if (locked > lock_limit && !capable(CAP_IPC_LOCK)) |
| 74 | ret = -ENOMEM; |
| 75 | else |
| 76 | current->mm->locked_vm += stt_pages; |
| 77 | } else { |
| 78 | if (WARN_ON_ONCE(stt_pages > current->mm->locked_vm)) |
| 79 | stt_pages = current->mm->locked_vm; |
| 80 | |
| 81 | current->mm->locked_vm -= stt_pages; |
| 82 | } |
| 83 | |
| 84 | pr_debug("[%d] RLIMIT_MEMLOCK KVM %c%ld %ld/%ld%s\n", current->pid, |
| 85 | inc ? '+' : '-', |
| 86 | stt_pages << PAGE_SHIFT, |
| 87 | current->mm->locked_vm << PAGE_SHIFT, |
| 88 | rlimit(RLIMIT_MEMLOCK), |
| 89 | ret ? " - exceeded" : ""); |
| 90 | |
| 91 | up_write(¤t->mm->mmap_sem); |
| 92 | |
| 93 | return ret; |
| 94 | } |
| 95 | |
| 96 | static void kvm_spapr_tce_iommu_table_free(struct rcu_head *head) |
| 97 | { |
| 98 | struct kvmppc_spapr_tce_iommu_table *stit = container_of(head, |
| 99 | struct kvmppc_spapr_tce_iommu_table, rcu); |
| 100 | |
| 101 | iommu_tce_table_put(stit->tbl); |
| 102 | |
| 103 | kfree(stit); |
| 104 | } |
| 105 | |
| 106 | static void kvm_spapr_tce_liobn_put(struct kref *kref) |
| 107 | { |
| 108 | struct kvmppc_spapr_tce_iommu_table *stit = container_of(kref, |
| 109 | struct kvmppc_spapr_tce_iommu_table, kref); |
| 110 | |
| 111 | list_del_rcu(&stit->next); |
| 112 | |
| 113 | call_rcu(&stit->rcu, kvm_spapr_tce_iommu_table_free); |
| 114 | } |
| 115 | |
| 116 | extern void kvm_spapr_tce_release_iommu_group(struct kvm *kvm, |
| 117 | struct iommu_group *grp) |
| 118 | { |
| 119 | int i; |
| 120 | struct kvmppc_spapr_tce_table *stt; |
| 121 | struct kvmppc_spapr_tce_iommu_table *stit, *tmp; |
| 122 | struct iommu_table_group *table_group = NULL; |
| 123 | |
| 124 | list_for_each_entry_rcu(stt, &kvm->arch.spapr_tce_tables, list) { |
| 125 | |
| 126 | table_group = iommu_group_get_iommudata(grp); |
| 127 | if (WARN_ON(!table_group)) |
| 128 | continue; |
| 129 | |
| 130 | list_for_each_entry_safe(stit, tmp, &stt->iommu_tables, next) { |
| 131 | for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i) { |
| 132 | if (table_group->tables[i] != stit->tbl) |
| 133 | continue; |
| 134 | |
| 135 | kref_put(&stit->kref, kvm_spapr_tce_liobn_put); |
| 136 | } |
| 137 | } |
| 138 | } |
| 139 | } |
| 140 | |
| 141 | extern long kvm_spapr_tce_attach_iommu_group(struct kvm *kvm, int tablefd, |
| 142 | struct iommu_group *grp) |
| 143 | { |
| 144 | struct kvmppc_spapr_tce_table *stt = NULL; |
| 145 | bool found = false; |
| 146 | struct iommu_table *tbl = NULL; |
| 147 | struct iommu_table_group *table_group; |
| 148 | long i; |
| 149 | struct kvmppc_spapr_tce_iommu_table *stit; |
| 150 | struct fd f; |
| 151 | |
| 152 | f = fdget(tablefd); |
| 153 | if (!f.file) |
| 154 | return -EBADF; |
| 155 | |
| 156 | list_for_each_entry_rcu(stt, &kvm->arch.spapr_tce_tables, list) { |
| 157 | if (stt == f.file->private_data) { |
| 158 | found = true; |
| 159 | break; |
| 160 | } |
| 161 | } |
| 162 | |
| 163 | fdput(f); |
| 164 | |
| 165 | if (!found) |
| 166 | return -EINVAL; |
