diff options
| author | Linus Torvalds <torvalds@linux-foundation.org> | 2008-04-25 15:48:08 -0400 |
|---|---|---|
| committer | Linus Torvalds <torvalds@linux-foundation.org> | 2008-04-25 15:48:08 -0400 |
| commit | bf16ae250999e76aff0491a362073a552db965fc (patch) | |
| tree | 9b012f0f4e9cc146648fe8914346452563f999d9 | |
| parent | 0b79dada976198cb1a4c043068e3b44d5cab2a5a (diff) | |
| parent | 1526a756fba5b1f2eb5001b8e8de2a0ea1bd2c66 (diff) | |
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-x86-pat
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/x86/linux-2.6-x86-pat:
generic: add ioremap_wc() interface wrapper
/dev/mem: make promisc the default
pat: cleanups
x86: PAT use reserve free memtype in mmap of /dev/mem
x86: PAT phys_mem_access_prot_allowed for dev/mem mmap
x86: PAT avoid aliasing in /dev/mem read/write
devmem: add range_is_allowed() check to mmap of /dev/mem
x86: introduce /dev/mem restrictions with a config option
| -rw-r--r-- | arch/x86/Kconfig.debug | 11 | ||||
| -rw-r--r-- | arch/x86/mm/init_32.c | 19 | ||||
| -rw-r--r-- | arch/x86/mm/init_64.c | 20 | ||||
| -rw-r--r-- | arch/x86/mm/ioremap.c | 29 | ||||
| -rw-r--r-- | arch/x86/mm/pat.c | 175 | ||||
| -rw-r--r-- | drivers/char/mem.c | 133 | ||||
| -rw-r--r-- | include/asm-generic/iomap.h | 4 | ||||
| -rw-r--r-- | include/asm-x86/io.h | 8 | ||||
| -rw-r--r-- | include/asm-x86/io_32.h | 6 | ||||
| -rw-r--r-- | include/asm-x86/io_64.h | 6 | ||||
| -rw-r--r-- | include/asm-x86/page.h | 1 | ||||
| -rw-r--r-- | include/asm-x86/pgtable.h | 9 |
12 files changed, 363 insertions, 58 deletions
diff --git a/arch/x86/Kconfig.debug b/arch/x86/Kconfig.debug index 610aaecc19f8..239fd9fba0a5 100644 --- a/arch/x86/Kconfig.debug +++ b/arch/x86/Kconfig.debug | |||
| @@ -5,6 +5,17 @@ config TRACE_IRQFLAGS_SUPPORT | |||
| 5 | 5 | ||
| 6 | source "lib/Kconfig.debug" | 6 | source "lib/Kconfig.debug" |
| 7 | 7 | ||
| 8 | config NONPROMISC_DEVMEM | ||
| 9 | bool "Disable promiscuous /dev/mem" | ||
| 10 | help | ||
| 11 | The /dev/mem file by default only allows userspace access to PCI | ||
| 12 | space and the BIOS code and data regions. This is sufficient for | ||
| 13 | dosemu and X and all common users of /dev/mem. With this config | ||
| 14 | option, you allow userspace access to all of memory, including | ||
| 15 | kernel and userspace memory. Accidental access to this is | ||
| 16 | obviously disasterous, but specific access can be used by people | ||
| 17 | debugging the kernel. | ||
| 18 | |||
| 8 | config EARLY_PRINTK | 19 | config EARLY_PRINTK |
| 9 | bool "Early printk" if EMBEDDED | 20 | bool "Early printk" if EMBEDDED |
| 10 | default y | 21 | default y |
diff --git a/arch/x86/mm/init_32.c b/arch/x86/mm/init_32.c index 08aa1878fad4..baf7c4f643c8 100644 --- a/arch/x86/mm/init_32.c +++ b/arch/x86/mm/init_32.c | |||
| @@ -227,6 +227,25 @@ static inline int page_kills_ppro(unsigned long pagenr) | |||
| 227 | return 0; | 227 | return 0; |
| 228 | } | 228 | } |
| 229 | 229 | ||
| 230 | /* | ||
| 231 | * devmem_is_allowed() checks to see if /dev/mem access to a certain address | ||
| 232 | * is valid. The argument is a physical page number. | ||
| 233 | * | ||
| 234 | * | ||
| 235 | * On x86, access has to be given to the first megabyte of ram because that area | ||
| 236 | * contains bios code and data regions used by X and dosemu and similar apps. | ||
| 237 | * Access has to be given to non-kernel-ram areas as well, these contain the PCI | ||
| 238 | * mmio resources as well as potential bios/acpi data regions. | ||
| 239 | */ | ||
| 240 | int devmem_is_allowed(unsigned long pagenr) | ||
| 241 | { | ||
| 242 | if (pagenr <= 256) | ||
| 243 | return 1; | ||
| 244 | if (!page_is_ram(pagenr)) | ||
| 245 | return 1; | ||
| 246 | return 0; | ||
| 247 | } | ||
| 248 | |||
| 230 | #ifdef CONFIG_HIGHMEM | 249 | #ifdef CONFIG_HIGHMEM |
| 231 | pte_t *kmap_pte; | 250 | pte_t *kmap_pte; |
| 232 | pgprot_t kmap_prot; | 251 | pgprot_t kmap_prot; |
diff --git a/arch/x86/mm/init_64.c b/arch/x86/mm/init_64.c index b798e7b92b17..0cca62663037 100644 --- a/arch/x86/mm/init_64.c +++ b/arch/x86/mm/init_64.c | |||
| @@ -663,6 +663,26 @@ EXPORT_SYMBOL_GPL(memory_add_physaddr_to_nid); | |||
| 663 | 663 | ||
