xref: /linux-6.15/include/linux/dax.h (revision efe4a1ac)
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 };
20 
21 #if IS_ENABLED(CONFIG_DAX)
22 struct dax_device *dax_get_by_host(const char *host);
23 void put_dax(struct dax_device *dax_dev);
24 #else
25 static inline struct dax_device *dax_get_by_host(const char *host)
26 {
27 	return NULL;
28 }
29 
30 static inline void put_dax(struct dax_device *dax_dev)
31 {
32 }
33 #endif
34 
35 int bdev_dax_pgoff(struct block_device *, sector_t, size_t, pgoff_t *pgoff);
36 #if IS_ENABLED(CONFIG_FS_DAX)
37 int __bdev_dax_supported(struct super_block *sb, int blocksize);
38 static inline int bdev_dax_supported(struct super_block *sb, int blocksize)
39 {
40 	return __bdev_dax_supported(sb, blocksize);
41 }
42 
43 static inline struct dax_device *fs_dax_get_by_host(const char *host)
44 {
45 	return dax_get_by_host(host);
46 }
47 
48 static inline void fs_put_dax(struct dax_device *dax_dev)
49 {
50 	put_dax(dax_dev);
51 }
52 
53 #else
54 static inline int bdev_dax_supported(struct super_block *sb, int blocksize)
55 {
56 	return -EOPNOTSUPP;
57 }
58 
59 static inline struct dax_device *fs_dax_get_by_host(const char *host)
60 {
61 	return NULL;
62 }
63 
64 static inline void fs_put_dax(struct dax_device *dax_dev)
65 {
66 }
67 #endif
68 
69 int dax_read_lock(void);
70 void dax_read_unlock(int id);
71 struct dax_device *alloc_dax(void *private, const char *host,
72 		const struct dax_operations *ops);
73 bool dax_alive(struct dax_device *dax_dev);
74 void kill_dax(struct dax_device *dax_dev);
75 void *dax_get_private(struct dax_device *dax_dev);
76 long dax_direct_access(struct dax_device *dax_dev, pgoff_t pgoff, long nr_pages,
77 		void **kaddr, pfn_t *pfn);
78 
79 /*
80  * We use lowest available bit in exceptional entry for locking, one bit for
81  * the entry size (PMD) and two more to tell us if the entry is a huge zero
82  * page (HZP) or an empty entry that is just used for locking.  In total four
83  * special bits.
84  *
85  * If the PMD bit isn't set the entry has size PAGE_SIZE, and if the HZP and
86  * EMPTY bits aren't set the entry is a normal DAX entry with a filesystem
87  * block allocation.
88  */
89 #define RADIX_DAX_SHIFT	(RADIX_TREE_EXCEPTIONAL_SHIFT + 4)
90 #define RADIX_DAX_ENTRY_LOCK (1 << RADIX_TREE_EXCEPTIONAL_SHIFT)
91 #define RADIX_DAX_PMD (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 1))
92 #define RADIX_DAX_HZP (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 2))
93 #define RADIX_DAX_EMPTY (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 3))
94 
95 static inline unsigned long dax_radix_sector(void *entry)
96 {
97 	return (unsigned long)entry >> RADIX_DAX_SHIFT;
98 }
99 
100 static inline void *dax_radix_locked_entry(sector_t sector, unsigned long flags)
101 {
102 	return (void *)(RADIX_TREE_EXCEPTIONAL_ENTRY | flags |
103 			((unsigned long)sector << RADIX_DAX_SHIFT) |
104 			RADIX_DAX_ENTRY_LOCK);
105 }
106 
107 ssize_t dax_iomap_rw(struct kiocb *iocb, struct iov_iter *iter,
108 		const struct iomap_ops *ops);
109 int dax_iomap_fault(struct vm_fault *vmf, enum page_entry_size pe_size,
110 		    const struct iomap_ops *ops);
111 int dax_delete_mapping_entry(struct address_space *mapping, pgoff_t index);
112 int dax_invalidate_mapping_entry_sync(struct address_space *mapping,
113 				      pgoff_t index);
114 void dax_wake_mapping_entry_waiter(struct address_space *mapping,
115 		pgoff_t index, void *entry, bool wake_all);
116 
117 #ifdef CONFIG_FS_DAX
118 int __dax_zero_page_range(struct block_device *bdev,
119 		struct dax_device *dax_dev, sector_t sector,
120 		unsigned int offset, unsigned int length);
121 #else
122 static inline int __dax_zero_page_range(struct block_device *bdev,
123 		struct dax_device *dax_dev, sector_t sector,
124 		unsigned int offset, unsigned int length)
125 {
126 	return -ENXIO;
127 }
128 #endif
129 
130 #ifdef CONFIG_FS_DAX_PMD
131 static inline unsigned int dax_radix_order(void *entry)
132 {
133 	if ((unsigned long)entry & RADIX_DAX_PMD)
134 		return PMD_SHIFT - PAGE_SHIFT;
135 	return 0;
136 }
137 #else
138 static inline unsigned int dax_radix_order(void *entry)
139 {
140 	return 0;
141 }
142 #endif
143 int dax_pfn_mkwrite(struct vm_fault *vmf);
144 
145 static inline bool vma_is_dax(struct vm_area_struct *vma)
146 {
147 	return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
148 }
149 
150 static inline bool dax_mapping(struct address_space *mapping)
151 {
152 	return mapping->host && IS_DAX(mapping->host);
153 }
154 
155 struct writeback_control;
156 int dax_writeback_mapping_range(struct address_space *mapping,
157 		struct block_device *bdev, struct writeback_control *wbc);
158 #endif
159