Commit | Line | Data |
---|---|---|
09f5bf4e BH |
1 | /* |
2 | * pNFS Objects layout driver high level definitions | |
3 | * | |
4 | * Copyright (C) 2007 Panasas Inc. [year of first publication] | |
5 | * All rights reserved. | |
6 | * | |
7 | * Benny Halevy <bhalevy@panasas.com> | |
aa281ac6 | 8 | * Boaz Harrosh <ooo@electrozaur.com> |
09f5bf4e BH |
9 | * |
10 | * This program is free software; you can redistribute it and/or modify | |
11 | * it under the terms of the GNU General Public License version 2 | |
12 | * See the file COPYING included with this distribution for more details. | |
13 | * | |
14 | * Redistribution and use in source and binary forms, with or without | |
15 | * modification, are permitted provided that the following conditions | |
16 | * are met: | |
17 | * | |
18 | * 1. Redistributions of source code must retain the above copyright | |
19 | * notice, this list of conditions and the following disclaimer. | |
20 | * 2. Redistributions in binary form must reproduce the above copyright | |
21 | * notice, this list of conditions and the following disclaimer in the | |
22 | * documentation and/or other materials provided with the distribution. | |
23 | * 3. Neither the name of the Panasas company nor the names of its | |
24 | * contributors may be used to endorse or promote products derived | |
25 | * from this software without specific prior written permission. | |
26 | * | |
27 | * THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED | |
28 | * WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF | |
29 | * MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE | |
30 | * DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE | |
31 | * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR | |
32 | * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF | |
33 | * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR | |
34 | * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF | |
35 | * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING | |
36 | * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS | |
37 | * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
38 | */ | |
39 | ||
18d98f6c SB |
40 | #include <linux/kmod.h> |
41 | #include <linux/moduleparam.h> | |
42 | #include <linux/ratelimit.h> | |
09f5bf4e BH |
43 | #include <scsi/osd_initiator.h> |
44 | #include "objlayout.h" | |
45 | ||
46 | #define NFSDBG_FACILITY NFSDBG_PNFS_LD | |
e51b841d BH |
47 | /* |
48 | * Create a objlayout layout structure for the given inode and return it. | |
49 | */ | |
50 | struct pnfs_layout_hdr * | |
51 | objlayout_alloc_layout_hdr(struct inode *inode, gfp_t gfp_flags) | |
52 | { | |
53 | struct objlayout *objlay; | |
54 | ||
55 | objlay = kzalloc(sizeof(struct objlayout), gfp_flags); | |
0aa61e78 TM |
56 | if (!objlay) |
57 | return NULL; | |
58 | spin_lock_init(&objlay->lock); | |
59 | INIT_LIST_HEAD(&objlay->err_list); | |
e51b841d BH |
60 | dprintk("%s: Return %p\n", __func__, objlay); |
61 | return &objlay->pnfs_layout; | |