| 664 | #endif /* CONFIG_MEMORY_HOTPLUG */ | 664 | #endif /* CONFIG_MEMORY_HOTPLUG */ |
| 665 | 665 | ||
| 666 | /* | ||
| 667 | * devmem_is_allowed() checks to see if /dev/mem access to a certain address | ||
| 668 | * is valid. The argument is a physical page number. | ||
| 669 | * | ||
| 670 | * | ||
| 671 | * On x86, access has to be given to the first megabyte of ram because that area | ||
| 672 | * contains bios code and data regions used by X and dosemu and similar apps. | ||
| 673 | * Access has to be given to non-kernel-ram areas as well, these contain the PCI | ||
| 674 | * mmio resources as well as potential bios/acpi data regions. | ||
| 675 | */ | ||
| 676 | int devmem_is_allowed(unsigned long pagenr) | ||
| 677 | { | ||
| 678 | if (pagenr <= 256) | ||
| 679 | return 1; | ||
| 680 | if (!page_is_ram(pagenr)) | ||
| 681 | return 1; | ||
| 682 | return 0; | ||
| 683 | } | ||
| 684 | |||
| 685 | |||
| 666 | static struct kcore_list kcore_mem, kcore_vmalloc, kcore_kernel, | 686 | static struct kcore_list kcore_mem, kcore_vmalloc, kcore_kernel, |
| 667 | kcore_modules, kcore_vsyscall; | 687 | kcore_modules, kcore_vsyscall; |
| 668 | 688 | ||
diff --git a/arch/x86/mm/ioremap.c b/arch/x86/mm/ioremap.c index 36a3f7ded626..d176b23110cc 100644 --- a/arch/x86/mm/ioremap.c +++ b/arch/x86/mm/ioremap.c | |||
| @@ -336,6 +336,35 @@ void iounmap(volatile void __iomem *addr) | |||
| 336 | } | 336 | } |
| 337 | EXPORT_SYMBOL(iounmap); | 337 | EXPORT_SYMBOL(iounmap); |
| 338 | 338 | ||
| 339 | /* | ||
| 340 | * Convert a physical pointer to a virtual kernel pointer for /dev/mem | ||
| 341 | * access | ||
| 342 | */ | ||
| 343 | void *xlate_dev_mem_ptr(unsigned long phys) | ||
| 344 | { | ||
| 345 | void *addr; | ||
| 346 | unsigned long start = phys & PAGE_MASK; | ||
| 347 | |||
| 348 | /* If page is RAM, we can use __va. Otherwise ioremap and unmap. */ | ||
| 349 | if (page_is_ram(start >> PAGE_SHIFT)) | ||
| 350 | return __va(phys); | ||
| 351 | |||
| 352 | addr = (void *)ioremap(start, PAGE_SIZE); | ||
| 353 | if (addr) | ||
| 354 | addr = (void *)((unsigned long)addr | (phys & ~PAGE_MASK)); | ||
| 355 | |||
| 356 | return addr; | ||
| 357 | } | ||
| 358 | |||
| 359 | void unxlate_dev_mem_ptr(unsigned long phys, void *addr) | ||
| 360 | { | ||
| 361 | if (page_is_ram(phys >> PAGE_SHIFT)) | ||
| 362 | return; | ||
| 363 | |||
| 364 | iounmap((void __iomem *)((unsigned long)addr & PAGE_MASK)); | ||
| 365 | return; | ||
| 366 | } | ||
| 367 | |||
| 339 | #ifdef CONFIG_X86_32 | 368 | #ifdef CONFIG_X86_32 |
| 340 | 369 | ||
| 341 | int __initdata early_ioremap_debug; | 370 | int __initdata early_ioremap_debug; |
diff --git a/arch/x86/mm/pat.c b/arch/x86/mm/pat.c index 72c0f6097402..ef8b64b89c7d 100644 --- a/arch/x86/mm/pat.c +++ b/arch/x86/mm/pat.c | |||
| @@ -11,6 +11,7 @@ | |||
| 11 | #include <linux/kernel.h> | 11 | #include <linux/kernel.h> |
| 12 | #include <linux/gfp.h> | 12 | #include <linux/gfp.h> |
| 13 | #include <linux/fs.h> | 13 | #include <linux/fs.h> |
| 14 | #include <linux/bootmem.h> | ||
| 14 | 15 | ||
| 15 | #include <asm/msr.h> | 16 | #include <asm/msr.h> |
| 16 | #include <asm/tlbflush.h> | 17 | #include <asm/tlbflush.h> |
| @@ -21,6 +22,7 @@ | |||
| 21 | #include <asm/cacheflush.h> | 22 | #include <asm/cacheflush.h> |
| 22 | #include <asm/fcntl.h> | 23 | #include <asm/fcntl.h> |
| 23 | #include <asm/mtrr.h> | 24 | #include <asm/mtrr.h> |
| 25 | #include <asm/io.h> | ||
| 24 | 26 | ||
| 25 | int pat_wc_enabled = 1; | 27 | int pat_wc_enabled = 1; |
| 26 | 28 | ||
| @@ -190,6 +192,21 @@ static int pat_x_mtrr_type(u64 start, u64 end, unsigned long prot, | |||
| 190 | return 0; | 192 | return 0; |
| 191 | } | 193 | } |
| 192 | 194 | ||
| 195 | /* | ||
| 196 | * req_type typically has one of the: | ||
| 197 | * - _PAGE_CACHE_WB | ||
| 198 | * - _PAGE_CACHE_WC | ||
| 199 | * - _PAGE_CACHE_UC_MINUS | ||
| 200 | * - _PAGE_CACHE_UC | ||
| 201 | * | ||
| 202 | * req_type will have a special case value '-1', when requester want to inherit | ||
| 203 | * the memory type from mtrr (if WB), existing PAT, defaulting to UC_MINUS. | ||
| 204 | * | ||
| 205 | * If ret_type is NULL, function will return an error if it cannot reserve the | ||
| 206 | * region with req_type. If ret_type is non-null, function will return | ||