| 167 | |
| 168 | table_group = iommu_group_get_iommudata(grp); |
| 169 | if (WARN_ON(!table_group)) |
| 170 | return -EFAULT; |
| 171 | |
| 172 | for (i = 0; i < IOMMU_TABLE_GROUP_MAX_TABLES; ++i) { |
| 173 | struct iommu_table *tbltmp = table_group->tables[i]; |
| 174 | |
| 175 | if (!tbltmp) |
| 176 | continue; |
| 177 | /* Make sure hardware table parameters are compatible */ |
| 178 | if ((tbltmp->it_page_shift <= stt->page_shift) && |
| 179 | (tbltmp->it_offset << tbltmp->it_page_shift == |
| 180 | stt->offset << stt->page_shift) && |
| 181 | (tbltmp->it_size << tbltmp->it_page_shift >= |
| 182 | stt->size << stt->page_shift)) { |
| 183 | /* |
| 184 | * Reference the table to avoid races with |
| 185 | * add/remove DMA windows. |
| 186 | */ |
| 187 | tbl = iommu_tce_table_get(tbltmp); |
| 188 | break; |
| 189 | } |
| 190 | } |
| 191 | if (!tbl) |
| 192 | return -EINVAL; |
| 193 | |
| 194 | list_for_each_entry_rcu(stit, &stt->iommu_tables, next) { |
| 195 | if (tbl != stit->tbl) |
| 196 | continue; |
| 197 | |
| 198 | if (!kref_get_unless_zero(&stit->kref)) { |
| 199 | /* stit is being destroyed */ |
| 200 | iommu_tce_table_put(tbl); |
| 201 | return -ENOTTY; |
| 202 | } |
| 203 | /* |
| 204 | * The table is already known to this KVM, we just increased |
| 205 | * its KVM reference counter and can return. |
| 206 | */ |
| 207 | return 0; |
| 208 | } |
| 209 | |
| 210 | stit = kzalloc(sizeof(*stit), GFP_KERNEL); |
| 211 | if (!stit) { |
| 212 | iommu_tce_table_put(tbl); |
| 213 | return -ENOMEM; |
| 214 | } |
| 215 | |
| 216 | stit->tbl = tbl; |
| 217 | kref_init(&stit->kref); |
| 218 | |
| 219 | list_add_rcu(&stit->next, &stt->iommu_tables); |
| 220 | |
| 221 | return 0; |
| 222 | } |
| 223 | |
| 224 | static void release_spapr_tce_table(struct rcu_head *head) |
| 225 | { |
| 226 | struct kvmppc_spapr_tce_table *stt = container_of(head, |
| 227 | struct kvmppc_spapr_tce_table, rcu); |
| 228 | unsigned long i, npages = kvmppc_tce_pages(stt->size); |
| 229 | |
| 230 | for (i = 0; i < npages; i++) |
| 231 | __free_page(stt->pages[i]); |
| 232 | |
| 233 | kfree(stt); |
| 234 | } |
| 235 | |
| 236 | static vm_fault_t kvm_spapr_tce_fault(struct vm_fault *vmf) |
| 237 | { |
| 238 | struct kvmppc_spapr_tce_table *stt = vmf->vma->vm_file->private_data; |
| 239 | struct page *page; |
| 240 | |
| 241 | if (vmf->pgoff >= kvmppc_tce_pages(stt->size)) |
| 242 | return VM_FAULT_SIGBUS; |
| 243 | |
| 244 | page = stt->pages[vmf->pgoff]; |
| 245 | get_page(page); |
| 246 | vmf->page = page; |
| 247 | return 0; |
| 248 | } |
| 249 | |
| 250 | static const struct vm_operations_struct kvm_spapr_tce_vm_ops = { |
| 251 | .fault = kvm_spapr_tce_fault, |
| 252 | }; |
| 253 | |
| 254 | static int kvm_spapr_tce_mmap(struct file *file, struct vm_area_struct *vma) |
| 255 | { |
| 256 | vma->vm_ops = &kvm_spapr_tce_vm_ops; |
| 257 | return 0; |
| 258 | } |
| 259 | |
| 260 | static int kvm_spapr_tce_release(struct inode *inode, struct file *filp) |