62 | } | |
63 | ||
64 | /* | |
65 | * Free an objlayout layout structure | |
66 | */ | |
67 | void | |
68 | objlayout_free_layout_hdr(struct pnfs_layout_hdr *lo) | |
69 | { | |
70 | struct objlayout *objlay = OBJLAYOUT(lo); | |
71 | ||
72 | dprintk("%s: objlay %p\n", __func__, objlay); | |
73 | ||
adb58535 | 74 | WARN_ON(!list_empty(&objlay->err_list)); |
e51b841d BH |
75 | kfree(objlay); |
76 | } | |
77 | ||
09f5bf4e BH |
78 | /* |
79 | * Unmarshall layout and store it in pnfslay. | |
80 | */ | |
81 | struct pnfs_layout_segment * | |
82 | objlayout_alloc_lseg(struct pnfs_layout_hdr *pnfslay, | |
83 | struct nfs4_layoutget_res *lgr, | |
84 | gfp_t gfp_flags) | |
85 | { | |
86 | int status = -ENOMEM; | |
87 | struct xdr_stream stream; | |
88 | struct xdr_buf buf = { | |
89 | .pages = lgr->layoutp->pages, | |
90 | .page_len = lgr->layoutp->len, | |
91 | .buflen = lgr->layoutp->len, | |
92 | .len = lgr->layoutp->len, | |
93 | }; | |
94 | struct page *scratch; | |
95 | struct pnfs_layout_segment *lseg; | |
96 | ||
97 | dprintk("%s: Begin pnfslay %p\n", __func__, pnfslay); | |
98 | ||
99 | scratch = alloc_page(gfp_flags); | |
100 | if (!scratch) | |
101 | goto err_nofree; | |
102 | ||
103 | xdr_init_decode(&stream, &buf, NULL); | |
104 | xdr_set_scratch_buffer(&stream, page_address(scratch), PAGE_SIZE); | |
105 | ||
106 | status = objio_alloc_lseg(&lseg, pnfslay, &lgr->range, &stream, gfp_flags); | |
107 | if (unlikely(status)) { | |
108 | dprintk("%s: objio_alloc_lseg Return err %d\n", __func__, | |
109 | status); | |
110 | goto err; | |
111 | } | |
112 | ||
113 | __free_page(scratch); | |
114 | ||
115 | dprintk("%s: Return %p\n", __func__, lseg); | |
116 | return lseg; | |
117 | ||
118 | err: | |
119 | __free_page(scratch); | |
120 | err_nofree: | |
121 | dprintk("%s: Err Return=>%d\n", __func__, status); | |
122 | return ERR_PTR(status); | |
123 | } | |
124 | ||
125 | /* | |
126 | * Free a layout segement | |
127 | */ | |
128 | void | |
129 | objlayout_free_lseg(struct pnfs_layout_segment *lseg) | |
130 | { | |
131 | dprintk("%s: freeing layout segment %p\n", __func__, lseg); | |
132 | ||
133 | if (unlikely(!lseg)) | |
134 | return; | |
135 | ||
136 | objio_free_lseg(lseg); | |
137 | } | |
138 | ||
04f83450 BH |
139 | /* |
140 | * I/O Operations | |
141 | */ | |
142 | static inline u64 | |
143 | end_offset(u64 start, u64 len) | |
144 | { | |
145 | u64 end; | |
146 | ||
147 | end = start + len; | |
148 | return end >= start ? end : NFS4_MAX_UINT64; | |
149 | } | |
150 | ||
17280175 | 151 | static void _fix_verify_io_params(struct pnfs_layout_segment *lseg, |
96218556 BH |
152 | struct page ***p_pages, unsigned *p_pgbase, |
153 | u64 offset, unsigned long count) | |
04f83450 | 154 | { |
04f83450 BH |
155 | u64 lseg_end_offset; |
156 | ||
04f83450 BH |
157 | BUG_ON(offset < lseg->pls_range.offset); |
158 | lseg_end_offset = end_offset(lseg->pls_range.offset, | |
159 | lseg->pls_range.length); | |
160 | BUG_ON(offset >= lseg_end_offset); | |