| 207 | * available type in ret_type in case of no error. In case of any error | ||
| 208 | * it will return a negative return value. | ||
| 209 | */ | ||
| 193 | int reserve_memtype(u64 start, u64 end, unsigned long req_type, | 210 | int reserve_memtype(u64 start, u64 end, unsigned long req_type, |
| 194 | unsigned long *ret_type) | 211 | unsigned long *ret_type) |
| 195 | { | 212 | { |
| @@ -200,9 +217,14 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type, | |||
| 200 | 217 | ||
| 201 | /* Only track when pat_wc_enabled */ | 218 | /* Only track when pat_wc_enabled */ |
| 202 | if (!pat_wc_enabled) { | 219 | if (!pat_wc_enabled) { |
| 203 | if (ret_type) | 220 | /* This is identical to page table setting without PAT */ |
| 204 | *ret_type = req_type; | 221 | if (ret_type) { |
| 205 | 222 | if (req_type == -1) { | |
| 223 | *ret_type = _PAGE_CACHE_WB; | ||
| 224 | } else { | ||
| 225 | *ret_type = req_type; | ||
| 226 | } | ||
| 227 | } | ||
| 206 | return 0; | 228 | return 0; |
| 207 | } | 229 | } |
| 208 | 230 | ||
| @@ -214,8 +236,29 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type, | |||
| 214 | return 0; | 236 | return 0; |
| 215 | } | 237 | } |
| 216 | 238 | ||
| 217 | req_type &= _PAGE_CACHE_MASK; | 239 | if (req_type == -1) { |
| 218 | err = pat_x_mtrr_type(start, end, req_type, &actual_type); | 240 | /* |
| 241 | * Special case where caller wants to inherit from mtrr or | ||
| 242 | * existing pat mapping, defaulting to UC_MINUS in case of | ||
| 243 | * no match. | ||
| 244 | */ | ||
| 245 | u8 mtrr_type = mtrr_type_lookup(start, end); | ||
| 246 | if (mtrr_type == 0xFE) { /* MTRR match error */ | ||
| 247 | err = -1; | ||
| 248 | } | ||
| 249 | |||
| 250 | if (mtrr_type == MTRR_TYPE_WRBACK) { | ||
| 251 | req_type = _PAGE_CACHE_WB; | ||
| 252 | actual_type = _PAGE_CACHE_WB; | ||
| 253 | } else { | ||
| 254 | req_type = _PAGE_CACHE_UC_MINUS; | ||
| 255 | actual_type = _PAGE_CACHE_UC_MINUS; | ||
| 256 | } | ||
| 257 | } else { | ||
| 258 | req_type &= _PAGE_CACHE_MASK; | ||
| 259 | err = pat_x_mtrr_type(start, end, req_type, &actual_type); | ||
| 260 | } | ||
| 261 | |||
| 219 | if (err) { | 262 | if (err) { |
| 220 | if (ret_type) | 263 | if (ret_type) |
| 221 | *ret_type = actual_type; | 264 | *ret_type = actual_type; |
| @@ -241,7 +284,7 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type, | |||
| 241 | struct memtype *saved_ptr; | 284 | struct memtype *saved_ptr; |
| 242 | 285 | ||
| 243 | if (parse->start >= end) { | 286 | if (parse->start >= end) { |
| 244 | printk("New Entry\n"); | 287 | pr_debug("New Entry\n"); |
| 245 | list_add(&new_entry->nd, parse->nd.prev); | 288 | list_add(&new_entry->nd, parse->nd.prev); |
| 246 | new_entry = NULL; | 289 | new_entry = NULL; |
| 247 | break; | 290 | break; |
| @@ -343,7 +386,7 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type, | |||
| 343 | break; | 386 | break; |
| 344 | } | 387 | } |
| 345 | 388 | ||
| 346 | printk("Overlap at 0x%Lx-0x%Lx\n", | 389 | printk(KERN_INFO "Overlap at 0x%Lx-0x%Lx\n", |
| 347 | saved_ptr->start, saved_ptr->end); | 390 | saved_ptr->start, saved_ptr->end); |
| 348 | /* No conflict. Go ahead and add this new entry */ | 391 | /* No conflict. Go ahead and add this new entry */ |
| 349 | list_add(&new_entry->nd, &saved_ptr->nd); | 392 | list_add(&new_entry->nd, &saved_ptr->nd); |
| @@ -353,7 +396,7 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type, | |||
| 353 | } | 396 | } |
| 354 | 397 | ||
| 355 | if (err) { | 398 | if (err) { |
| 356 | printk( | 399 | printk(KERN_INFO |
| 357 | "reserve_memtype failed 0x%Lx-0x%Lx, track %s, req %s\n", | 400 | "reserve_memtype failed 0x%Lx-0x%Lx, track %s, req %s\n", |
| 358 | start, end, cattr_name(new_entry->type), | 401 | start, end, cattr_name(new_entry->type), |
| 359 | cattr_name(req_type)); | 402 | cattr_name(req_type)); |
| @@ -365,16 +408,16 @@ int reserve_memtype(u64 start, u64 end, unsigned long req_type, | |||
| 365 | if (new_entry) { | 408 | if (new_entry) { |
| 366 | /* No conflict. Not yet added to the list. Add to the tail */ | 409 | /* No conflict. Not yet added to the list. Add to the tail */ |
| 367 | list_add_tail(&new_entry->nd, &memtype_list); | 410 | list_add_tail(&new_entry->nd, &memtype_list); |
