2 * Copyright(c) 2017 Intel Corporation. All rights reserved.
4 * This program is free software; you can redistribute it and/or modify
5 * it under the terms of version 2 of the GNU General Public License as
6 * published by the Free Software Foundation.
8 * This program is distributed in the hope that it will be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
11 * General Public License for more details.
13 #include <linux/pagemap.h>
14 #include <linux/module.h>
15 #include <linux/mount.h>
16 #include <linux/magic.h>
17 #include <linux/genhd.h>
18 #include <linux/cdev.h>
19 #include <linux/hash.h>
20 #include <linux/slab.h>
21 #include <linux/dax.h>
24 static dev_t dax_devt
;
25 DEFINE_STATIC_SRCU(dax_srcu
);
26 static struct vfsmount
*dax_mnt
;
27 static DEFINE_IDA(dax_minor_ida
);
28 static struct kmem_cache
*dax_cache __read_mostly
;
29 static struct super_block
*dax_superblock __read_mostly
;
31 #define DAX_HASH_SIZE (PAGE_SIZE / sizeof(struct hlist_head))
32 static struct hlist_head dax_host_list
[DAX_HASH_SIZE
];
33 static DEFINE_SPINLOCK(dax_host_lock
);
35 int dax_read_lock(void)
37 return srcu_read_lock(&dax_srcu
);
39 EXPORT_SYMBOL_GPL(dax_read_lock
);
41 void dax_read_unlock(int id
)
43 srcu_read_unlock(&dax_srcu
, id
);
45 EXPORT_SYMBOL_GPL(dax_read_unlock
);
48 int bdev_dax_pgoff(struct block_device
*bdev
, sector_t sector
, size_t size
,
51 phys_addr_t phys_off
= (get_start_sect(bdev
) + sector
) * 512;
54 *pgoff
= PHYS_PFN(phys_off
);
55 if (phys_off
% PAGE_SIZE
|| size
% PAGE_SIZE
)
59 EXPORT_SYMBOL(bdev_dax_pgoff
);
62 * __bdev_dax_supported() - Check if the device supports dax for filesystem
63 * @sb: The superblock of the device
64 * @blocksize: The block size of the device
66 * This is a library function for filesystems to check if the block device
67 * can be mounted with dax option.
69 * Return: negative errno if unsupported, 0 if supported.
71 int __bdev_dax_supported(struct super_block
*sb
, int blocksize
)
73 struct block_device
*bdev
= sb
->s_bdev
;
74 struct dax_device
*dax_dev
;
81 if (blocksize
!= PAGE_SIZE
) {
82 pr_err("VFS (%s): error: unsupported blocksize for dax\n",
87 err
= bdev_dax_pgoff(bdev
, 0, PAGE_SIZE
, &pgoff
);
89 pr_err("VFS (%s): error: unaligned partition for dax\n",
94 dax_dev
= dax_get_by_host(bdev
->bd_disk
->disk_name
);
96 pr_err("VFS (%s): error: device does not support dax\n",
101 id
= dax_read_lock();
102 len
= dax_direct_access(dax_dev
, pgoff
, 1, &kaddr
, &pfn
);
108 pr_err("VFS (%s): error: dax access failed (%ld)",
110 return len
< 0 ? len
: -EIO
;
115 EXPORT_SYMBOL_GPL(__bdev_dax_supported
);
119 * struct dax_device - anchor object for dax services
121 * @cdev: optional character interface for "device dax"
122 * @host: optional name for lookups where the device path is not available
123 * @private: dax driver private data
124 * @alive: !alive + rcu grace period == no new operations / mappings
127 struct hlist_node list
;
133 const struct dax_operations
*ops
;
137 * dax_direct_access() - translate a device pgoff to an absolute pfn
138 * @dax_dev: a dax_device instance representing the logical memory range
139 * @pgoff: offset in pages from the start of the device to translate
140 * @nr_pages: number of consecutive pages caller can handle relative to @pfn
141 * @kaddr: output parameter that returns a virtual address mapping of pfn
142 * @pfn: output parameter that returns an absolute pfn translation of @pgoff
144 * Return: negative errno if an error occurs, otherwise the number of
145 * pages accessible at the device relative @pgoff.
147 long dax_direct_access(struct dax_device
*dax_dev
, pgoff_t pgoff
, long nr_pages
,
148 void **kaddr
, pfn_t
*pfn
)
153 * The device driver is allowed to sleep, in order to make the
154 * memory directly accessible.
