]>
Commit | Line | Data |
---|---|---|
155e7524 FI |
1 | /* |
2 | * linux/fs/nfs/blocklayout/blocklayout.c | |
3 | * | |
4 | * Module for the NFSv4.1 pNFS block layout driver. | |
5 | * | |
6 | * Copyright (c) 2006 The Regents of the University of Michigan. | |
7 | * All rights reserved. | |
8 | * | |
9 | * Andy Adamson <andros@citi.umich.edu> | |
10 | * Fred Isaman <iisaman@umich.edu> | |
11 | * | |
12 | * permission is granted to use, copy, create derivative works and | |
13 | * redistribute this software and such derivative works for any purpose, | |
14 | * so long as the name of the university of michigan is not used in | |
15 | * any advertising or publicity pertaining to the use or distribution | |
16 | * of this software without specific, written prior authorization. if | |
17 | * the above copyright notice or any other identification of the | |
18 | * university of michigan is included in any copy of any portion of | |
19 | * this software, then the disclaimer below must also be included. | |
20 | * | |
21 | * this software is provided as is, without representation from the | |
22 | * university of michigan as to its fitness for any purpose, and without | |
23 | * warranty by the university of michigan of any kind, either express | |
24 | * or implied, including without limitation the implied warranties of | |
25 | * merchantability and fitness for a particular purpose. the regents | |
26 | * of the university of michigan shall not be liable for any damages, | |
27 | * including special, indirect, incidental, or consequential damages, | |
28 | * with respect to any claim arising out or in connection with the use | |
29 | * of the software, even if it has been or is hereafter advised of the | |
30 | * possibility of such damages. | |
31 | */ | |
32 | #include <linux/module.h> | |
33 | #include <linux/init.h> | |
fe0a9b74 JR |
34 | #include <linux/mount.h> |
35 | #include <linux/namei.h> | |
155e7524 FI |
36 | |
37 | #include "blocklayout.h" | |
38 | ||
39 | #define NFSDBG_FACILITY NFSDBG_PNFS_LD | |
40 | ||
41 | MODULE_LICENSE("GPL"); | |
42 | MODULE_AUTHOR("Andy Adamson <andros@citi.umich.edu>"); | |
43 | MODULE_DESCRIPTION("The NFSv4.1 pNFS Block layout driver"); | |
44 | ||
fe0a9b74 JR |
45 | struct dentry *bl_device_pipe; |
46 | wait_queue_head_t bl_wq; | |
47 | ||
155e7524 FI |
48 | static enum pnfs_try_status |
49 | bl_read_pagelist(struct nfs_read_data *rdata) | |
50 | { | |
51 | return PNFS_NOT_ATTEMPTED; | |
52 | } | |
53 | ||
54 | static enum pnfs_try_status | |
55 | bl_write_pagelist(struct nfs_write_data *wdata, | |
56 | int sync) | |
57 | { | |
58 | return PNFS_NOT_ATTEMPTED; | |
59 | } | |
60 | ||
9e692969 | 61 | /* FIXME - range ignored */ |
155e7524 | 62 | static void |
9e692969 | 63 | release_extents(struct pnfs_block_layout *bl, struct pnfs_layout_range *range) |
155e7524 | 64 | { |
