Pull libnvdimm fixes from Dan Williams: "1/ Fixes to the libnvdimm 'pfn' device that establishes a reserved area for storing a struct page array. 2/ Fixes for dax operations on a raw block device to prevent pagecache collisions with dax mappings. 3/ A fix for pfn_t usage in vm_insert_mixed that lead to a null pointer de-reference. These have received build success notification from the kbuild robot across 153 configs and pass the latest ndctl tests" * 'libnvdimm-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/nvdimm/nvdimm: phys_to_pfn_t: use phys_addr_t mm: fix pfn_t to page conversion in vm_insert_mixed block: use DAX for partition table reads block: revert runtime dax control of the raw block device fs, block: force direct-I/O for dax-enabled block devices devm_memremap_pages: fix vmem_altmap lifetime + alignment handling libnvdimm, pfn: fix restoring memmap location libnvdimm: fix mode determination for e820 devices
Unified · Split
| @@ -434,42 +434,6 @@ bool blkdev_dax_capable(struct block_device *bdev) · 433 unmodified lines | ||
| 434 | 434 | |
| 435 | 435 | return true; |
| 436 | 436 | } |
| 437 | − | |
| 438 | −static int blkdev_daxset(struct block_device *bdev, unsigned long argp) | |
| 439 | −{ | |
| 440 | − unsigned long arg; | |
| 441 | − int rc = 0; | |
| 442 | − | |
| 443 | − if (!capable(CAP_SYS_ADMIN)) | |
| 444 | − return -EACCES; | |
| 445 | − | |
| 446 | − if (get_user(arg, (int __user *)(argp))) | |
| 447 | − return -EFAULT; | |
| 448 | − arg = !!arg; | |
| 449 | − if (arg == !!(bdev->bd_inode->i_flags & S_DAX)) | |
| 450 | − return 0; | |
| 451 | − | |
| 452 | − if (arg) | |
| 453 | − arg = S_DAX; | |
| 454 | − | |
| 455 | − if (arg && !blkdev_dax_capable(bdev)) | |
| 456 | − return -ENOTTY; | |
| 457 | − | |
| 458 | − inode_lock(bdev->bd_inode); | |
| 459 | − if (bdev->bd_map_count == 0) | |
| 460 | − inode_set_flags(bdev->bd_inode, arg, S_DAX); | |
| 461 | − else | |
| 462 | − rc = -EBUSY; | |
| 463 | − inode_unlock(bdev->bd_inode); | |
| 464 | − return rc; | |
| 465 | −} | |
| 466 | −#else | |
| 467 | −static int blkdev_daxset(struct block_device *bdev, int arg) | |
| 468 | −{ | |
| 469 | − if (arg) | |
| 470 | − return -ENOTTY; | |
| 471 | − return 0; | |
| 472 | −} | |
| 473 | 437 | #endif |
| 474 | 438 | |
| 475 | 439 | static int blkdev_flushbuf(struct block_device *bdev, fmode_t mode, |
| @@ -634,8 +598,6 @@ int blkdev_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd, · 158 unmodified lines | ||
| 634 | 598 | case BLKTRACESETUP: |
| 635 | 599 | case BLKTRACETEARDOWN: |
| 636 | 600 | return blk_trace_ioctl(bdev, cmd, argp); |
| 637 | − case BLKDAXSET: | |
| 638 | − return blkdev_daxset(bdev, arg); | |
| 639 | 601 | case BLKDAXGET: |
| 640 | 602 | return put_int(arg, !!(bdev->bd_inode->i_flags & S_DAX)); |