| 261 | { |
| 262 | struct kvmppc_spapr_tce_table *stt = filp->private_data; |
| 263 | struct kvmppc_spapr_tce_iommu_table *stit, *tmp; |
| 264 | struct kvm *kvm = stt->kvm; |
| 265 | |
| 266 | mutex_lock(&kvm->lock); |
| 267 | list_del_rcu(&stt->list); |
| 268 | mutex_unlock(&kvm->lock); |
| 269 | |
| 270 | list_for_each_entry_safe(stit, tmp, &stt->iommu_tables, next) { |
| 271 | WARN_ON(!kref_read(&stit->kref)); |
| 272 | while (1) { |
| 273 | if (kref_put(&stit->kref, kvm_spapr_tce_liobn_put)) |
| 274 | break; |
| 275 | } |
| 276 | } |
| 277 | |
| 278 | kvm_put_kvm(stt->kvm); |
| 279 | |
| 280 | kvmppc_account_memlimit( |
| 281 | kvmppc_stt_pages(kvmppc_tce_pages(stt->size)), false); |
| 282 | call_rcu(&stt->rcu, release_spapr_tce_table); |
| 283 | |
| 284 | return 0; |
| 285 | } |
| 286 | |
| 287 | static const struct file_operations kvm_spapr_tce_fops = { |
| 288 | .mmap = kvm_spapr_tce_mmap, |
| 289 | .release = kvm_spapr_tce_release, |
| 290 | }; |
| 291 | |
| 292 | long kvm_vm_ioctl_create_spapr_tce(struct kvm *kvm, |
| 293 | struct kvm_create_spapr_tce_64 *args) |
| 294 | { |
| 295 | struct kvmppc_spapr_tce_table *stt = NULL; |
| 296 | struct kvmppc_spapr_tce_table *siter; |
| 297 | unsigned long npages, size = args->size; |
| 298 | int ret = -ENOMEM; |
| 299 | int i; |
| 300 | |
| 301 | if (!args->size || args->page_shift < 12 || args->page_shift > 34 || |
| 302 | (args->offset + args->size > (ULLONG_MAX >> args->page_shift))) |
| 303 | return -EINVAL; |
| 304 | |
| 305 | npages = kvmppc_tce_pages(size); |
| 306 | ret = kvmppc_account_memlimit(kvmppc_stt_pages(npages), true); |
| 307 | if (ret) |
| 308 | return ret; |
| 309 | |
| 310 | ret = -ENOMEM; |
| 311 | stt = kzalloc(sizeof(*stt) + npages * sizeof(struct page *), |
| 312 | GFP_KERNEL); |
| 313 | if (!stt) |
| 314 | goto fail_acct; |
| 315 | |
| 316 | stt->liobn = args->liobn; |
| 317 | stt->page_shift = args->page_shift; |
| 318 | stt->offset = args->offset; |
| 319 | stt->size = size; |
| 320 | stt->kvm = kvm; |
| 321 | INIT_LIST_HEAD_RCU(&stt->iommu_tables); |
| 322 | |
| 323 | for (i = 0; i < npages; i++) { |
| 324 | stt->pages[i] = alloc_page(GFP_KERNEL | __GFP_ZERO); |
| 325 | if (!stt->pages[i]) |
| 326 | goto fail; |
| 327 | } |
| 328 | |
| 329 | mutex_lock(&kvm->lock); |
| 330 | |
| 331 | /* Check this LIOBN hasn't been previously allocated */ |
| 332 | ret = 0; |
| 333 | list_for_each_entry(siter, &kvm->arch.spapr_tce_tables, list) { |
| 334 | if (siter->liobn == args->liobn) { |
| 335 | ret = -EBUSY; |
| 336 | break; |
| 337 | } |
| 338 | } |
| 339 | |
| 340 | kvm_get_kvm(kvm); |
| 341 | if (!ret) |
| 342 | ret = anon_inode_getfd("kvm-spapr-tce", &kvm_spapr_tce_fops, |
| 343 | stt, O_RDWR | O_CLOEXEC); |
| 344 | |
| 345 | if (ret >= 0) |
| 346 | list_add_rcu(&stt->list, &kvm->arch.spapr_tce_tables); |
| 347 | else |
| 348 | kvm_put_kvm(kvm); |
| 349 | |
| 350 | mutex_unlock(&kvm->lock); |
| 351 | |
| 352 | if (ret >= 0) |