9e692969 FI |
65 | int i; |
66 | struct pnfs_block_extent *be; | |
67 | ||
68 | spin_lock(&bl->bl_ext_lock); | |
69 | for (i = 0; i < EXTENT_LISTS; i++) { | |
70 | while (!list_empty(&bl->bl_extents[i])) { | |
71 | be = list_first_entry(&bl->bl_extents[i], | |
72 | struct pnfs_block_extent, | |
73 | be_node); | |
74 | list_del(&be->be_node); | |
75 | bl_put_extent(be); | |
76 | } | |
77 | } | |
78 | spin_unlock(&bl->bl_ext_lock); | |
155e7524 FI |
79 | } |
80 | ||
155e7524 FI |
81 | static void |
82 | release_inval_marks(struct pnfs_inval_markings *marks) | |
83 | { | |
c1c2a4cd FI |
84 | struct pnfs_inval_tracking *pos, *temp; |
85 | ||
86 | list_for_each_entry_safe(pos, temp, &marks->im_tree.mtt_stub, it_link) { | |
87 | list_del(&pos->it_link); | |
88 | kfree(pos); | |
89 | } | |
155e7524 FI |
90 | return; |
91 | } | |
92 | ||
93 | static void bl_free_layout_hdr(struct pnfs_layout_hdr *lo) | |
94 | { | |
95 | struct pnfs_block_layout *bl = BLK_LO2EXT(lo); | |
96 | ||
97 | dprintk("%s enter\n", __func__); | |
98 | release_extents(bl, NULL); | |
99 | release_inval_marks(&bl->bl_inval); | |
100 | kfree(bl); | |
101 | } | |
102 | ||
103 | static struct pnfs_layout_hdr *bl_alloc_layout_hdr(struct inode *inode, | |
104 | gfp_t gfp_flags) | |
105 | { | |
106 | struct pnfs_block_layout *bl; | |
107 | ||
108 | dprintk("%s enter\n", __func__); | |
109 | bl = kzalloc(sizeof(*bl), gfp_flags); | |
110 | if (!bl) | |
111 | return NULL; | |
112 | spin_lock_init(&bl->bl_ext_lock); | |
113 | INIT_LIST_HEAD(&bl->bl_extents[0]); | |
114 | INIT_LIST_HEAD(&bl->bl_extents[1]); | |
115 | INIT_LIST_HEAD(&bl->bl_commit); | |
116 | INIT_LIST_HEAD(&bl->bl_committing); | |
117 | bl->bl_count = 0; | |
118 | bl->bl_blocksize = NFS_SERVER(inode)->pnfs_blksize >> SECTOR_SHIFT; | |
119 | BL_INIT_INVAL_MARKS(&bl->bl_inval, bl->bl_blocksize); | |
120 | return &bl->bl_layout; | |
121 | } | |
122 | ||
a60d2ebd | 123 | static void bl_free_lseg(struct pnfs_layout_segment *lseg) |
155e7524 | 124 | { |
a60d2ebd FI |
125 | dprintk("%s enter\n", __func__); |
126 | kfree(lseg); | |
155e7524 FI |
127 | } |
128 | ||
a60d2ebd FI |
129 | /* We pretty much ignore lseg, and store all data layout wide, so we |
130 | * can correctly merge. | |
131 | */ | |
132 | static struct pnfs_layout_segment *bl_alloc_lseg(struct pnfs_layout_hdr *lo, | |
133 | struct nfs4_layoutget_res *lgr, | |
134 | gfp_t gfp_flags) | |
155e7524 | 135 | { |
a60d2ebd FI |
136 | struct pnfs_layout_segment *lseg; |
137 | int status; | |
138 | ||
139 | dprintk("%s enter\n", __func__); | |
140 | lseg = kzalloc(sizeof(*lseg), gfp_flags); | |
141 | if (!lseg) | |
142 | return ERR_PTR(-ENOMEM); | |
143 | status = nfs4_blk_process_layoutget(lo, lgr, gfp_flags); | |
144 | if (status) { | |
145 | /* We don't want to call the full-blown bl_free_lseg, | |
146 | * since on error extents were not touched. | |
