include/linux/dax.h

   1 #ifndef _LINUX_DAX_H
   2 #define _LINUX_DAX_H
   3
   4 #include <linux/fs.h>
   5 #include <linux/mm.h>
   6 #include <linux/radix-tree.h>
   7 #include <asm/pgtable.h>
   8
   9 struct iomap_ops;
  10 struct dax_device;
  11 struct dax_operations {
  12         /*
  13          * direct_access: translate a device-relative
  14          * logical-page-offset into an absolute physical pfn. Return the
  15          * number of pages available for DAX at that pfn.
  16          */
  17         long (*direct_access)(struct dax_device *, pgoff_t, long,
  18                         void **, pfn_t *);
  19         /* copy_from_iter: required operation for fs-dax direct-i/o */
  20         size_t (*copy_from_iter)(struct dax_device *, pgoff_t, void *, size_t,
  21                         struct iov_iter *);
  22         /* flush: optional driver-specific cache management after writes */
  23         void (*flush)(struct dax_device *, pgoff_t, void *, size_t);
  24 };
  25
  26 extern struct attribute_group dax_attribute_group;
  27
  28 #if IS_ENABLED(CONFIG_DAX)
  29 struct dax_device *dax_get_by_host(const char *host);
  30 void put_dax(struct dax_device *dax_dev);
  31 #else
  32 static inline struct dax_device *dax_get_by_host(const char *host)
  33 {
  34         return NULL;
  35 }
  36
  37 static inline void put_dax(struct dax_device *dax_dev)
  38 {
  39 }
  40 #endif
  41
  42 int bdev_dax_pgoff(struct block_device *, sector_t, size_t, pgoff_t *pgoff);
  43 #if IS_ENABLED(CONFIG_FS_DAX)
  44 int __bdev_dax_supported(struct super_block *sb, int blocksize);
  45 static inline int bdev_dax_supported(struct super_block *sb, int blocksize)
  46 {
  47         return __bdev_dax_supported(sb, blocksize);
  48 }
  49
  50 static inline struct dax_device *fs_dax_get_by_host(const char *host)
  51 {
  52         return dax_get_by_host(host);
  53 }
  54
  55 static inline void fs_put_dax(struct dax_device *dax_dev)
  56 {
  57         put_dax(dax_dev);
  58 }
  59
  60 #else
  61 static inline int bdev_dax_supported(struct super_block *sb, int blocksize)
  62 {
  63         return -EOPNOTSUPP;
  64 }
  65
  66 static inline struct dax_device *fs_dax_get_by_host(const char *host)
  67 {
  68         return NULL;
  69 }
  70
  71 static inline void fs_put_dax(struct dax_device *dax_dev)
  72 {
  73 }
  74 #endif
  75
  76 int dax_read_lock(void);
  77 void dax_read_unlock(int id);
  78 struct dax_device *alloc_dax(void *private, const char *host,
  79                 const struct dax_operations *ops);
  80 bool dax_alive(struct dax_device *dax_dev);
  81 void kill_dax(struct dax_device *dax_dev);
  82 void *dax_get_private(struct dax_device *dax_dev);
  83 long dax_direct_access(struct dax_device *dax_dev, pgoff_t pgoff, long nr_pages,
  84                 void **kaddr, pfn_t *pfn);
  85 size_t dax_copy_from_iter(struct dax_device *dax_dev, pgoff_t pgoff, void *addr,
  86                 size_t bytes, struct iov_iter *i);
  87 void dax_flush(struct dax_device *dax_dev, pgoff_t pgoff, void *addr,
  88                 size_t size);
  89 void dax_write_cache(struct dax_device *dax_dev, bool wc);
  90
  91 /*
  92  * We use lowest available bit in exceptional entry for locking, one bit for
  93  * the entry size (PMD) and two more to tell us if the entry is a huge zero
  94  * page (HZP) or an empty entry that is just used for locking.  In total four
  95  * special bits.
  96  *
  97  * If the PMD bit isn't set the entry has size PAGE_SIZE, and if the HZP and
  98  * EMPTY bits aren't set the entry is a normal DAX entry with a filesystem
  99  * block allocation.
 100  */
 101 #define RADIX_DAX_SHIFT (RADIX_TREE_EXCEPTIONAL_SHIFT + 4)
 102 #define RADIX_DAX_ENTRY_LOCK (1 << RADIX_TREE_EXCEPTIONAL_SHIFT)
 103 #define RADIX_DAX_PMD (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 1))
 104 #define RADIX_DAX_HZP (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 2))
 105 #define RADIX_DAX_EMPTY (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 3))
 106
 107 static inline unsigned long dax_radix_sector(void *entry)
 108 {
 109         return (unsigned long)entry >> RADIX_DAX_SHIFT;
 110 }
 111
 112 static inline void *dax_radix_locked_entry(sector_t sector, unsigned long flags)
 113 {
 114         return (void *)(RADIX_TREE_EXCEPTIONAL_ENTRY | flags |
 115                         ((unsigned long)sector << RADIX_DAX_SHIFT) |
 116                         RADIX_DAX_ENTRY_LOCK);
 117 }
 118
 119 ssize_t dax_iomap_rw(struct kiocb *iocb, struct iov_iter *iter,
 120                 const struct iomap_ops *ops);
 121 int dax_iomap_fault(struct vm_fault *vmf, enum page_entry_size pe_size,
 122                     const struct iomap_ops *ops);
 123 int dax_delete_mapping_entry(struct address_space *mapping, pgoff_t index);
 124 int dax_invalidate_mapping_entry_sync(struct address_space *mapping,
 125                                       pgoff_t index);
 126 void dax_wake_mapping_entry_waiter(struct address_space *mapping,
 127                 pgoff_t index, void *entry, bool wake_all);
 128
 129 #ifdef CONFIG_FS_DAX
 130 int __dax_zero_page_range(struct block_device *bdev,
 131                 struct dax_device *dax_dev, sector_t sector,
 132                 unsigned int offset, unsigned int length);
 133 #else
 134 static inline int __dax_zero_page_range(struct block_device *bdev,
 135                 struct dax_device *dax_dev, sector_t sector,
 136                 unsigned int offset, unsigned int length)
 137 {
 138         return -ENXIO;
 139 }
 140 #endif
 141
 142 #ifdef CONFIG_FS_DAX_PMD
 143 static inline unsigned int dax_radix_order(void *entry)
 144 {
 145         if ((unsigned long)entry & RADIX_DAX_PMD)
 146                 return PMD_SHIFT - PAGE_SHIFT;
 147         return 0;
 148 }
 149 #else
 150 static inline unsigned int dax_radix_order(void *entry)
 151 {
 152         return 0;
 153 }
 154 #endif
 155 int dax_pfn_mkwrite(struct vm_fault *vmf);
 156
 157 static inline bool dax_mapping(struct address_space *mapping)
 158 {
 159         return mapping->host && IS_DAX(mapping->host);
 160 }
 161
 162 struct writeback_control;
 163 int dax_writeback_mapping_range(struct address_space *mapping,
 164                 struct block_device *bdev, struct writeback_control *wbc);
 165 #endif