| 353 | return ret; |
| 354 | |
| 355 | fail: |
| 356 | for (i = 0; i < npages; i++) |
| 357 | if (stt->pages[i]) |
| 358 | __free_page(stt->pages[i]); |
| 359 | |
| 360 | kfree(stt); |
| 361 | fail_acct: |
| 362 | kvmppc_account_memlimit(kvmppc_stt_pages(npages), false); |
| 363 | return ret; |
| 364 | } |
| 365 | |
| 366 | static long kvmppc_tce_validate(struct kvmppc_spapr_tce_table *stt, |
| 367 | unsigned long tce) |
| 368 | { |
| 369 | unsigned long gpa = tce & ~(TCE_PCI_READ | TCE_PCI_WRITE); |
| 370 | enum dma_data_direction dir = iommu_tce_direction(tce); |
| 371 | struct kvmppc_spapr_tce_iommu_table *stit; |
| 372 | unsigned long ua = 0; |
| 373 | |
| 374 | /* Allow userspace to poison TCE table */ |
| 375 | if (dir == DMA_NONE) |
| 376 | return H_SUCCESS; |
| 377 | |
| 378 | if (iommu_tce_check_gpa(stt->page_shift, gpa)) |
| 379 | return H_TOO_HARD; |
| 380 | |
| 381 | if (kvmppc_tce_to_ua(stt->kvm, tce, &ua, NULL)) |
| 382 | return H_TOO_HARD; |
| 383 | |
| 384 | list_for_each_entry_rcu(stit, &stt->iommu_tables, next) { |
| 385 | unsigned long hpa = 0; |
| 386 | struct mm_iommu_table_group_mem_t *mem; |
| 387 | long shift = stit->tbl->it_page_shift; |
| 388 | |
| 389 | mem = mm_iommu_lookup(stt->kvm->mm, ua, 1ULL << shift); |
| 390 | if (!mem) |
| 391 | return H_TOO_HARD; |
| 392 | |
| 393 | if (mm_iommu_ua_to_hpa(mem, ua, shift, &hpa)) |
| 394 | return H_TOO_HARD; |
| 395 | } |
| 396 | |
| 397 | return H_SUCCESS; |
| 398 | } |
| 399 | |
| 400 | static void kvmppc_clear_tce(struct mm_struct *mm, struct iommu_table *tbl, |
| 401 | unsigned long entry) |
| 402 | { |
| 403 | unsigned long hpa = 0; |
| 404 | enum dma_data_direction dir = DMA_NONE; |
| 405 | |
| 406 | iommu_tce_xchg(mm, tbl, entry, &hpa, &dir); |
| 407 | } |
| 408 | |
| 409 | static long kvmppc_tce_iommu_mapped_dec(struct kvm *kvm, |
| 410 | struct iommu_table *tbl, unsigned long entry) |
| 411 | { |
| 412 | struct mm_iommu_table_group_mem_t *mem = NULL; |
| 413 | const unsigned long pgsize = 1ULL << tbl->it_page_shift; |
| 414 | __be64 *pua = IOMMU_TABLE_USERSPACE_ENTRY_RO(tbl, entry); |
| 415 | |
| 416 | if (!pua) |
| 417 | return H_SUCCESS; |
| 418 | |
| 419 | mem = mm_iommu_lookup(kvm->mm, be64_to_cpu(*pua), pgsize); |
| 420 | if (!mem) |
| 421 | return H_TOO_HARD; |
| 422 | |
| 423 | mm_iommu_mapped_dec(mem); |
| 424 | |
| 425 | *pua = cpu_to_be64(0); |
| 426 | |
| 427 | return H_SUCCESS; |
| 428 | } |
| 429 | |
| 430 | static long kvmppc_tce_iommu_do_unmap(struct kvm *kvm, |
| 431 | struct iommu_table *tbl, unsigned long entry) |
| 432 | { |
| 433 | enum dma_data_direction dir = DMA_NONE; |
| 434 | unsigned long hpa = 0; |
| 435 | long ret; |
| 436 | |
| 437 | if (WARN_ON_ONCE(iommu_tce_xchg(kvm->mm, tbl, entry, &hpa, &dir))) |
| 438 | return H_TOO_HARD; |
| 439 | |
| 440 | if (dir == DMA_NONE) |
| 441 | return H_SUCCESS; |
| 442 | |
| 443 | ret = kvmppc_tce_iommu_mapped_dec(kvm, tbl, entry); |
| 444 | if (ret != H_SUCCESS) |