96218556 | 161 | WARN_ON(offset + count > lseg_end_offset); |
04f83450 | 162 | |
96218556 BH |
163 | if (*p_pgbase > PAGE_SIZE) { |
164 | dprintk("%s: pgbase(0x%x) > PAGE_SIZE\n", __func__, *p_pgbase); | |
165 | *p_pages += *p_pgbase >> PAGE_SHIFT; | |
166 | *p_pgbase &= ~PAGE_MASK; | |
04f83450 | 167 | } |
04f83450 BH |
168 | } |
169 | ||
170 | /* | |
171 | * I/O done common code | |
172 | */ | |
173 | static void | |
e2e04355 | 174 | objlayout_iodone(struct objlayout_io_res *oir) |
04f83450 | 175 | { |
e2e04355 BH |
176 | if (likely(oir->status >= 0)) { |
177 | objio_free_result(oir); | |
adb58535 | 178 | } else { |
e2e04355 | 179 | struct objlayout *objlay = oir->objlay; |
adb58535 BH |
180 | |
181 | spin_lock(&objlay->lock); | |
a0fe8bf4 | 182 | objlay->delta_space_valid = OBJ_DSU_INVALID; |
e2e04355 | 183 | list_add(&objlay->err_list, &oir->err_list); |
adb58535 BH |
184 | spin_unlock(&objlay->lock); |
185 | } | |
186 | } | |
187 | ||
188 | /* | |
189 | * objlayout_io_set_result - Set an osd_error code on a specific osd comp. | |
190 | * | |
191 | * The @index component IO failed (error returned from target). Register | |
192 | * the error for later reporting at layout-return. | |
193 | */ | |
194 | void | |
e2e04355 | 195 | objlayout_io_set_result(struct objlayout_io_res *oir, unsigned index, |
adb58535 BH |
196 | struct pnfs_osd_objid *pooid, int osd_error, |
197 | u64 offset, u64 length, bool is_write) | |
198 | { | |
e2e04355 | 199 | struct pnfs_osd_ioerr *ioerr = &oir->ioerrs[index]; |
adb58535 | 200 | |
e2e04355 | 201 | BUG_ON(index >= oir->num_comps); |
adb58535 BH |
202 | if (osd_error) { |
203 | ioerr->oer_component = *pooid; | |
204 | ioerr->oer_comp_offset = offset; | |
205 | ioerr->oer_comp_length = length; | |
206 | ioerr->oer_iswrite = is_write; | |
207 | ioerr->oer_errno = osd_error; | |
208 | ||
209 | dprintk("%s: err[%d]: errno=%d is_write=%d dev(%llx:%llx) " | |
210 | "par=0x%llx obj=0x%llx offset=0x%llx length=0x%llx\n", | |
211 | __func__, index, ioerr->oer_errno, | |
212 | ioerr->oer_iswrite, | |
213 | _DEVID_LO(&ioerr->oer_component.oid_device_id), | |
214 | _DEVID_HI(&ioerr->oer_component.oid_device_id), | |
215 | ioerr->oer_component.oid_partition_id, | |
216 | ioerr->oer_component.oid_object_id, | |
217 | ioerr->oer_comp_offset, | |
218 | ioerr->oer_comp_length); | |
219 | } else { | |
220 | /* User need not call if no error is reported */ | |
221 | ioerr->oer_errno = 0; | |
222 | } | |
04f83450 BH |
223 | } |
224 | ||
225 | /* Function scheduled on rpc workqueue to call ->nfs_readlist_complete(). | |
226 | * This is because the osd completion is called with ints-off from | |
227 | * the block layer | |
228 | */ | |
229 | static void _rpc_read_complete(struct work_struct *work) | |
230 | { | |
231 | struct rpc_task *task; | |
d45f60c6 | 232 | struct nfs_pgio_header *hdr; |
04f83450 BH |
233 | |
234 | dprintk("%s enter\n", __func__); | |
235 | task = container_of(work, struct rpc_task, u.tk_work); | |
