xref: /linux-6.15/include/linux/dax.h (revision 90bb7664)
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 int bdev_dax_pgoff(struct block_device *, sector_t, size_t, pgoff_t *pgoff);
22 #if IS_ENABLED(CONFIG_FS_DAX)
23 int __bdev_dax_supported(struct super_block *sb, int blocksize);
24 static inline int bdev_dax_supported(struct super_block *sb, int blocksize)
25 {
26 	return __bdev_dax_supported(sb, blocksize);
27 }
28 #else
29 static inline int bdev_dax_supported(struct super_block *sb, int blocksize)
30 {
31 	return -EOPNOTSUPP;
32 }
33 #endif
34 
35 #if IS_ENABLED(CONFIG_DAX)
36 struct dax_device *dax_get_by_host(const char *host);
37 void put_dax(struct dax_device *dax_dev);
38 #else
39 static inline struct dax_device *dax_get_by_host(const char *host)
40 {
41 	return NULL;
42 }
43 
44 static inline void put_dax(struct dax_device *dax_dev)
45 {
46 }
47 #endif
48 
49 int dax_read_lock(void);
50 void dax_read_unlock(int id);
51 struct dax_device *alloc_dax(void *private, const char *host,
52 		const struct dax_operations *ops);
53 bool dax_alive(struct dax_device *dax_dev);
54 void kill_dax(struct dax_device *dax_dev);
55 void *dax_get_private(struct dax_device *dax_dev);
56 long dax_direct_access(struct dax_device *dax_dev, pgoff_t pgoff, long nr_pages,
57 		void **kaddr, pfn_t *pfn);
58 
59 /*
60  * We use lowest available bit in exceptional entry for locking, one bit for
61  * the entry size (PMD) and two more to tell us if the entry is a huge zero
62  * page (HZP) or an empty entry that is just used for locking.  In total four
63  * special bits.
64  *
65  * If the PMD bit isn't set the entry has size PAGE_SIZE, and if the HZP and
66  * EMPTY bits aren't set the entry is a normal DAX entry with a filesystem
67  * block allocation.
68  */
69 #define RADIX_DAX_SHIFT	(RADIX_TREE_EXCEPTIONAL_SHIFT + 4)
70 #define RADIX_DAX_ENTRY_LOCK (1 << RADIX_TREE_EXCEPTIONAL_SHIFT)
71 #define RADIX_DAX_PMD (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 1))
72 #define RADIX_DAX_HZP (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 2))
73 #define RADIX_DAX_EMPTY (1 << (RADIX_TREE_EXCEPTIONAL_SHIFT + 3))
74 
75 static inline unsigned long dax_radix_sector(void *entry)
76 {
77 	return (unsigned long)entry >> RADIX_DAX_SHIFT;
78 }
79 
80 static inline void *dax_radix_locked_entry(sector_t sector, unsigned long flags)
81 {
82 	return (void *)(RADIX_TREE_EXCEPTIONAL_ENTRY | flags |
83 			((unsigned long)sector << RADIX_DAX_SHIFT) |
84 			RADIX_DAX_ENTRY_LOCK);
85 }
86 
87 ssize_t dax_iomap_rw(struct kiocb *iocb, struct iov_iter *iter,
88 		const struct iomap_ops *ops);
89 int dax_iomap_fault(struct vm_fault *vmf, enum page_entry_size pe_size,
90 		    const struct iomap_ops *ops);
91 int dax_delete_mapping_entry(struct address_space *mapping, pgoff_t index);
92 int dax_invalidate_mapping_entry_sync(struct address_space *mapping,
93 				      pgoff_t index);
94 void dax_wake_mapping_entry_waiter(struct address_space *mapping,
95 		pgoff_t index, void *entry, bool wake_all);
96 
97 #ifdef CONFIG_FS_DAX
98 int __dax_zero_page_range(struct block_device *bdev,
99 		struct dax_device *dax_dev, sector_t sector,
100 		unsigned int offset, unsigned int length);
101 #else
102 static inline int __dax_zero_page_range(struct block_device *bdev,
103 		struct dax_device *dax_dev, sector_t sector,
104 		unsigned int offset, unsigned int length)
105 {
106 	return -ENXIO;
107 }
108 #endif
109 
110 #ifdef CONFIG_FS_DAX_PMD
111 static inline unsigned int dax_radix_order(void *entry)
112 {
113 	if ((unsigned long)entry & RADIX_DAX_PMD)
114 		return PMD_SHIFT - PAGE_SHIFT;
115 	return 0;
116 }
117 #else
118 static inline unsigned int dax_radix_order(void *entry)
119 {
120 	return 0;
121 }
122 #endif
123 int dax_pfn_mkwrite(struct vm_fault *vmf);
124 
125 static inline bool vma_is_dax(struct vm_area_struct *vma)
126 {
127 	return vma->vm_file && IS_DAX(vma->vm_file->f_mapping->host);
128 }
129 
130 static inline bool dax_mapping(struct address_space *mapping)
131 {
132 	return mapping->host && IS_DAX(mapping->host);
133 }
134 
135 struct writeback_control;
136 int dax_writeback_mapping_range(struct address_space *mapping,
137 		struct block_device *bdev, struct writeback_control *wbc);
138 #endif
139