| 368 | printk("New Entry\n"); | 411 | pr_debug("New Entry\n"); |
| 369 | } | 412 | } |
| 370 | 413 | ||
| 371 | if (ret_type) { | 414 | if (ret_type) { |
| 372 | printk( | 415 | pr_debug( |
| 373 | "reserve_memtype added 0x%Lx-0x%Lx, track %s, req %s, ret %s\n", | 416 | "reserve_memtype added 0x%Lx-0x%Lx, track %s, req %s, ret %s\n", |
| 374 | start, end, cattr_name(actual_type), | 417 | start, end, cattr_name(actual_type), |
| 375 | cattr_name(req_type), cattr_name(*ret_type)); | 418 | cattr_name(req_type), cattr_name(*ret_type)); |
| 376 | } else { | 419 | } else { |
| 377 | printk( | 420 | pr_debug( |
| 378 | "reserve_memtype added 0x%Lx-0x%Lx, track %s, req %s\n", | 421 | "reserve_memtype added 0x%Lx-0x%Lx, track %s, req %s\n", |
| 379 | start, end, cattr_name(actual_type), | 422 | start, end, cattr_name(actual_type), |
| 380 | cattr_name(req_type)); | 423 | cattr_name(req_type)); |
| @@ -411,11 +454,115 @@ int free_memtype(u64 start, u64 end) | |||
| 411 | spin_unlock(&memtype_lock); | 454 | spin_unlock(&memtype_lock); |
| 412 | 455 | ||
| 413 | if (err) { | 456 | if (err) { |
| 414 | printk(KERN_DEBUG "%s:%d freeing invalid memtype %Lx-%Lx\n", | 457 | printk(KERN_INFO "%s:%d freeing invalid memtype %Lx-%Lx\n", |
| 415 | current->comm, current->pid, start, end); | 458 | current->comm, current->pid, start, end); |
| 416 | } | 459 | } |
| 417 | 460 | ||
| 418 | printk( "free_memtype request 0x%Lx-0x%Lx\n", start, end); | 461 | pr_debug("free_memtype request 0x%Lx-0x%Lx\n", start, end); |
| 419 | return err; | 462 | return err; |
| 420 | } | 463 | } |
| 421 | 464 | ||
| 465 | |||
| 466 | /* | ||
| 467 | * /dev/mem mmap interface. The memtype used for mapping varies: | ||
| 468 | * - Use UC for mappings with O_SYNC flag | ||
| 469 | * - Without O_SYNC flag, if there is any conflict in reserve_memtype, | ||
| 470 | * inherit the memtype from existing mapping. | ||
| 471 | * - Else use UC_MINUS memtype (for backward compatibility with existing | ||
| 472 | * X drivers. | ||
| 473 | */ | ||
| 474 | pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, | ||
| 475 | unsigned long size, pgprot_t vma_prot) | ||
| 476 | { | ||
| 477 | return vma_prot; | ||
| 478 | } | ||
| 479 | |||
| 480 | int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn, | ||
| 481 | unsigned long size, pgprot_t *vma_prot) | ||
| 482 | { | ||
| 483 | u64 offset = ((u64) pfn) << PAGE_SHIFT; | ||
| 484 | unsigned long flags = _PAGE_CACHE_UC_MINUS; | ||
| 485 | unsigned long ret_flags; | ||
| 486 | int retval; | ||
| 487 | |||
| 488 | if (file->f_flags & O_SYNC) { | ||
| 489 | flags = _PAGE_CACHE_UC; | ||
| 490 | } | ||
| 491 | |||
| 492 | #ifdef CONFIG_X86_32 | ||
| 493 | /* | ||
| 494 | * On the PPro and successors, the MTRRs are used to set | ||
| 495 | * memory types for physical addresses outside main memory, | ||
| 496 | * so blindly setting UC or PWT on those pages is wrong. | ||
| 497 | * For Pentiums and earlier, the surround logic should disable | ||
| 498 | * caching for the high addresses through the KEN pin, but | ||
| 499 | * we maintain the tradition of paranoia in this code. | ||
| 500 | */ | ||
| 501 | if (!pat_wc_enabled && | ||
| 502 | ! ( test_bit(X86_FEATURE_MTRR, boot_cpu_data.x86_capability) || | ||
| 503 | test_bit(X86_FEATURE_K6_MTRR, boot_cpu_data.x86_capability) || | ||
| 504 | test_bit(X86_FEATURE_CYRIX_ARR, boot_cpu_data.x86_capability) || | ||
| 505 | test_bit(X86_FEATURE_CENTAUR_MCR, boot_cpu_data.x86_capability)) && | ||
| 506 | (pfn << PAGE_SHIFT) >= __pa(high_memory)) { | ||
| 507 | flags = _PAGE_CACHE_UC; | ||
| 508 | } | ||
| 509 | #endif | ||
| 510 | |||
| 511 | /* | ||
| 512 | * With O_SYNC, we can only take UC mapping. Fail if we cannot. | ||
| 513 | * Without O_SYNC, we want to get | ||
| 514 | * - WB for WB-able memory and no other conflicting mappings | ||
| 515 | * - UC_MINUS for non-WB-able memory with no other conflicting mappings | ||
| 516 | * - Inherit from confliting mappings otherwise | ||
| 517 | */ | ||
| 518 | if (flags != _PAGE_CACHE_UC_MINUS) { | ||
| 519 | retval = reserve_memtype(offset, offset + size, flags, NULL); | ||
| 520 | } else { | ||
| 521 | retval = reserve_memtype(offset, offset + size, -1, &ret_flags); | ||