d45f60c6 | 236 | hdr = container_of(task, struct nfs_pgio_header, task); |
04f83450 | 237 | |
d45f60c6 | 238 | pnfs_ld_read_done(hdr); |
04f83450 BH |
239 | } |
240 | ||
241 | void | |
e2e04355 | 242 | objlayout_read_done(struct objlayout_io_res *oir, ssize_t status, bool sync) |
04f83450 | 243 | { |
d45f60c6 | 244 | struct nfs_pgio_header *hdr = oir->rpcdata; |
04f83450 | 245 | |
d45f60c6 | 246 | oir->status = hdr->task.tk_status = status; |
4cdc685c | 247 | if (status >= 0) |
d45f60c6 | 248 | hdr->res.count = status; |
5c0b4129 | 249 | else |
d45f60c6 | 250 | hdr->pnfs_error = status; |
e2e04355 BH |
251 | objlayout_iodone(oir); |
252 | /* must not use oir after this point */ | |
04f83450 | 253 | |
96218556 | 254 | dprintk("%s: Return status=%zd eof=%d sync=%d\n", __func__, |
d45f60c6 | 255 | status, hdr->res.eof, sync); |
96218556 | 256 | |
04f83450 | 257 | if (sync) |
d45f60c6 | 258 | pnfs_ld_read_done(hdr); |
04f83450 | 259 | else { |
d45f60c6 WAA |
260 | INIT_WORK(&hdr->task.u.tk_work, _rpc_read_complete); |
261 | schedule_work(&hdr->task.u.tk_work); | |
04f83450 BH |
262 | } |
263 | } | |
264 | ||
265 | /* | |
266 | * Perform sync or async reads. | |
267 | */ | |
268 | enum pnfs_try_status | |
d45f60c6 | 269 | objlayout_read_pagelist(struct nfs_pgio_header *hdr) |
04f83450 | 270 | { |
cd841605 | 271 | struct inode *inode = hdr->inode; |
d45f60c6 WAA |
272 | loff_t offset = hdr->args.offset; |
273 | size_t count = hdr->args.count; | |
e6c40fe3 | 274 | int err; |
04f83450 BH |
275 | loff_t eof; |
276 | ||
cd841605 | 277 | eof = i_size_read(inode); |
04f83450 BH |
278 | if (unlikely(offset + count > eof)) { |
279 | if (offset >= eof) { | |
e6c40fe3 | 280 | err = 0; |
d45f60c6 WAA |
281 | hdr->res.count = 0; |
282 | hdr->res.eof = 1; | |
4cdc685c | 283 | /*FIXME: do we need to call pnfs_ld_read_done() */ |
04f83450 BH |
284 | goto out; |
285 | } | |
286 | count = eof - offset; | |
287 | } | |
288 | ||
d45f60c6 WAA |
289 | hdr->res.eof = (offset + count) >= eof; |
290 | _fix_verify_io_params(hdr->lseg, &hdr->args.pages, | |
291 | &hdr->args.pgbase, | |
292 | hdr->args.offset, hdr->args.count); | |
4cdc685c | 293 | |
e6c40fe3 | 294 | dprintk("%s: inode(%lx) offset 0x%llx count 0x%Zx eof=%d\n", |
d45f60c6 | 295 | __func__, inode->i_ino, offset, count, hdr->res.eof); |
04f83450 | 296 | |
d45f60c6 | 297 | err = objio_read_pagelist(hdr); |
04f83450 | 298 | out: |
e6c40fe3 | 299 | if (unlikely(err)) { |
cd841605 | 300 | hdr->pnfs_error = err; |
e6c40fe3 BH |
301 | dprintk("%s: Returned Error %d\n", __func__, err); |
302 | return PNFS_NOT_ATTEMPTED; | |
303 | } | |
04f83450 BH |
304 | return PNFS_ATTEMPTED; |
305 | } | |
306 | ||
307 | /* Function scheduled on rpc workqueue to call ->nfs_writelist_complete(). | |
308 | * This is because the osd completion is called with ints-off from | |
309 | * the block layer | |
310 | */ | |
311 | static void _rpc_write_complete(struct work_struct *work) | |
312 | { | |
313 | struct rpc_task *task; | |
d45f60c6 | 314 | struct nfs_pgio_header *hdr; |