| 641 | 603 | break; |
| @@ -16,6 +16,7 @@ · 15 unmodified lines | ||
| 16 | 16 | #include <linux/kmod.h> |
| 17 | 17 | #include <linux/ctype.h> |
| 18 | 18 | #include <linux/genhd.h> |
| 19 | +#include <linux/dax.h> | |
| 19 | 20 | #include <linux/blktrace_api.h> |
| 20 | 21 | |
| 21 | 22 | #include "partitions/check.h" |
| @@ -550,13 +551,24 @@ int invalidate_partitions(struct gendisk *disk, struct block_device *bdev) · 528 unmodified lines | ||
| 550 | 551 | return 0; |
| 551 | 552 | } |
| 552 | 553 | |
| 553 | −unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p) | |
| 554 | +static struct page *read_pagecache_sector(struct block_device *bdev, sector_t n) | |
| 554 | 555 | { |
| 555 | 556 | struct address_space *mapping = bdev->bd_inode->i_mapping; |
| 557 | + | |
| 558 | + return read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_CACHE_SHIFT-9)), | |
| 559 | + NULL); | |
| 560 | +} | |
| 561 | + | |
| 562 | +unsigned char *read_dev_sector(struct block_device *bdev, sector_t n, Sector *p) | |
| 563 | +{ | |
| 556 | 564 | struct page *page; |
| 557 | 565 | |
| 558 | − page = read_mapping_page(mapping, (pgoff_t)(n >> (PAGE_CACHE_SHIFT-9)), | |
| 559 | − NULL); | |
| 566 | + /* don't populate page cache for dax capable devices */ | |
| 567 | + if (IS_DAX(bdev->bd_inode)) | |
| 568 | + page = read_dax_sector(bdev, n); | |
| 569 | + else | |
| 570 | + page = read_pagecache_sector(bdev, n); | |
| 571 | + | |
| 560 | 572 | if (!IS_ERR(page)) { |
| 561 | 573 | if (PageError(page)) |
| 562 | 574 | goto fail; |
| @@ -1277,10 +1277,12 @@ static ssize_t mode_show(struct device *dev, · 1276 unmodified lines | ||
| 1277 | 1277 | |
| 1278 | 1278 | device_lock(dev); |
| 1279 | 1279 | claim = ndns->claim; |
| 1280 | − if (pmem_should_map_pages(dev) || (claim && is_nd_pfn(claim))) | |
| 1281 | − mode = "memory"; | |
| 1282 | − else if (claim && is_nd_btt(claim)) | |
| 1280 | + if (claim && is_nd_btt(claim)) | |
| 1283 | 1281 | mode = "safe"; |
| 1282 | + else if (claim && is_nd_pfn(claim)) | |
| 1283 | + mode = "memory"; | |
| 1284 | + else if (!claim && pmem_should_map_pages(dev)) | |
| 1285 | + mode = "memory"; | |
| 1284 | 1286 | else |
| 1285 | 1287 | mode = "raw"; |
| 1286 | 1288 | rc = sprintf(buf, "%s\n", mode); |
| @@ -301,10 +301,8 @@ int nd_pfn_validate(struct nd_pfn *nd_pfn) · 300 unmodified lines | ||
| 301 | 301 | |
| 302 | 302 | switch (le32_to_cpu(pfn_sb->mode)) { |
| 303 | 303 | case PFN_MODE_RAM: |
| 304 | − break; | |
| 305 | 304 | case PFN_MODE_PMEM: |
| 306 | − /* TODO: allocate from PMEM support */ | |
| 307 | − return -ENOTTY; | |
| 305 | + break; | |
| 308 | 306 | default: |
| 309 | 307 | return -ENXIO; |
| 310 | 308 | } |
| @@ -1736,37 +1736,13 @@ static int blkdev_dax_pmd_fault(struct vm_area_struct *vma, unsigned long addr, · 1735 unmodified lines | ||