| 445 | iommu_tce_xchg(kvm->mm, tbl, entry, &hpa, &dir); |
| 446 | |
| 447 | return ret; |
| 448 | } |
| 449 | |
| 450 | static long kvmppc_tce_iommu_unmap(struct kvm *kvm, |
| 451 | struct kvmppc_spapr_tce_table *stt, struct iommu_table *tbl, |
| 452 | unsigned long entry) |
| 453 | { |
| 454 | unsigned long i, ret = H_SUCCESS; |
| 455 | unsigned long subpages = 1ULL << (stt->page_shift - tbl->it_page_shift); |
| 456 | unsigned long io_entry = entry * subpages; |
| 457 | |
| 458 | for (i = 0; i < subpages; ++i) { |
| 459 | ret = kvmppc_tce_iommu_do_unmap(kvm, tbl, io_entry + i); |
| 460 | if (ret != H_SUCCESS) |
| 461 | break; |
| 462 | } |
| 463 | |
| 464 | return ret; |
| 465 | } |
| 466 | |
| 467 | long kvmppc_tce_iommu_do_map(struct kvm *kvm, struct iommu_table *tbl, |
| 468 | unsigned long entry, unsigned long ua, |
| 469 | enum dma_data_direction dir) |
| 470 | { |
| 471 | long ret; |
| 472 | unsigned long hpa; |
| 473 | __be64 *pua = IOMMU_TABLE_USERSPACE_ENTRY(tbl, entry); |
| 474 | struct mm_iommu_table_group_mem_t *mem; |
| 475 | |
| 476 | if (!pua) |
| 477 | /* it_userspace allocation might be delayed */ |
| 478 | return H_TOO_HARD; |
| 479 | |
| 480 | mem = mm_iommu_lookup(kvm->mm, ua, 1ULL << tbl->it_page_shift); |
| 481 | if (!mem) |
| 482 | /* This only handles v2 IOMMU type, v1 is handled via ioctl() */ |
| 483 | return H_TOO_HARD; |
| 484 | |
| 485 | if (WARN_ON_ONCE(mm_iommu_ua_to_hpa(mem, ua, tbl->it_page_shift, &hpa))) |
| 486 | return H_TOO_HARD; |
| 487 | |
| 488 | if (mm_iommu_mapped_inc(mem)) |
| 489 | return H_TOO_HARD; |
| 490 | |
| 491 | ret = iommu_tce_xchg(kvm->mm, tbl, entry, &hpa, &dir); |
| 492 | if (WARN_ON_ONCE(ret)) { |
| 493 | mm_iommu_mapped_dec(mem); |
| 494 | return H_TOO_HARD; |
| 495 | } |
| 496 | |
| 497 | if (dir != DMA_NONE) |
| 498 | kvmppc_tce_iommu_mapped_dec(kvm, tbl, entry); |
| 499 | |
| 500 | *pua = cpu_to_be64(ua); |
| 501 | |
| 502 | return 0; |
| 503 | } |
| 504 | |
| 505 | static long kvmppc_tce_iommu_map(struct kvm *kvm, |
| 506 | struct kvmppc_spapr_tce_table *stt, struct iommu_table *tbl, |
| 507 | unsigned long entry, unsigned long ua, |
| 508 | enum dma_data_direction dir) |
| 509 | { |
| 510 | unsigned long i, pgoff, ret = H_SUCCESS; |
| 511 | unsigned long subpages = 1ULL << (stt->page_shift - tbl->it_page_shift); |
| 512 | unsigned long io_entry = entry * subpages; |
| 513 | |
| 514 | for (i = 0, pgoff = 0; i < subpages; |
| 515 | ++i, pgoff += IOMMU_PAGE_SIZE(tbl)) { |
| 516 | |
| 517 | ret = kvmppc_tce_iommu_do_map(kvm, tbl, |
| 518 | io_entry + i, ua + pgoff, dir); |
| 519 | if (ret != H_SUCCESS) |
| 520 | break; |
| 521 | } |
| 522 | |
| 523 | return ret; |
| 524 | } |
| 525 | |
| 526 | long kvmppc_h_put_tce(struct kvm_vcpu *vcpu, unsigned long liobn, |
| 527 | unsigned long ioba, unsigned long tce) |
| 528 | { |
| 529 | struct kvmppc_spapr_tce_table *stt; |
| 530 | long ret, idx; |
| 531 | struct kvmppc_spapr_tce_iommu_table *stit; |
| 532 | unsigned long entry, ua = 0; |