04f83450 BH |
315 | |
316 | dprintk("%s enter\n", __func__); | |
317 | task = container_of(work, struct rpc_task, u.tk_work); | |
d45f60c6 | 318 | hdr = container_of(task, struct nfs_pgio_header, task); |
04f83450 | 319 | |
d45f60c6 | 320 | pnfs_ld_write_done(hdr); |
04f83450 BH |
321 | } |
322 | ||
323 | void | |
e2e04355 | 324 | objlayout_write_done(struct objlayout_io_res *oir, ssize_t status, bool sync) |
04f83450 | 325 | { |
d45f60c6 | 326 | struct nfs_pgio_header *hdr = oir->rpcdata; |
04f83450 | 327 | |
d45f60c6 | 328 | oir->status = hdr->task.tk_status = status; |
04f83450 | 329 | if (status >= 0) { |
d45f60c6 | 330 | hdr->res.count = status; |
c65e6254 | 331 | hdr->verf.committed = oir->committed; |
5c0b4129 | 332 | } else { |
d45f60c6 | 333 | hdr->pnfs_error = status; |
96218556 | 334 | } |
e2e04355 | 335 | objlayout_iodone(oir); |
96218556 BH |
336 | /* must not use oir after this point */ |
337 | ||
338 | dprintk("%s: Return status %zd committed %d sync=%d\n", __func__, | |
c65e6254 | 339 | status, hdr->verf.committed, sync); |
04f83450 BH |
340 | |
341 | if (sync) | |
d45f60c6 | 342 | pnfs_ld_write_done(hdr); |
04f83450 | 343 | else { |
d45f60c6 WAA |
344 | INIT_WORK(&hdr->task.u.tk_work, _rpc_write_complete); |
345 | schedule_work(&hdr->task.u.tk_work); | |
04f83450 BH |
346 | } |
347 | } | |
348 | ||
349 | /* | |
350 | * Perform sync or async writes. | |
351 | */ | |
352 | enum pnfs_try_status | |
d45f60c6 | 353 | objlayout_write_pagelist(struct nfs_pgio_header *hdr, int how) |
04f83450 | 354 | { |
e6c40fe3 | 355 | int err; |
04f83450 | 356 | |
d45f60c6 WAA |
357 | _fix_verify_io_params(hdr->lseg, &hdr->args.pages, |
358 | &hdr->args.pgbase, | |
359 | hdr->args.offset, hdr->args.count); | |
04f83450 | 360 | |
d45f60c6 | 361 | err = objio_write_pagelist(hdr, how); |
e6c40fe3 | 362 | if (unlikely(err)) { |
cd841605 | 363 | hdr->pnfs_error = err; |
e6c40fe3 BH |
364 | dprintk("%s: Returned Error %d\n", __func__, err); |
365 | return PNFS_NOT_ATTEMPTED; | |
366 | } | |
04f83450 BH |
367 | return PNFS_ATTEMPTED; |
368 | } | |
369 | ||
a0fe8bf4 BH |
370 | void |
371 | objlayout_encode_layoutcommit(struct pnfs_layout_hdr *pnfslay, | |
372 | struct xdr_stream *xdr, | |
373 | const struct nfs4_layoutcommit_args *args) | |
374 | { | |
375 | struct objlayout *objlay = OBJLAYOUT(pnfslay); | |
376 | struct pnfs_osd_layoutupdate lou; | |
377 | __be32 *start; | |
378 | ||
379 | dprintk("%s: Begin\n", __func__); | |
380 | ||
381 | spin_lock(&objlay->lock); | |
382 | lou.dsu_valid = (objlay->delta_space_valid == OBJ_DSU_VALID); | |
383 | lou.dsu_delta = objlay->delta_space_used; | |
384 | objlay->delta_space_used = 0; | |
385 | objlay->delta_space_valid = OBJ_DSU_INIT; | |
386 | lou.olu_ioerr_flag = !list_empty(&objlay->err_list); | |
387 | spin_unlock(&objlay->lock); | |
388 | ||
389 | start = xdr_reserve_space(xdr, 4); | |
390 | ||
391 | BUG_ON(pnfs_osd_xdr_encode_layoutupdate(xdr, &lou)); | |
392 | ||