161 if (!dax_alive(dax_dev
))
167 avail
= dax_dev
->ops
->direct_access(dax_dev
, pgoff
, nr_pages
,
171 return min(avail
, nr_pages
);
173 EXPORT_SYMBOL_GPL(dax_direct_access
);
175 bool dax_alive(struct dax_device
*dax_dev
)
177 lockdep_assert_held(&dax_srcu
);
178 return dax_dev
->alive
;
180 EXPORT_SYMBOL_GPL(dax_alive
);
182 static int dax_host_hash(const char *host
)
184 return hashlen_hash(hashlen_string("DAX", host
)) % DAX_HASH_SIZE
;
188 * Note, rcu is not protecting the liveness of dax_dev, rcu is ensuring
189 * that any fault handlers or operations that might have seen
190 * dax_alive(), have completed. Any operations that start after
191 * synchronize_srcu() has run will abort upon seeing !dax_alive().
193 void kill_dax(struct dax_device
*dax_dev
)
198 dax_dev
->alive
= false;
200 synchronize_srcu(&dax_srcu
);
202 spin_lock(&dax_host_lock
);
203 hlist_del_init(&dax_dev
->list
);
204 spin_unlock(&dax_host_lock
);
206 dax_dev
->private = NULL
;
208 EXPORT_SYMBOL_GPL(kill_dax
);
210 static struct inode
*dax_alloc_inode(struct super_block
*sb
)
212 struct dax_device
*dax_dev
;
214 dax_dev
= kmem_cache_alloc(dax_cache
, GFP_KERNEL
);
215 return &dax_dev
->inode
;
218 static struct dax_device
*to_dax_dev(struct inode
*inode
)
220 return container_of(inode
, struct dax_device
, inode
);
223 static void dax_i_callback(struct rcu_head
*head
)
225 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
226 struct dax_device
*dax_dev
= to_dax_dev(inode
);
228 kfree(dax_dev
->host
);
229 dax_dev
->host
= NULL
;
230 ida_simple_remove(&dax_minor_ida
, MINOR(inode
->i_rdev
));
231 kmem_cache_free(dax_cache
, dax_dev
);
234 static void dax_destroy_inode(struct inode
*inode
)
236 struct dax_device
*dax_dev
= to_dax_dev(inode
);
238 WARN_ONCE(dax_dev
->alive
,
239 "kill_dax() must be called before final iput()\n");
240 call_rcu(&inode
->i_rcu
, dax_i_callback
);
243 static const struct super_operations dax_sops
= {
244 .statfs
= simple_statfs
,
245 .alloc_inode
= dax_alloc_inode
,
246 .destroy_inode
= dax_destroy_inode
,
247 .drop_inode
= generic_delete_inode
,
250 static struct dentry
*dax_mount(struct file_system_type
*fs_type
,
251 int flags
, const char *dev_name
, void *data
)
253 return mount_pseudo(fs_type
, "dax:", &dax_sops
, NULL
, DAXFS_MAGIC
);
256 static struct file_system_type dax_fs_type
= {
259 .kill_sb
= kill_anon_super
,
262 static int dax_test(struct inode
*inode
, void *data
)
264 dev_t devt
= *(dev_t
*) data
;
266 return inode
->i_rdev
== devt
;
269 static int dax_set(struct inode
*inode
, void *data
)
271 dev_t devt
= *(dev_t
*) data
;
273 inode
->i_rdev
= devt
;
277 static struct dax_device
*dax_dev_get(dev_t devt
)
279 struct dax_device
*dax_dev
;
282 inode
= iget5_locked(dax_superblock
, hash_32(devt
+ DAXFS_MAGIC
, 31),
283 dax_test
, dax_set
, &devt
);
288 dax_dev
= to_dax_dev(inode
);
289 if (inode
->i_state
& I_NEW
) {
290 dax_dev
->alive
= true;
291 inode
->i_cdev
= &dax_dev
->cdev
;
292 inode
->i_mode
= S_IFCHR
;
293 inode
->i_flags
= S_DAX
;
294 mapping_set_gfp_mask(&inode
->i_data
, GFP_USER
);
295 unlock_new_inode(inode
);
301 static void dax_add_host(struct dax_device
*dax_dev
, const char *host
)
306 * Unconditionally init dax_dev since it's coming from a
307 * non-zeroed slab cache