| 1736 | 1736 | return __dax_pmd_fault(vma, addr, pmd, flags, blkdev_get_block, NULL); |
| 1737 | 1737 | } |
| 1738 | 1738 | |
| 1739 | −static void blkdev_vm_open(struct vm_area_struct *vma) | |
| 1740 | −{ | |
| 1741 | − struct inode *bd_inode = bdev_file_inode(vma->vm_file); | |
| 1742 | − struct block_device *bdev = I_BDEV(bd_inode); | |
| 1743 | − | |
| 1744 | − inode_lock(bd_inode); | |
| 1745 | − bdev->bd_map_count++; | |
| 1746 | − inode_unlock(bd_inode); | |
| 1747 | −} | |
| 1748 | − | |
| 1749 | −static void blkdev_vm_close(struct vm_area_struct *vma) | |
| 1750 | −{ | |
| 1751 | − struct inode *bd_inode = bdev_file_inode(vma->vm_file); | |
| 1752 | − struct block_device *bdev = I_BDEV(bd_inode); | |
| 1753 | − | |
| 1754 | − inode_lock(bd_inode); | |
| 1755 | − bdev->bd_map_count--; | |
| 1756 | − inode_unlock(bd_inode); | |
| 1757 | −} | |
| 1758 | − | |
| 1759 | 1739 | static const struct vm_operations_struct blkdev_dax_vm_ops = { |
| 1760 | − .open = blkdev_vm_open, | |
| 1761 | − .close = blkdev_vm_close, | |
| 1762 | 1740 | .fault = blkdev_dax_fault, |
| 1763 | 1741 | .pmd_fault = blkdev_dax_pmd_fault, |
| 1764 | 1742 | .pfn_mkwrite = blkdev_dax_fault, |
| 1765 | 1743 | }; |
| 1766 | 1744 | |
| 1767 | 1745 | static const struct vm_operations_struct blkdev_default_vm_ops = { |
| 1768 | − .open = blkdev_vm_open, | |
| 1769 | − .close = blkdev_vm_close, | |
| 1770 | 1746 | .fault = filemap_fault, |
| 1771 | 1747 | .map_pages = filemap_map_pages, |
| 1772 | 1748 | }; |
| @@ -1774,18 +1750,14 @@ static const struct vm_operations_struct blkdev_default_vm_ops = { · 1 unmodified line | ||
| 1774 | 1750 | static int blkdev_mmap(struct file *file, struct vm_area_struct *vma) |
| 1775 | 1751 | { |
| 1776 | 1752 | struct inode *bd_inode = bdev_file_inode(file); |
| 1777 | − struct block_device *bdev = I_BDEV(bd_inode); | |
| 1778 | 1753 | |
| 1779 | 1754 | file_accessed(file); |
| 1780 | − inode_lock(bd_inode); | |
| 1781 | − bdev->bd_map_count++; | |
| 1782 | 1755 | if (IS_DAX(bd_inode)) { |
| 1783 | 1756 | vma->vm_ops = &blkdev_dax_vm_ops; |
| 1784 | 1757 | vma->vm_flags |= VM_MIXEDMAP | VM_HUGEPAGE; |
| 1785 | 1758 | } else { |
| 1786 | 1759 | vma->vm_ops = &blkdev_default_vm_ops; |
| 1787 | 1760 | } |
| 1788 | − inode_unlock(bd_inode); | |
| 1789 | 1761 | |
| 1790 | 1762 | return 0; |
| 1791 | 1763 | } |
| @@ -58,6 +58,26 @@ static void dax_unmap_atomic(struct block_device *bdev, · 57 unmodified lines | ||
| 58 | 58 | blk_queue_exit(bdev->bd_queue); |
| 59 | 59 | } |
| 60 | 60 | |
| 61 | +struct page *read_dax_sector(struct block_device *bdev, sector_t n) | |
| 62 | +{ | |
| 63 | + struct page *page = alloc_pages(GFP_KERNEL, 0); | |
| 64 | + struct blk_dax_ctl dax = { | |