147 | */ | |
148 | kfree(lseg); | |
149 | return ERR_PTR(status); | |
150 | } | |
151 | return lseg; | |
155e7524 FI |
152 | } |
153 | ||
154 | static void | |
155 | bl_encode_layoutcommit(struct pnfs_layout_hdr *lo, struct xdr_stream *xdr, | |
156 | const struct nfs4_layoutcommit_args *arg) | |
157 | { | |
158 | } | |
159 | ||
160 | static void | |
161 | bl_cleanup_layoutcommit(struct nfs4_layoutcommit_data *lcdata) | |
162 | { | |
163 | } | |
164 | ||
2f9fd182 FI |
165 | static void free_blk_mountid(struct block_mount_id *mid) |
166 | { | |
167 | if (mid) { | |
168 | struct pnfs_block_dev *dev; | |
169 | spin_lock(&mid->bm_lock); | |
170 | while (!list_empty(&mid->bm_devlist)) { | |
171 | dev = list_first_entry(&mid->bm_devlist, | |
172 | struct pnfs_block_dev, | |
173 | bm_node); | |
174 | list_del(&dev->bm_node); | |
175 | bl_free_block_dev(dev); | |
176 | } | |
177 | spin_unlock(&mid->bm_lock); | |
178 | kfree(mid); | |
179 | } | |
180 | } | |
181 | ||
182 | /* This is mostly copied from the filelayout's get_device_info function. | |
183 | * It seems much of this should be at the generic pnfs level. | |
184 | */ | |
185 | static struct pnfs_block_dev * | |
186 | nfs4_blk_get_deviceinfo(struct nfs_server *server, const struct nfs_fh *fh, | |
187 | struct nfs4_deviceid *d_id) | |
188 | { | |
189 | struct pnfs_device *dev; | |
190 | struct pnfs_block_dev *rv = NULL; | |
191 | u32 max_resp_sz; | |
192 | int max_pages; | |
193 | struct page **pages = NULL; | |
194 | int i, rc; | |
195 | ||
196 | /* | |
197 | * Use the session max response size as the basis for setting | |
198 | * GETDEVICEINFO's maxcount | |
199 | */ | |
200 | max_resp_sz = server->nfs_client->cl_session->fc_attrs.max_resp_sz; | |
201 | max_pages = max_resp_sz >> PAGE_SHIFT; | |
202 | dprintk("%s max_resp_sz %u max_pages %d\n", | |
203 | __func__, max_resp_sz, max_pages); | |
204 | ||
205 | dev = kmalloc(sizeof(*dev), GFP_NOFS); | |
206 | if (!dev) { | |
207 | dprintk("%s kmalloc failed\n", __func__); | |
208 | return NULL; | |
209 | } | |
210 | ||
211 | pages = kzalloc(max_pages * sizeof(struct page *), GFP_NOFS); | |
212 | if (pages == NULL) { | |
213 | kfree(dev); | |
214 | return NULL; | |
215 | } | |
216 | for (i = 0; i < max_pages; i++) { | |
217 | pages[i] = alloc_page(GFP_NOFS); | |
218 | if (!pages[i]) | |
219 | goto out_free; | |
220 | } | |
221 | ||
222 | memcpy(&dev->dev_id, d_id, sizeof(*d_id)); | |
223 | dev->layout_type = LAYOUT_BLOCK_VOLUME; | |
224 | dev->pages = pages; | |
225 | dev->pgbase = 0; | |
226 | dev->pglen = PAGE_SIZE * max_pages; | |
227 | dev->mincount = 0; | |
228 | ||
229 | dprintk("%s: dev_id: %s\n", __func__, dev->dev_id.data); | |
230 | rc = nfs4_proc_getdeviceinfo(server, dev); | |
231 | dprintk("%s getdevice info returns %d\n", __func__, rc); | |
232 | if (rc) | |