| 533 | enum dma_data_direction dir; |
| 534 | |
| 535 | /* udbg_printf("H_PUT_TCE(): liobn=0x%lx ioba=0x%lx, tce=0x%lx\n", */ |
| 536 | /* liobn, ioba, tce); */ |
| 537 | |
| 538 | stt = kvmppc_find_table(vcpu->kvm, liobn); |
| 539 | if (!stt) |
| 540 | return H_TOO_HARD; |
| 541 | |
| 542 | ret = kvmppc_ioba_validate(stt, ioba, 1); |
| 543 | if (ret != H_SUCCESS) |
| 544 | return ret; |
| 545 | |
| 546 | idx = srcu_read_lock(&vcpu->kvm->srcu); |
| 547 | |
| 548 | ret = kvmppc_tce_validate(stt, tce); |
| 549 | if (ret != H_SUCCESS) |
| 550 | goto unlock_exit; |
| 551 | |
| 552 | dir = iommu_tce_direction(tce); |
| 553 | |
| 554 | if ((dir != DMA_NONE) && kvmppc_tce_to_ua(vcpu->kvm, tce, &ua, NULL)) { |
| 555 | ret = H_PARAMETER; |
| 556 | goto unlock_exit; |
| 557 | } |
| 558 | |
| 559 | entry = ioba >> stt->page_shift; |
| 560 | |
| 561 | list_for_each_entry_lockless(stit, &stt->iommu_tables, next) { |
| 562 | if (dir == DMA_NONE) |
| 563 | ret = kvmppc_tce_iommu_unmap(vcpu->kvm, stt, |
| 564 | stit->tbl, entry); |
| 565 | else |
| 566 | ret = kvmppc_tce_iommu_map(vcpu->kvm, stt, stit->tbl, |
| 567 | entry, ua, dir); |
| 568 | |
| 569 | if (ret != H_SUCCESS) { |
| 570 | kvmppc_clear_tce(vcpu->kvm->mm, stit->tbl, entry); |
| 571 | goto unlock_exit; |
| 572 | } |
| 573 | } |
| 574 | |
| 575 | kvmppc_tce_put(stt, entry, tce); |
| 576 | |
| 577 | unlock_exit: |
| 578 | srcu_read_unlock(&vcpu->kvm->srcu, idx); |
| 579 | |
| 580 | return ret; |
| 581 | } |
| 582 | EXPORT_SYMBOL_GPL(kvmppc_h_put_tce); |
| 583 | |
| 584 | long kvmppc_h_put_tce_indirect(struct kvm_vcpu *vcpu, |
| 585 | unsigned long liobn, unsigned long ioba, |
| 586 | unsigned long tce_list, unsigned long npages) |
| 587 | { |
| 588 | struct kvmppc_spapr_tce_table *stt; |
| 589 | long i, ret = H_SUCCESS, idx; |
| 590 | unsigned long entry, ua = 0; |
| 591 | u64 __user *tces; |
| 592 | u64 tce; |
| 593 | struct kvmppc_spapr_tce_iommu_table *stit; |
| 594 | |
| 595 | stt = kvmppc_find_table(vcpu->kvm, liobn); |
| 596 | if (!stt) |
| 597 | return H_TOO_HARD; |
| 598 | |
| 599 | entry = ioba >> stt->page_shift; |
| 600 | /* |
| 601 | * SPAPR spec says that the maximum size of the list is 512 TCEs |
| 602 | * so the whole table fits in 4K page |
| 603 | */ |
| 604 | if (npages > 512) |
| 605 | return H_PARAMETER; |
| 606 | |
| 607 | if (tce_list & (SZ_4K - 1)) |
| 608 | return H_PARAMETER; |
| 609 | |
| 610 | ret = kvmppc_ioba_validate(stt, ioba, npages); |
| 611 | if (ret != H_SUCCESS) |
| 612 | return ret; |
| 613 | |
| 614 | idx = srcu_read_lock(&vcpu->kvm->srcu); |
| 615 | if (kvmppc_tce_to_ua(vcpu->kvm, tce_list, &ua, NULL)) { |
| 616 | ret = H_TOO_HARD; |
| 617 | goto unlock_exit; |
| 618 | } |
| 619 | tces = (u64 __user *) ua; |
| 620 | |
| 621 | for (i = 0; i < npages; ++i) { |
| 622 | if (get_user(tce, tces + i)) { |
| 623 | ret = H_TOO_HARD; |
| 624 | goto unlock_exit; |
| 625 | } |
| 626 | tce = be64_to_cpu(tce); |
| 627 | |