309 INIT_HLIST_NODE(&dax_dev
->list
);
310 dax_dev
->host
= host
;
314 hash
= dax_host_hash(host
);
315 spin_lock(&dax_host_lock
);
316 hlist_add_head(&dax_dev
->list
, &dax_host_list
[hash
]);
317 spin_unlock(&dax_host_lock
);
320 struct dax_device
*alloc_dax(void *private, const char *__host
,
321 const struct dax_operations
*ops
)
323 struct dax_device
*dax_dev
;
328 host
= kstrdup(__host
, GFP_KERNEL
);
332 minor
= ida_simple_get(&dax_minor_ida
, 0, MINORMASK
+1, GFP_KERNEL
);
336 devt
= MKDEV(MAJOR(dax_devt
), minor
);
337 dax_dev
= dax_dev_get(devt
);
341 dax_add_host(dax_dev
, host
);
343 dax_dev
->private = private;
347 ida_simple_remove(&dax_minor_ida
, minor
);
352 EXPORT_SYMBOL_GPL(alloc_dax
);
354 void put_dax(struct dax_device
*dax_dev
)
358 iput(&dax_dev
->inode
);
360 EXPORT_SYMBOL_GPL(put_dax
);
363 * dax_get_by_host() - temporary lookup mechanism for filesystem-dax
364 * @host: alternate name for the device registered by a dax driver
366 struct dax_device
*dax_get_by_host(const char *host
)
368 struct dax_device
*dax_dev
, *found
= NULL
;
374 hash
= dax_host_hash(host
);
376 id
= dax_read_lock();
377 spin_lock(&dax_host_lock
);
378 hlist_for_each_entry(dax_dev
, &dax_host_list
[hash
], list
) {
379 if (!dax_alive(dax_dev
)
380 || strcmp(host
, dax_dev
->host
) != 0)
383 if (igrab(&dax_dev
->inode
))
387 spin_unlock(&dax_host_lock
);
392 EXPORT_SYMBOL_GPL(dax_get_by_host
);
395 * inode_dax: convert a public inode into its dax_dev
396 * @inode: An inode with i_cdev pointing to a dax_dev
398 * Note this is not equivalent to to_dax_dev() which is for private
399 * internal use where we know the inode filesystem type == dax_fs_type.
401 struct dax_device
*inode_dax(struct inode
*inode
)
403 struct cdev
*cdev
= inode
->i_cdev
;
405 return container_of(cdev
, struct dax_device
, cdev
);
407 EXPORT_SYMBOL_GPL(inode_dax
);
409 struct inode
*dax_inode(struct dax_device
*dax_dev
)
411 return &dax_dev
->inode
;
413 EXPORT_SYMBOL_GPL(dax_inode
);
415 void *dax_get_private(struct dax_device
*dax_dev
)
417 return dax_dev
->private;
419 EXPORT_SYMBOL_GPL(dax_get_private
);
421 static void init_once(void *_dax_dev
)
423 struct dax_device
*dax_dev
= _dax_dev
;
424 struct inode
*inode
= &dax_dev
->inode
;
426 inode_init_once(inode
);
429 static int __dax_fs_init(void)
433 dax_cache
= kmem_cache_create("dax_cache", sizeof(struct dax_device
), 0,
434 (SLAB_HWCACHE_ALIGN
|SLAB_RECLAIM_ACCOUNT
|
435 SLAB_MEM_SPREAD
|SLAB_ACCOUNT
),
440 rc
= register_filesystem(&dax_fs_type
);
442 goto err_register_fs
;
444 dax_mnt
= kern_mount(&dax_fs_type
);
445 if (IS_ERR(dax_mnt
)) {
446 rc
= PTR_ERR(dax_mnt
);
449 dax_superblock
= dax_mnt
->mnt_sb
;
454 unregister_filesystem(&dax_fs_type
);
456 kmem_cache_destroy(dax_cache
);
461 static void __dax_fs_exit(void)
463 kern_unmount(dax_mnt
);
464 unregister_filesystem(&dax_fs_type
);
465 kmem_cache_destroy(dax_cache
);
468 static int __init
dax_fs_init(void)
472 rc
= __dax_fs_init();
476 rc
= alloc_chrdev_region(&dax_devt
, 0, MINORMASK
+1, "dax");
482 static void __exit
dax_fs_exit(void)
484 unregister_chrdev_region(dax_devt
, MINORMASK
+1);
485 ida_destroy(&dax_minor_ida
);
489 MODULE_AUTHOR("Intel Corporation");
490 MODULE_LICENSE("GPL v2");
491 subsys_initcall(dax_fs_init
);
492 module_exit(dax_fs_exit
);