393 | *start = cpu_to_be32((xdr->p - start - 1) * 4); | |
394 | ||
395 | dprintk("%s: Return delta_space_used %lld err %d\n", __func__, | |
396 | lou.dsu_delta, lou.olu_ioerr_flag); | |
397 | } | |
398 | ||
adb58535 BH |
399 | static int |
400 | err_prio(u32 oer_errno) | |
401 | { | |
402 | switch (oer_errno) { | |
403 | case 0: | |
404 | return 0; | |
405 | ||
406 | case PNFS_OSD_ERR_RESOURCE: | |
407 | return OSD_ERR_PRI_RESOURCE; | |
408 | case PNFS_OSD_ERR_BAD_CRED: | |
409 | return OSD_ERR_PRI_BAD_CRED; | |
410 | case PNFS_OSD_ERR_NO_ACCESS: | |
411 | return OSD_ERR_PRI_NO_ACCESS; | |
412 | case PNFS_OSD_ERR_UNREACHABLE: | |
413 | return OSD_ERR_PRI_UNREACHABLE; | |
414 | case PNFS_OSD_ERR_NOT_FOUND: | |
415 | return OSD_ERR_PRI_NOT_FOUND; | |
416 | case PNFS_OSD_ERR_NO_SPACE: | |
417 | return OSD_ERR_PRI_NO_SPACE; | |
418 | default: | |
419 | WARN_ON(1); | |
420 | /* fallthrough */ | |
421 | case PNFS_OSD_ERR_EIO: | |
422 | return OSD_ERR_PRI_EIO; | |
423 | } | |
424 | } | |
425 | ||
426 | static void | |
427 | merge_ioerr(struct pnfs_osd_ioerr *dest_err, | |
428 | const struct pnfs_osd_ioerr *src_err) | |
429 | { | |
430 | u64 dest_end, src_end; | |
431 | ||
432 | if (!dest_err->oer_errno) { | |
433 | *dest_err = *src_err; | |
434 | /* accumulated device must be blank */ | |
435 | memset(&dest_err->oer_component.oid_device_id, 0, | |
436 | sizeof(dest_err->oer_component.oid_device_id)); | |
437 | ||
438 | return; | |
439 | } | |
440 | ||
441 | if (dest_err->oer_component.oid_partition_id != | |
442 | src_err->oer_component.oid_partition_id) | |
443 | dest_err->oer_component.oid_partition_id = 0; | |
444 | ||
445 | if (dest_err->oer_component.oid_object_id != | |
446 | src_err->oer_component.oid_object_id) | |
447 | dest_err->oer_component.oid_object_id = 0; | |
448 | ||
449 | if (dest_err->oer_comp_offset > src_err->oer_comp_offset) | |
450 | dest_err->oer_comp_offset = src_err->oer_comp_offset; | |
451 | ||
452 | dest_end = end_offset(dest_err->oer_comp_offset, | |
453 | dest_err->oer_comp_length); | |
454 | src_end = end_offset(src_err->oer_comp_offset, | |
455 | src_err->oer_comp_length); | |
456 | if (dest_end < src_end) | |
457 | dest_end = src_end; | |
458 | ||
459 | dest_err->oer_comp_length = dest_end - dest_err->oer_comp_offset; | |
460 | ||
461 | if ((src_err->oer_iswrite == dest_err->oer_iswrite) && | |
462 | (err_prio(src_err->oer_errno) > err_prio(dest_err->oer_errno))) { | |
463 | dest_err->oer_errno = src_err->oer_errno; | |
464 | } else if (src_err->oer_iswrite) { | |
465 | dest_err->oer_iswrite = true; | |
466 | dest_err->oer_errno = src_err->oer_errno; | |
467 | } | |
468 | } | |
469 | ||
470 | static void | |
471 | encode_accumulated_error(struct objlayout *objlay, __be32 *p) | |
472 | { | |
e2e04355 | 473 | struct objlayout_io_res *oir, *tmp; |
adb58535 BH |
474 | struct pnfs_osd_ioerr accumulated_err = {.oer_errno = 0}; |
475 | ||
e2e04355 | 476 | list_for_each_entry_safe(oir, tmp, &objlay->err_list, err_list) { |
adb58535 BH |
477 | unsigned i; |