| 522 | } | ||
| 523 | |||
| 524 | if (retval < 0) | ||
| 525 | return 0; | ||
| 526 | |||
| 527 | flags = ret_flags; | ||
| 528 | |||
| 529 | if (pfn <= max_pfn_mapped && | ||
| 530 | ioremap_change_attr((unsigned long)__va(offset), size, flags) < 0) { | ||
| 531 | free_memtype(offset, offset + size); | ||
| 532 | printk(KERN_INFO | ||
| 533 | "%s:%d /dev/mem ioremap_change_attr failed %s for %Lx-%Lx\n", | ||
| 534 | current->comm, current->pid, | ||
| 535 | cattr_name(flags), | ||
| 536 | offset, offset + size); | ||
| 537 | return 0; | ||
| 538 | } | ||
| 539 | |||
| 540 | *vma_prot = __pgprot((pgprot_val(*vma_prot) & ~_PAGE_CACHE_MASK) | | ||
| 541 | flags); | ||
| 542 | return 1; | ||
| 543 | } | ||
| 544 | |||
| 545 | void map_devmem(unsigned long pfn, unsigned long size, pgprot_t vma_prot) | ||
| 546 | { | ||
| 547 | u64 addr = (u64)pfn << PAGE_SHIFT; | ||
| 548 | unsigned long flags; | ||
| 549 | unsigned long want_flags = (pgprot_val(vma_prot) & _PAGE_CACHE_MASK); | ||
| 550 | |||
| 551 | reserve_memtype(addr, addr + size, want_flags, &flags); | ||
| 552 | if (flags != want_flags) { | ||
| 553 | printk(KERN_INFO | ||
| 554 | "%s:%d /dev/mem expected mapping type %s for %Lx-%Lx, got %s\n", | ||
| 555 | current->comm, current->pid, | ||
| 556 | cattr_name(want_flags), | ||
| 557 | addr, addr + size, | ||
| 558 | cattr_name(flags)); | ||
| 559 | } | ||
| 560 | } | ||
| 561 | |||
| 562 | void unmap_devmem(unsigned long pfn, unsigned long size, pgprot_t vma_prot) | ||
| 563 | { | ||
| 564 | u64 addr = (u64)pfn << PAGE_SHIFT; | ||
| 565 | |||
| 566 | free_memtype(addr, addr + size); | ||
| 567 | } | ||
| 568 | |||
diff --git a/drivers/char/mem.c b/drivers/char/mem.c index 20070b7c573d..e83623ead441 100644 --- a/drivers/char/mem.c +++ b/drivers/char/mem.c | |||
| @@ -41,36 +41,7 @@ | |||
| 41 | */ | 41 | */ |
| 42 | static inline int uncached_access(struct file *file, unsigned long addr) | 42 | static inline int uncached_access(struct file *file, unsigned long addr) |
| 43 | { | 43 | { |
| 44 | #if defined(__i386__) && !defined(__arch_um__) | 44 | #if defined(CONFIG_IA64) |
| 45 | /* | ||
| 46 | * On the PPro and successors, the MTRRs are used to set | ||
| 47 | * memory types for physical addresses outside main memory, | ||
| 48 | * so blindly setting PCD or PWT on those pages is wrong. | ||
| 49 | * For Pentiums and earlier, the surround logic should disable | ||
| 50 | * caching for the high addresses through the KEN pin, but | ||
| 51 | * we maintain the tradition of paranoia in this code. | ||
| 52 | */ | ||
| 53 | if (file->f_flags & O_SYNC) | ||
| 54 | return 1; | ||
| 55 | return !( test_bit(X86_FEATURE_MTRR, boot_cpu_data.x86_capability) || | ||
| 56 | test_bit(X86_FEATURE_K6_MTRR, boot_cpu_data.x86_capability) || | ||
| 57 | test_bit(X86_FEATURE_CYRIX_ARR, boot_cpu_data.x86_capability) || | ||
| 58 | test_bit(X86_FEATURE_CENTAUR_MCR, boot_cpu_data.x86_capability) ) | ||
| 59 | && addr >= __pa(high_memory); | ||
| 60 | #elif defined(__x86_64__) && !defined(__arch_um__) | ||
| 61 | /* | ||
| 62 | * This is broken because it can generate memory type aliases, | ||
| 63 | * which can cause cache corruptions | ||
| 64 | * But it is only available for root and we have to be bug-to-bug | ||
| 65 | * compatible with i386. | ||
| 66 | */ | ||
| 67 | if (file->f_flags & O_SYNC) | ||
| 68 | return 1; | ||
| 69 | /* same behaviour as i386. PAT always set to cached and MTRRs control the | ||
| 70 | caching behaviour. | ||
| 71 | Hopefully a full PAT implementation will fix that soon. */ | ||
| 72 | return 0; | ||
| 73 | #elif defined(CONFIG_IA64) | ||
| 74 | /* | 45 | /* |
| 75 | * On ia64, we ignore O_SYNC because we cannot tolerate memory attribute aliases. | 46 | * On ia64, we ignore O_SYNC because we cannot tolerate memory attribute aliases. |
| 76 | */ | 47 | */ |
| @@ -108,6 +79,36 @@ static inline int valid_mmap_phys_addr_range(unsigned long pfn, size_t size) | |||
| 108 | } | 79 | } |
| 109 | #endif | 80 | #endif |
| 110 | 81 | ||
| 82 | #ifdef CONFIG_NONPROMISC_DEVMEM | ||
| 83 | static inline int range_is_allowed(unsigned long pfn, unsigned long size) | ||
| 84 | { | ||
| 85 | u64 from = ((u64)pfn) << PAGE_SHIFT; | ||