| 628 | ret = kvmppc_tce_validate(stt, tce); |
| 629 | if (ret != H_SUCCESS) |
| 630 | goto unlock_exit; |
| 631 | } |
| 632 | |
| 633 | for (i = 0; i < npages; ++i) { |
| 634 | /* |
| 635 | * This looks unsafe, because we validate, then regrab |
| 636 | * the TCE from userspace which could have been changed by |
| 637 | * another thread. |
| 638 | * |
| 639 | * But it actually is safe, because the relevant checks will be |
| 640 | * re-executed in the following code. If userspace tries to |
| 641 | * change this dodgily it will result in a messier failure mode |
| 642 | * but won't threaten the host. |
| 643 | */ |
| 644 | if (get_user(tce, tces + i)) { |
| 645 | ret = H_TOO_HARD; |
| 646 | goto unlock_exit; |
| 647 | } |
| 648 | tce = be64_to_cpu(tce); |
| 649 | |
| 650 | if (kvmppc_tce_to_ua(vcpu->kvm, tce, &ua, NULL)) |
| 651 | return H_PARAMETER; |
| 652 | |
| 653 | list_for_each_entry_lockless(stit, &stt->iommu_tables, next) { |
| 654 | ret = kvmppc_tce_iommu_map(vcpu->kvm, stt, |
| 655 | stit->tbl, entry + i, ua, |
| 656 | iommu_tce_direction(tce)); |
| 657 | |
| 658 | if (ret != H_SUCCESS) { |
| 659 | kvmppc_clear_tce(vcpu->kvm->mm, stit->tbl, |
| 660 | entry); |
| 661 | goto unlock_exit; |
| 662 | } |
| 663 | } |
| 664 | |
| 665 | kvmppc_tce_put(stt, entry + i, tce); |
| 666 | } |
| 667 | |
| 668 | unlock_exit: |
| 669 | srcu_read_unlock(&vcpu->kvm->srcu, idx); |
| 670 | |
| 671 | return ret; |
| 672 | } |
| 673 | EXPORT_SYMBOL_GPL(kvmppc_h_put_tce_indirect); |
| 674 | |
| 675 | long kvmppc_h_stuff_tce(struct kvm_vcpu *vcpu, |
| 676 | unsigned long liobn, unsigned long ioba, |
| 677 | unsigned long tce_value, unsigned long npages) |
| 678 | { |
| 679 | struct kvmppc_spapr_tce_table *stt; |
| 680 | long i, ret; |
| 681 | struct kvmppc_spapr_tce_iommu_table *stit; |
| 682 | |
| 683 | stt = kvmppc_find_table(vcpu->kvm, liobn); |
| 684 | if (!stt) |
| 685 | return H_TOO_HARD; |
| 686 | |
| 687 | ret = kvmppc_ioba_validate(stt, ioba, npages); |
| 688 | if (ret != H_SUCCESS) |
| 689 | return ret; |
| 690 | |
| 691 | /* Check permission bits only to allow userspace poison TCE for debug */ |
| 692 | if (tce_value & (TCE_PCI_WRITE | TCE_PCI_READ)) |
| 693 | return H_PARAMETER; |
| 694 | |
| 695 | list_for_each_entry_lockless(stit, &stt->iommu_tables, next) { |
| 696 | unsigned long entry = ioba >> stt->page_shift; |
| 697 | |
| 698 | for (i = 0; i < npages; ++i) { |
| 699 | ret = kvmppc_tce_iommu_unmap(vcpu->kvm, stt, |
| 700 | stit->tbl, entry + i); |
| 701 | |
| 702 | if (ret == H_SUCCESS) |
| 703 | continue; |
| 704 | |
| 705 | if (ret == H_TOO_HARD) |
| 706 | return ret; |
| 707 | |
| 708 | WARN_ON_ONCE(1); |
| 709 | kvmppc_clear_tce(vcpu->kvm->mm, stit->tbl, entry); |
| 710 | } |
| 711 | } |
| 712 | |
| 713 | for (i = 0; i < npages; ++i, ioba += (1ULL << stt->page_shift)) |
| 714 | kvmppc_tce_put(stt, ioba >> stt->page_shift, tce_value); |
| 715 | |
| 716 | return H_SUCCESS; |
| 717 | } |
| 718 | EXPORT_SYMBOL_GPL(kvmppc_h_stuff_tce); |