478 | ||
e2e04355 BH |
479 | for (i = 0; i < oir->num_comps; i++) { |
480 | struct pnfs_osd_ioerr *ioerr = &oir->ioerrs[i]; | |
adb58535 BH |
481 | |
482 | if (!ioerr->oer_errno) | |
483 | continue; | |
484 | ||
a030889a WAA |
485 | printk(KERN_ERR "NFS: %s: err[%d]: errno=%d " |
486 | "is_write=%d dev(%llx:%llx) par=0x%llx " | |
487 | "obj=0x%llx offset=0x%llx length=0x%llx\n", | |
adb58535 BH |
488 | __func__, i, ioerr->oer_errno, |
489 | ioerr->oer_iswrite, | |
490 | _DEVID_LO(&ioerr->oer_component.oid_device_id), | |
491 | _DEVID_HI(&ioerr->oer_component.oid_device_id), | |
492 | ioerr->oer_component.oid_partition_id, | |
493 | ioerr->oer_component.oid_object_id, | |
494 | ioerr->oer_comp_offset, | |
495 | ioerr->oer_comp_length); | |
496 | ||
497 | merge_ioerr(&accumulated_err, ioerr); | |
498 | } | |
e2e04355 BH |
499 | list_del(&oir->err_list); |
500 | objio_free_result(oir); | |
adb58535 BH |
501 | } |
502 | ||
503 | pnfs_osd_xdr_encode_ioerr(p, &accumulated_err); | |
504 | } | |
505 | ||
506 | void | |
507 | objlayout_encode_layoutreturn(struct pnfs_layout_hdr *pnfslay, | |
508 | struct xdr_stream *xdr, | |
509 | const struct nfs4_layoutreturn_args *args) | |
510 | { | |
511 | struct objlayout *objlay = OBJLAYOUT(pnfslay); | |
e2e04355 | 512 | struct objlayout_io_res *oir, *tmp; |
adb58535 BH |
513 | __be32 *start; |
514 | ||
515 | dprintk("%s: Begin\n", __func__); | |
516 | start = xdr_reserve_space(xdr, 4); | |
517 | BUG_ON(!start); | |
518 | ||
519 | spin_lock(&objlay->lock); | |
520 | ||
e2e04355 | 521 | list_for_each_entry_safe(oir, tmp, &objlay->err_list, err_list) { |
adb58535 BH |
522 | __be32 *last_xdr = NULL, *p; |
523 | unsigned i; | |
524 | int res = 0; | |
525 | ||
e2e04355 BH |
526 | for (i = 0; i < oir->num_comps; i++) { |
527 | struct pnfs_osd_ioerr *ioerr = &oir->ioerrs[i]; | |
adb58535 BH |
528 | |
529 | if (!ioerr->oer_errno) | |
530 | continue; | |
531 | ||
532 | dprintk("%s: err[%d]: errno=%d is_write=%d " | |
533 | "dev(%llx:%llx) par=0x%llx obj=0x%llx " | |
534 | "offset=0x%llx length=0x%llx\n", | |
535 | __func__, i, ioerr->oer_errno, | |
536 | ioerr->oer_iswrite, | |
537 | _DEVID_LO(&ioerr->oer_component.oid_device_id), | |
538 | _DEVID_HI(&ioerr->oer_component.oid_device_id), | |
539 | ioerr->oer_component.oid_partition_id, | |
540 | ioerr->oer_component.oid_object_id, | |
541 | ioerr->oer_comp_offset, | |
542 | ioerr->oer_comp_length); | |
543 | ||
544 | p = pnfs_osd_xdr_ioerr_reserve_space(xdr); | |
545 | if (unlikely(!p)) { | |
546 | res = -E2BIG; | |
547 | break; /* accumulated_error */ | |
548 | } | |
549 | ||
550 | last_xdr = p; | |
e2e04355 | 551 | pnfs_osd_xdr_encode_ioerr(p, &oir->ioerrs[i]); |
adb58535 BH |
552 | } |
553 | ||
554 | /* TODO: use xdr_write_pages */ | |
555 | if (unlikely(res)) { | |
556 | /* no space for even one error descriptor */ | |
557 | BUG_ON(!last_xdr); | |
558 | ||
559 | /* we've encountered a situation with lots and lots of | |
560 | * errors and no space to encode them all. Use the last | |