| 86 | u64 to = from + size; | ||
| 87 | u64 cursor = from; | ||
| 88 | |||
| 89 | while (cursor < to) { | ||
| 90 | if (!devmem_is_allowed(pfn)) { | ||
| 91 | printk(KERN_INFO | ||
| 92 | "Program %s tried to access /dev/mem between %Lx->%Lx.\n", | ||
| 93 | current->comm, from, to); | ||
| 94 | return 0; | ||
| 95 | } | ||
| 96 | cursor += PAGE_SIZE; | ||
| 97 | pfn++; | ||
| 98 | } | ||
| 99 | return 1; | ||
| 100 | } | ||
| 101 | #else | ||
| 102 | static inline int range_is_allowed(unsigned long pfn, unsigned long size) | ||
| 103 | { | ||
| 104 | return 1; | ||
| 105 | } | ||
| 106 | #endif | ||
| 107 | |||
| 108 | void __attribute__((weak)) unxlate_dev_mem_ptr(unsigned long phys, void *addr) | ||
| 109 | { | ||
| 110 | } | ||
| 111 | |||
| 111 | /* | 112 | /* |
| 112 | * This funcion reads the *physical* memory. The f_pos points directly to the | 113 | * This funcion reads the *physical* memory. The f_pos points directly to the |
| 113 | * memory location. | 114 | * memory location. |
| @@ -150,15 +151,25 @@ static ssize_t read_mem(struct file * file, char __user * buf, | |||
| 150 | 151 | ||
| 151 | sz = min_t(unsigned long, sz, count); | 152 | sz = min_t(unsigned long, sz, count); |
| 152 | 153 | ||
| 154 | if (!range_is_allowed(p >> PAGE_SHIFT, count)) | ||
| 155 | return -EPERM; | ||
| 156 | |||
| 153 | /* | 157 | /* |
| 154 | * On ia64 if a page has been mapped somewhere as | 158 | * On ia64 if a page has been mapped somewhere as |
| 155 | * uncached, then it must also be accessed uncached | 159 | * uncached, then it must also be accessed uncached |
| 156 | * by the kernel or data corruption may occur | 160 | * by the kernel or data corruption may occur |
| 157 | */ | 161 | */ |
| 158 | ptr = xlate_dev_mem_ptr(p); | 162 | ptr = xlate_dev_mem_ptr(p); |
| 163 | if (!ptr) | ||
| 164 | return -EFAULT; | ||
| 159 | 165 | ||
| 160 | if (copy_to_user(buf, ptr, sz)) | 166 | if (copy_to_user(buf, ptr, sz)) { |
| 167 | unxlate_dev_mem_ptr(p, ptr); | ||
| 161 | return -EFAULT; | 168 | return -EFAULT; |
| 169 | } | ||
| 170 | |||
| 171 | unxlate_dev_mem_ptr(p, ptr); | ||
| 172 | |||
| 162 | buf += sz; | 173 | buf += sz; |
| 163 | p += sz; | 174 | p += sz; |
| 164 | count -= sz; | 175 | count -= sz; |
| @@ -207,20 +218,32 @@ static ssize_t write_mem(struct file * file, const char __user * buf, | |||
| 207 | 218 | ||
| 208 | sz = min_t(unsigned long, sz, count); | 219 | sz = min_t(unsigned long, sz, count); |
| 209 | 220 | ||
| 221 | if (!range_is_allowed(p >> PAGE_SHIFT, sz)) | ||
| 222 | return -EPERM; | ||
| 223 | |||
| 210 | /* | 224 | /* |
| 211 | * On ia64 if a page has been mapped somewhere as | 225 | * On ia64 if a page has been mapped somewhere as |
| 212 | * uncached, then it must also be accessed uncached | 226 | * uncached, then it must also be accessed uncached |
| 213 | * by the kernel or data corruption may occur | 227 | * by the kernel or data corruption may occur |
| 214 | */ | 228 | */ |
| 215 | ptr = xlate_dev_mem_ptr(p); | 229 | ptr = xlate_dev_mem_ptr(p); |
| 230 | if (!ptr) { | ||
| 231 | if (written) | ||
| 232 | break; | ||
| 233 | return -EFAULT; | ||
| 234 | } | ||
| 216 | 235 | ||
| 217 | copied = copy_from_user(ptr, buf, sz); | 236 | copied = copy_from_user(ptr, buf, sz); |
| 218 | if (copied) { | 237 | if (copied) { |
| 219 | written += sz - copied; | 238 | written += sz - copied; |
| 239 | unxlate_dev_mem_ptr(p, ptr); | ||
| 220 | if (written) | 240 | if (written) |
| 221 | break; | 241 | break; |
| 222 | return -EFAULT; | 242 | return -EFAULT; |
| 223 | } | 243 | } |
| 244 | |||
| 245 | unxlate_dev_mem_ptr(p, ptr); | ||
| 246 | |||
| 224 | buf += sz; | 247 | buf += sz; |
| 225 | p += sz; | 248 | p += sz; |
| 226 | count -= sz; | 249 | count -= sz; |
| @@ -231,6 +254,12 @@ static ssize_t write_mem(struct file * file, const char __user * buf, | |||
| 231 | return written; | 254 | return written; |
| 232 | } | 255 | } |
| 233 | 256 | ||
| 257 | int __attribute__((weak)) phys_mem_access_prot_allowed(struct file *file, | ||
| 258 | unsigned long pfn, unsigned long size, pgprot_t *vma_prot) | ||
| 259 | { | ||
| 260 | return 1; | ||
| 261 | } | ||
| 262 | |||
| 234 | #ifndef __HAVE_PHYS_MEM_ACCESS_PROT | 263 | #ifndef __HAVE_PHYS_MEM_ACCESS_PROT |