233 | goto out_free; | |
234 | ||
235 | rv = nfs4_blk_decode_device(server, dev); | |
236 | out_free: | |
237 | for (i = 0; i < max_pages; i++) | |
238 | __free_page(pages[i]); | |
239 | kfree(pages); | |
240 | kfree(dev); | |
241 | return rv; | |
242 | } | |
243 | ||
155e7524 FI |
244 | static int |
245 | bl_set_layoutdriver(struct nfs_server *server, const struct nfs_fh *fh) | |
246 | { | |
2f9fd182 FI |
247 | struct block_mount_id *b_mt_id = NULL; |
248 | struct pnfs_devicelist *dlist = NULL; | |
249 | struct pnfs_block_dev *bdev; | |
250 | LIST_HEAD(block_disklist); | |
251 | int status = 0, i; | |
252 | ||
155e7524 | 253 | dprintk("%s enter\n", __func__); |
2f9fd182 FI |
254 | |
255 | if (server->pnfs_blksize == 0) { | |
256 | dprintk("%s Server did not return blksize\n", __func__); | |
257 | return -EINVAL; | |
258 | } | |
259 | b_mt_id = kzalloc(sizeof(struct block_mount_id), GFP_NOFS); | |
260 | if (!b_mt_id) { | |
261 | status = -ENOMEM; | |
262 | goto out_error; | |
263 | } | |
264 | /* Initialize nfs4 block layout mount id */ | |
265 | spin_lock_init(&b_mt_id->bm_lock); | |
266 | INIT_LIST_HEAD(&b_mt_id->bm_devlist); | |
267 | ||
268 | dlist = kmalloc(sizeof(struct pnfs_devicelist), GFP_NOFS); | |
269 | if (!dlist) { | |
270 | status = -ENOMEM; | |
271 | goto out_error; | |
272 | } | |
273 | dlist->eof = 0; | |
274 | while (!dlist->eof) { | |
275 | status = nfs4_proc_getdevicelist(server, fh, dlist); | |
276 | if (status) | |
277 | goto out_error; | |
278 | dprintk("%s GETDEVICELIST numdevs=%i, eof=%i\n", | |
279 | __func__, dlist->num_devs, dlist->eof); | |
280 | for (i = 0; i < dlist->num_devs; i++) { | |
281 | bdev = nfs4_blk_get_deviceinfo(server, fh, | |
282 | &dlist->dev_id[i]); | |
283 | if (!bdev) { | |
284 | status = -ENODEV; | |
285 | goto out_error; | |
286 | } | |
287 | spin_lock(&b_mt_id->bm_lock); | |
288 | list_add(&bdev->bm_node, &b_mt_id->bm_devlist); | |
289 | spin_unlock(&b_mt_id->bm_lock); | |
290 | } | |
291 | } | |
292 | dprintk("%s SUCCESS\n", __func__); | |
293 | server->pnfs_ld_data = b_mt_id; | |
294 | ||
295 | out_return: | |
296 | kfree(dlist); | |
297 | return status; | |
298 | ||
299 | out_error: | |
300 | free_blk_mountid(b_mt_id); | |
301 | goto out_return; | |
155e7524 FI |
302 | } |
303 | ||
304 | static int | |
305 | bl_clear_layoutdriver(struct nfs_server *server) | |
306 | { | |
2f9fd182 FI |
307 | struct block_mount_id *b_mt_id = server->pnfs_ld_data; |
308 | ||
155e7524 | 309 | dprintk("%s enter\n", __func__); |
2f9fd182 FI |
310 | free_blk_mountid(b_mt_id); |
311 | dprintk("%s RETURNS\n", __func__); | |
155e7524 FI |
312 | return 0; |
313 | } | |
314 | ||
e9643fe8 BH |
315 | static const struct nfs_pageio_ops bl_pg_read_ops = { |
316 | .pg_init = pnfs_generic_pg_init_read, | |
317 | .pg_test = pnfs_generic_pg_test, | |
318 | .pg_doio = pnfs_generic_pg_readpages, | |