| 65 | + .size = PAGE_SIZE, | |
| 66 | + .sector = n & ~((((int) PAGE_SIZE) / 512) - 1), | |
| 67 | + }; | |
| 68 | + long rc; | |
| 69 | + | |
| 70 | + if (!page) | |
| 71 | + return ERR_PTR(-ENOMEM); | |
| 72 | + | |
| 73 | + rc = dax_map_atomic(bdev, &dax); | |
| 74 | + if (rc < 0) | |
| 75 | + return ERR_PTR(rc); | |
| 76 | + memcpy_from_pmem(page_address(page), dax.addr, PAGE_SIZE); | |
| 77 | + dax_unmap_atomic(bdev, &dax); | |
| 78 | + return page; | |
| 79 | +} | |
| 80 | + | |
| 61 | 81 | /* |
| 62 | 82 | * dax_clear_blocks() is called from within transaction context from XFS, |
| 63 | 83 | * and hence this means the stack from this point must follow GFP_NOFS |
| @@ -14,6 +14,17 @@ int dax_fault(struct vm_area_struct *, struct vm_fault *, get_block_t, · 13 unmodified lines | ||
| 14 | 14 | dax_iodone_t); |
| 15 | 15 | int __dax_fault(struct vm_area_struct *, struct vm_fault *, get_block_t, |
| 16 | 16 | dax_iodone_t); |
| 17 | + | |
| 18 | +#ifdef CONFIG_FS_DAX | |
| 19 | +struct page *read_dax_sector(struct block_device *bdev, sector_t n); | |
| 20 | +#else | |
| 21 | +static inline struct page *read_dax_sector(struct block_device *bdev, | |
| 22 | + sector_t n) | |
| 23 | +{ | |
| 24 | + return ERR_PTR(-ENXIO); | |
| 25 | +} | |
| 26 | +#endif | |
| 27 | + | |
| 17 | 28 | #ifdef CONFIG_TRANSPARENT_HUGEPAGE |
| 18 | 29 | int dax_pmd_fault(struct vm_area_struct *, unsigned long addr, pmd_t *, |
| 19 | 30 | unsigned int flags, get_block_t, dax_iodone_t); |
| @@ -484,9 +484,6 @@ struct block_device { · 483 unmodified lines | ||
| 484 | 484 | int bd_fsfreeze_count; |
| 485 | 485 | /* Mutex for freeze */ |
| 486 | 486 | struct mutex bd_fsfreeze_mutex; |
| 487 | −#ifdef CONFIG_FS_DAX | |
| 488 | − int bd_map_count; | |
| 489 | −#endif | |
| 490 | 487 | }; |
| 491 | 488 | |
| 492 | 489 | /* |
| @@ -2907,7 +2904,7 @@ extern void replace_mount_options(struct super_block *sb, char *options); · 2414 unmodified lines | ||
| 2907 | 2904 | |
| 2908 | 2905 | static inline bool io_is_direct(struct file *filp) |
| 2909 | 2906 | { |
| 2910 | − return (filp->f_flags & O_DIRECT) || IS_DAX(file_inode(filp)); | |
| 2907 | + return (filp->f_flags & O_DIRECT) || IS_DAX(filp->f_mapping->host); | |
| 2911 | 2908 | } |
| 2912 | 2909 | |
| 2913 | 2910 | static inline int iocb_flags(struct file *file) |
| @@ -29,7 +29,7 @@ static inline pfn_t pfn_to_pfn_t(unsigned long pfn) · 28 unmodified lines | ||
| 29 | 29 | return __pfn_to_pfn_t(pfn, 0); |
| 30 | 30 | } |
| 31 | 31 | |
| 32 | −extern pfn_t phys_to_pfn_t(dma_addr_t addr, unsigned long flags); | |
| 32 | +extern pfn_t phys_to_pfn_t(phys_addr_t addr, unsigned long flags); | |
| 33 | 33 | |
| 34 | 34 | static inline bool pfn_t_has_page(pfn_t pfn) |
| 35 | 35 | { |
| @@ -48,7 +48,7 @@ static inline struct page *pfn_t_to_page(pfn_t pfn) · 12 unmodified lines | ||