| 235 | static pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, | 264 | static pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, |
| 236 | unsigned long size, pgprot_t vma_prot) | 265 | unsigned long size, pgprot_t vma_prot) |
| @@ -271,6 +300,35 @@ static inline int private_mapping_ok(struct vm_area_struct *vma) | |||
| 271 | } | 300 | } |
| 272 | #endif | 301 | #endif |
| 273 | 302 | ||
| 303 | void __attribute__((weak)) | ||
| 304 | map_devmem(unsigned long pfn, unsigned long len, pgprot_t prot) | ||
| 305 | { | ||
| 306 | /* nothing. architectures can override. */ | ||
| 307 | } | ||
| 308 | |||
| 309 | void __attribute__((weak)) | ||
| 310 | unmap_devmem(unsigned long pfn, unsigned long len, pgprot_t prot) | ||
| 311 | { | ||
| 312 | /* nothing. architectures can override. */ | ||
| 313 | } | ||
| 314 | |||
| 315 | static void mmap_mem_open(struct vm_area_struct *vma) | ||
| 316 | { | ||
| 317 | map_devmem(vma->vm_pgoff, vma->vm_end - vma->vm_start, | ||
| 318 | vma->vm_page_prot); | ||
| 319 | } | ||
| 320 | |||
| 321 | static void mmap_mem_close(struct vm_area_struct *vma) | ||
| 322 | { | ||
| 323 | unmap_devmem(vma->vm_pgoff, vma->vm_end - vma->vm_start, | ||
| 324 | vma->vm_page_prot); | ||
| 325 | } | ||
| 326 | |||
| 327 | static struct vm_operations_struct mmap_mem_ops = { | ||
| 328 | .open = mmap_mem_open, | ||
| 329 | .close = mmap_mem_close | ||
| 330 | }; | ||
| 331 | |||
| 274 | static int mmap_mem(struct file * file, struct vm_area_struct * vma) | 332 | static int mmap_mem(struct file * file, struct vm_area_struct * vma) |
| 275 | { | 333 | { |
| 276 | size_t size = vma->vm_end - vma->vm_start; | 334 | size_t size = vma->vm_end - vma->vm_start; |
| @@ -281,17 +339,28 @@ static int mmap_mem(struct file * file, struct vm_area_struct * vma) | |||
| 281 | if (!private_mapping_ok(vma)) | 339 | if (!private_mapping_ok(vma)) |
| 282 | return -ENOSYS; | 340 | return -ENOSYS; |
| 283 | 341 | ||
| 342 | if (!range_is_allowed(vma->vm_pgoff, size)) | ||
| 343 | return -EPERM; | ||
| 344 | |||
| 345 | if (!phys_mem_access_prot_allowed(file, vma->vm_pgoff, size, | ||
| 346 | &vma->vm_page_prot)) | ||
| 347 | return -EINVAL; | ||
| 348 | |||
| 284 | vma->vm_page_prot = phys_mem_access_prot(file, vma->vm_pgoff, | 349 | vma->vm_page_prot = phys_mem_access_prot(file, vma->vm_pgoff, |
| 285 | size, | 350 | size, |
| 286 | vma->vm_page_prot); | 351 | vma->vm_page_prot); |
| 287 | 352 | ||
| 353 | vma->vm_ops = &mmap_mem_ops; | ||
| 354 | |||
| 288 | /* Remap-pfn-range will mark the range VM_IO and VM_RESERVED */ | 355 | /* Remap-pfn-range will mark the range VM_IO and VM_RESERVED */ |
| 289 | if (remap_pfn_range(vma, | 356 | if (remap_pfn_range(vma, |
| 290 | vma->vm_start, | 357 | vma->vm_start, |
| 291 | vma->vm_pgoff, | 358 | vma->vm_pgoff, |
| 292 | size, | 359 | size, |
| 293 | vma->vm_page_prot)) | 360 | vma->vm_page_prot)) { |
| 361 | unmap_devmem(vma->vm_pgoff, size, vma->vm_page_prot); | ||
| 294 | return -EAGAIN; | 362 | return -EAGAIN; |
| 363 | } | ||
| 295 | return 0; | 364 | return 0; |
| 296 | } | 365 | } |
| 297 | 366 | ||
diff --git a/include/asm-generic/iomap.h b/include/asm-generic/iomap.h index 67dc84cd1343..76b0cc5637f8 100644 --- a/include/asm-generic/iomap.h +++ b/include/asm-generic/iomap.h | |||
| @@ -60,6 +60,10 @@ extern void iowrite32_rep(void __iomem *port, const void *buf, unsigned long cou | |||
| 60 | extern void __iomem *ioport_map(unsigned long port, unsigned int nr); | 60 | extern void __iomem *ioport_map(unsigned long port, unsigned int nr); |
| 61 | extern void ioport_unmap(void __iomem *); | 61 | extern void ioport_unmap(void __iomem *); |
| 62 | 62 | ||
| 63 | #ifndef ARCH_HAS_IOREMAP_WC | ||
| 64 | #define ioremap_wc ioremap_nocache | ||
| 65 | #endif | ||
| 66 | |||
| 63 | /* Create a virtual mapping cookie for a PCI BAR (memory or IO) */ | 67 | /* Create a virtual mapping cookie for a PCI BAR (memory or IO) */ |
| 64 | struct pci_dev; | 68 | struct pci_dev; |
| 65 | extern void __iomem *pci_iomap(struct pci_dev *dev, int bar, unsigned long max); | 69 | extern void __iomem *pci_iomap(struct pci_dev *dev, int bar, unsigned long max); |