319 | }; | |
320 | ||
321 | static const struct nfs_pageio_ops bl_pg_write_ops = { | |
322 | .pg_init = pnfs_generic_pg_init_write, | |
323 | .pg_test = pnfs_generic_pg_test, | |
324 | .pg_doio = pnfs_generic_pg_writepages, | |
325 | }; | |
326 | ||
155e7524 FI |
327 | static struct pnfs_layoutdriver_type blocklayout_type = { |
328 | .id = LAYOUT_BLOCK_VOLUME, | |
329 | .name = "LAYOUT_BLOCK_VOLUME", | |
330 | .read_pagelist = bl_read_pagelist, | |
331 | .write_pagelist = bl_write_pagelist, | |
332 | .alloc_layout_hdr = bl_alloc_layout_hdr, | |
333 | .free_layout_hdr = bl_free_layout_hdr, | |
334 | .alloc_lseg = bl_alloc_lseg, | |
335 | .free_lseg = bl_free_lseg, | |
336 | .encode_layoutcommit = bl_encode_layoutcommit, | |
337 | .cleanup_layoutcommit = bl_cleanup_layoutcommit, | |
338 | .set_layoutdriver = bl_set_layoutdriver, | |
339 | .clear_layoutdriver = bl_clear_layoutdriver, | |
e9643fe8 BH |
340 | .pg_read_ops = &bl_pg_read_ops, |
341 | .pg_write_ops = &bl_pg_write_ops, | |
155e7524 FI |
342 | }; |
343 | ||
fe0a9b74 JR |
344 | static const struct rpc_pipe_ops bl_upcall_ops = { |
345 | .upcall = bl_pipe_upcall, | |
346 | .downcall = bl_pipe_downcall, | |
347 | .destroy_msg = bl_pipe_destroy_msg, | |
348 | }; | |
349 | ||
155e7524 FI |
350 | static int __init nfs4blocklayout_init(void) |
351 | { | |
fe0a9b74 JR |
352 | struct vfsmount *mnt; |
353 | struct path path; | |
155e7524 FI |
354 | int ret; |
355 | ||
356 | dprintk("%s: NFSv4 Block Layout Driver Registering...\n", __func__); | |
357 | ||
358 | ret = pnfs_register_layoutdriver(&blocklayout_type); | |
fe0a9b74 JR |
359 | if (ret) |
360 | goto out; | |
361 | ||
362 | init_waitqueue_head(&bl_wq); | |
363 | ||
364 | mnt = rpc_get_mount(); | |
365 | if (IS_ERR(mnt)) { | |
366 | ret = PTR_ERR(mnt); | |
367 | goto out_remove; | |
368 | } | |
369 | ||
370 | ret = vfs_path_lookup(mnt->mnt_root, | |
371 | mnt, | |
372 | NFS_PIPE_DIRNAME, 0, &path); | |
373 | if (ret) | |
374 | goto out_remove; | |
375 | ||
376 | bl_device_pipe = rpc_mkpipe(path.dentry, "blocklayout", NULL, | |
377 | &bl_upcall_ops, 0); | |
378 | if (IS_ERR(bl_device_pipe)) { | |
379 | ret = PTR_ERR(bl_device_pipe); | |
380 | goto out_remove; | |
381 | } | |
382 | out: | |
383 | return ret; | |
384 | ||
385 | out_remove: | |
386 | pnfs_unregister_layoutdriver(&blocklayout_type); | |
155e7524 FI |
387 | return ret; |
388 | } | |
389 | ||
390 | static void __exit nfs4blocklayout_exit(void) | |
391 | { | |
392 | dprintk("%s: NFSv4 Block Layout Driver Unregistering...\n", | |
393 | __func__); | |
394 | ||
395 | pnfs_unregister_layoutdriver(&blocklayout_type); | |
fe0a9b74 | 396 | rpc_unlink(bl_device_pipe); |
155e7524 FI |
397 | } |
398 | ||
399 | MODULE_ALIAS("nfs-layouttype4-3"); | |
400 | ||
401 | module_init(nfs4blocklayout_init); | |
402 | module_exit(nfs4blocklayout_exit); |