| 48 | 48 | return NULL; |
| 49 | 49 | } |
| 50 | 50 | |
| 51 | −static inline dma_addr_t pfn_t_to_phys(pfn_t pfn) | |
| 51 | +static inline phys_addr_t pfn_t_to_phys(pfn_t pfn) | |
| 52 | 52 | { |
| 53 | 53 | return PFN_PHYS(pfn_t_to_pfn(pfn)); |
| 54 | 54 | } |
| @@ -222,7 +222,6 @@ struct fsxattr { · 221 unmodified lines | ||
| 222 | 222 | #define BLKSECDISCARD _IO(0x12,125) |
| 223 | 223 | #define BLKROTATIONAL _IO(0x12,126) |
| 224 | 224 | #define BLKZEROOUT _IO(0x12,127) |
| 225 | −#define BLKDAXSET _IO(0x12,128) | |
| 226 | 225 | #define BLKDAXGET _IO(0x12,129) |
| 227 | 226 | |
| 228 | 227 | #define BMAP_IOCTL 1 /* obsolete - kept for compatibility */ |
| @@ -150,7 +150,7 @@ void devm_memunmap(struct device *dev, void *addr) · 149 unmodified lines | ||
| 150 | 150 | } |
| 151 | 151 | EXPORT_SYMBOL(devm_memunmap); |
| 152 | 152 | |
| 153 | −pfn_t phys_to_pfn_t(dma_addr_t addr, unsigned long flags) | |
| 153 | +pfn_t phys_to_pfn_t(phys_addr_t addr, unsigned long flags) | |
| 154 | 154 | { |
| 155 | 155 | return __pfn_to_pfn_t(addr >> PAGE_SHIFT, flags); |
| 156 | 156 | } |
| @@ -183,7 +183,11 @@ EXPORT_SYMBOL(put_zone_device_page); · 26 unmodified lines | ||
| 183 | 183 | |
| 184 | 184 | static void pgmap_radix_release(struct resource *res) |
| 185 | 185 | { |
| 186 | − resource_size_t key; | |
| 186 | + resource_size_t key, align_start, align_size, align_end; | |
| 187 | + | |
| 188 | + align_start = res->start & ~(SECTION_SIZE - 1); | |
| 189 | + align_size = ALIGN(resource_size(res), SECTION_SIZE); | |
| 190 | + align_end = align_start + align_size - 1; | |
| 187 | 191 | |
| 188 | 192 | mutex_lock(&pgmap_lock); |
| 189 | 193 | for (key = res->start; key <= res->end; key += SECTION_SIZE) |
| @@ -226,12 +230,11 @@ static void devm_memremap_pages_release(struct device *dev, void *data) · 36 unmodified lines | ||
| 226 | 230 | percpu_ref_put(pgmap->ref); |
| 227 | 231 | } |
| 228 | 232 | |
| 229 | − pgmap_radix_release(res); | |
| 230 | − | |
| 231 | 233 | /* pages are dead and unused, undo the arch mapping */ |
| 232 | 234 | align_start = res->start & ~(SECTION_SIZE - 1); |
| 233 | 235 | align_size = ALIGN(resource_size(res), SECTION_SIZE); |
| 234 | 236 | arch_remove_memory(align_start, align_size); |
| 237 | + pgmap_radix_release(res); | |
| 235 | 238 | dev_WARN_ONCE(dev, pgmap->altmap && pgmap->altmap->alloc, |
| 236 | 239 | "%s: failed to free all reserved pages\n", __func__); |
| 237 | 240 | } |
| @@ -267,7 +270,7 @@ void *devm_memremap_pages(struct device *dev, struct resource *res, · 29 unmodified lines | ||
| 267 | 270 | { |
| 268 | 271 | int is_ram = region_intersects(res->start, resource_size(res), |
| 269 | 272 | "System RAM"); |
| 270 | − resource_size_t key, align_start, align_size; | |