diff --git a/include/asm-x86/io.h b/include/asm-x86/io.h index 7b292d386713..d5b11f60dbd0 100644 --- a/include/asm-x86/io.h +++ b/include/asm-x86/io.h | |||
| @@ -1,3 +1,6 @@ | |||
| 1 | #ifndef _ASM_X86_IO_H | ||
| 2 | #define _ASM_X86_IO_H | ||
| 3 | |||
| 1 | #define ARCH_HAS_IOREMAP_WC | 4 | #define ARCH_HAS_IOREMAP_WC |
| 2 | 5 | ||
| 3 | #ifdef CONFIG_X86_32 | 6 | #ifdef CONFIG_X86_32 |
| @@ -5,7 +8,12 @@ | |||
| 5 | #else | 8 | #else |
| 6 | # include "io_64.h" | 9 | # include "io_64.h" |
| 7 | #endif | 10 | #endif |
| 11 | |||
| 12 | extern void *xlate_dev_mem_ptr(unsigned long phys); | ||
| 13 | extern void unxlate_dev_mem_ptr(unsigned long phys, void *addr); | ||
| 14 | |||
| 8 | extern int ioremap_change_attr(unsigned long vaddr, unsigned long size, | 15 | extern int ioremap_change_attr(unsigned long vaddr, unsigned long size, |
| 9 | unsigned long prot_val); | 16 | unsigned long prot_val); |
| 10 | extern void __iomem *ioremap_wc(unsigned long offset, unsigned long size); | 17 | extern void __iomem *ioremap_wc(unsigned long offset, unsigned long size); |
| 11 | 18 | ||
| 19 | #endif /* _ASM_X86_IO_H */ | ||
diff --git a/include/asm-x86/io_32.h b/include/asm-x86/io_32.h index 509045f5fda2..6e73467a4fb1 100644 --- a/include/asm-x86/io_32.h +++ b/include/asm-x86/io_32.h | |||
| @@ -49,12 +49,6 @@ | |||
| 49 | #include <linux/vmalloc.h> | 49 | #include <linux/vmalloc.h> |
| 50 | 50 | ||
| 51 | /* | 51 | /* |
| 52 | * Convert a physical pointer to a virtual kernel pointer for /dev/mem | ||
| 53 | * access | ||
| 54 | */ | ||
| 55 | #define xlate_dev_mem_ptr(p) __va(p) | ||
| 56 | |||
| 57 | /* | ||
| 58 | * Convert a virtual cached pointer to an uncached pointer | 52 | * Convert a virtual cached pointer to an uncached pointer |
| 59 | */ | 53 | */ |
| 60 | #define xlate_dev_kmem_ptr(p) p | 54 | #define xlate_dev_kmem_ptr(p) p |
diff --git a/include/asm-x86/io_64.h b/include/asm-x86/io_64.h index c2f5eef47b88..0930bedf9e4d 100644 --- a/include/asm-x86/io_64.h +++ b/include/asm-x86/io_64.h | |||
| @@ -308,12 +308,6 @@ extern int iommu_bio_merge; | |||
| 308 | #define BIO_VMERGE_BOUNDARY iommu_bio_merge | 308 | #define BIO_VMERGE_BOUNDARY iommu_bio_merge |
| 309 | 309 | ||
| 310 | /* | 310 | /* |
| 311 | * Convert a physical pointer to a virtual kernel pointer for /dev/mem | ||
| 312 | * access | ||
| 313 | */ | ||
| 314 | #define xlate_dev_mem_ptr(p) __va(p) | ||
| 315 | |||
| 316 | /* | ||
| 317 | * Convert a virtual cached pointer to an uncached pointer | 311 | * Convert a virtual cached pointer to an uncached pointer |
| 318 | */ | 312 | */ |
| 319 | #define xlate_dev_kmem_ptr(p) p | 313 | #define xlate_dev_kmem_ptr(p) p |
diff --git a/include/asm-x86/page.h b/include/asm-x86/page.h index 6724a4bc6b7a..b381f4a5a0bd 100644 --- a/include/asm-x86/page.h +++ b/include/asm-x86/page.h | |||
| @@ -47,6 +47,7 @@ | |||
| 47 | #ifndef __ASSEMBLY__ | 47 | #ifndef __ASSEMBLY__ |
| 48 | 48 | ||
| 49 | extern int page_is_ram(unsigned long pagenr); | 49 | extern int page_is_ram(unsigned long pagenr); |
| 50 | extern int devmem_is_allowed(unsigned long pagenr); | ||
| 50 | 51 | ||
| 51 | extern unsigned long max_pfn_mapped; | 52 | extern unsigned long max_pfn_mapped; |
| 52 | 53 | ||
diff --git a/include/asm-x86/pgtable.h b/include/asm-x86/pgtable.h index b8a08bd7bd48..a496d6335d3b 100644 --- a/include/asm-x86/pgtable.h +++ b/include/asm-x86/pgtable.h | |||
| @@ -288,6 +288,15 @@ static inline pte_t pte_modify(pte_t pte, pgprot_t newprot) | |||
| 288 | 288 | ||
| 289 | #define canon_pgprot(p) __pgprot(pgprot_val(p) & __supported_pte_mask) | 289 | #define canon_pgprot(p) __pgprot(pgprot_val(p) & __supported_pte_mask) |
| 290 | 290 | ||
| 291 | #ifndef __ASSEMBLY__ | ||
| 292 | #define __HAVE_PHYS_MEM_ACCESS_PROT | ||
| 293 | struct file; | ||
| 294 | pgprot_t phys_mem_access_prot(struct file *file, unsigned long pfn, | ||
| 295 | unsigned long size, pgprot_t vma_prot); | ||
| 296 | int phys_mem_access_prot_allowed(struct file *file, unsigned long pfn, | ||
| 297 | unsigned long size, pgprot_t *vma_prot); | ||
| 298 | #endif | ||
| 299 | |||
| 291 | #ifdef CONFIG_PARAVIRT | 300 | #ifdef CONFIG_PARAVIRT |
| 292 | #include <asm/paravirt.h> | 301 | #include <asm/paravirt.h> |
| 293 | #else /* !CONFIG_PARAVIRT */ | 302 | #else /* !CONFIG_PARAVIRT */ |