| 273 | + resource_size_t key, align_start, align_size, align_end; | |
| 271 | 274 | struct dev_pagemap *pgmap; |
| 272 | 275 | struct page_map *page_map; |
| 273 | 276 | unsigned long pfn; |
| @@ -309,7 +312,10 @@ void *devm_memremap_pages(struct device *dev, struct resource *res, · 35 unmodified lines | ||
| 309 | 312 | |
| 310 | 313 | mutex_lock(&pgmap_lock); |
| 311 | 314 | error = 0; |
| 312 | − for (key = res->start; key <= res->end; key += SECTION_SIZE) { | |
| 315 | + align_start = res->start & ~(SECTION_SIZE - 1); | |
| 316 | + align_size = ALIGN(resource_size(res), SECTION_SIZE); | |
| 317 | + align_end = align_start + align_size - 1; | |
| 318 | + for (key = align_start; key <= align_end; key += SECTION_SIZE) { | |
| 313 | 319 | struct dev_pagemap *dup; |
| 314 | 320 | |
| 315 | 321 | rcu_read_lock(); |
| @@ -336,8 +342,6 @@ void *devm_memremap_pages(struct device *dev, struct resource *res, · 20 unmodified lines | ||
| 336 | 342 | if (nid < 0) |
| 337 | 343 | nid = numa_mem_id(); |
| 338 | 344 | |
| 339 | − align_start = res->start & ~(SECTION_SIZE - 1); | |
| 340 | − align_size = ALIGN(resource_size(res), SECTION_SIZE); | |
| 341 | 345 | error = arch_add_memory(nid, align_start, align_size, true); |
| 342 | 346 | if (error) |
| 343 | 347 | goto err_add_memory; |
| @@ -1591,10 +1591,15 @@ int vm_insert_mixed(struct vm_area_struct *vma, unsigned long addr, · 1590 unmodified lines | ||
| 1591 | 1591 | * than insert_pfn). If a zero_pfn were inserted into a VM_MIXEDMAP |
| 1592 | 1592 | * without pte special, it would there be refcounted as a normal page. |
| 1593 | 1593 | */ |
| 1594 | − if (!HAVE_PTE_SPECIAL && pfn_t_valid(pfn)) { | |
| 1594 | + if (!HAVE_PTE_SPECIAL && !pfn_t_devmap(pfn) && pfn_t_valid(pfn)) { | |
| 1595 | 1595 | struct page *page; |
| 1596 | 1596 | |
| 1597 | − page = pfn_t_to_page(pfn); | |
| 1597 | + /* | |
| 1598 | + * At this point we are committed to insert_page() | |
| 1599 | + * regardless of whether the caller specified flags that | |
| 1600 | + * result in pfn_t_has_page() == false. | |
| 1601 | + */ | |
| 1602 | + page = pfn_to_page(pfn_t_to_pfn(pfn)); | |
| 1598 | 1603 | return insert_page(vma, addr, page, vma->vm_page_prot); |
| 1599 | 1604 | } |
| 1600 | 1605 | return insert_pfn(vma, addr, pfn, vma->vm_page_prot); |
| @@ -113,7 +113,7 @@ void *__wrap_devm_memremap_pages(struct device *dev, struct resource *res, · 112 unmodified lines | ||
| 113 | 113 | } |
| 114 | 114 | EXPORT_SYMBOL(__wrap_devm_memremap_pages); |
| 115 | 115 | |
| 116 | −pfn_t __wrap_phys_to_pfn_t(dma_addr_t addr, unsigned long flags) | |
| 116 | +pfn_t __wrap_phys_to_pfn_t(phys_addr_t addr, unsigned long flags) | |
| 117 | 117 | { |
| 118 | 118 | struct nfit_test_resource *nfit_res = get_nfit_res(addr); |
| 119 | 119 | |