561 | * available slot to report the union of all the | |
562 | * remaining errors. | |
563 | */ | |
564 | encode_accumulated_error(objlay, last_xdr); | |
565 | goto loop_done; | |
566 | } | |
e2e04355 BH |
567 | list_del(&oir->err_list); |
568 | objio_free_result(oir); | |
adb58535 BH |
569 | } |
570 | loop_done: | |
571 | spin_unlock(&objlay->lock); | |
572 | ||
573 | *start = cpu_to_be32((xdr->p - start - 1) * 4); | |
574 | dprintk("%s: Return\n", __func__); | |
575 | } | |
576 | ||
577 | ||
b6c05f16 BH |
578 | /* |
579 | * Get Device Info API for io engines | |
580 | */ | |
581 | struct objlayout_deviceinfo { | |
582 | struct page *page; | |
583 | struct pnfs_osd_deviceaddr da; /* This must be last */ | |
584 | }; | |
585 | ||
586 | /* Initialize and call nfs_getdeviceinfo, then decode and return a | |
587 | * "struct pnfs_osd_deviceaddr *" Eventually objlayout_put_deviceinfo() | |
588 | * should be called. | |
589 | */ | |
590 | int objlayout_get_deviceinfo(struct pnfs_layout_hdr *pnfslay, | |
591 | struct nfs4_deviceid *d_id, struct pnfs_osd_deviceaddr **deviceaddr, | |
592 | gfp_t gfp_flags) | |
593 | { | |
594 | struct objlayout_deviceinfo *odi; | |
595 | struct pnfs_device pd; | |
b6c05f16 BH |
596 | struct page *page, **pages; |
597 | u32 *p; | |
598 | int err; | |
599 | ||
600 | page = alloc_page(gfp_flags); | |
601 | if (!page) | |
602 | return -ENOMEM; | |
603 | ||
604 | pages = &page; | |
605 | pd.pages = pages; | |
606 | ||
607 | memcpy(&pd.dev_id, d_id, sizeof(*d_id)); | |
608 | pd.layout_type = LAYOUT_OSD2_OBJECTS; | |
609 | pd.pages = &page; | |
610 | pd.pgbase = 0; | |
611 | pd.pglen = PAGE_SIZE; | |
612 | pd.mincount = 0; | |
52fcac98 | 613 | pd.maxcount = PAGE_SIZE; |
b6c05f16 | 614 | |
cd5875fe TM |
615 | err = nfs4_proc_getdeviceinfo(NFS_SERVER(pnfslay->plh_inode), &pd, |
616 | pnfslay->plh_lc_cred); | |
b6c05f16 BH |
617 | dprintk("%s nfs_getdeviceinfo returned %d\n", __func__, err); |
618 | if (err) | |
619 | goto err_out; | |
620 | ||
621 | p = page_address(page); | |
622 | odi = kzalloc(sizeof(*odi), gfp_flags); | |
623 | if (!odi) { | |
624 | err = -ENOMEM; | |
625 | goto err_out; | |
626 | } | |
627 | pnfs_osd_xdr_decode_deviceaddr(&odi->da, p); | |
628 | odi->page = page; | |
629 | *deviceaddr = &odi->da; | |
630 | return 0; | |
631 | ||
632 | err_out: | |
633 | __free_page(page); | |
634 | return err; | |
635 | } | |
636 | ||
637 | void objlayout_put_deviceinfo(struct pnfs_osd_deviceaddr *deviceaddr) | |
638 | { | |
639 | struct objlayout_deviceinfo *odi = container_of(deviceaddr, | |
640 | struct objlayout_deviceinfo, | |
641 | da); | |
642 | ||
643 | __free_page(odi->page); | |
644 | kfree(odi); | |
645 | } | |
18d98f6c SB |
646 | |
647 | enum { | |
648 | OBJLAYOUT_MAX_URI_LEN = 256, OBJLAYOUT_MAX_OSDNAME_LEN = 64, | |
649 | OBJLAYOUT_MAX_SYSID_HEX_LEN = OSD_SYSTEMID_LEN * 2 + 1, | |
650 | OSD_LOGIN_UPCALL_PATHLEN = 256 | |
651 | }; | |
652 | ||
653 | static char osd_login_prog[OSD_LOGIN_UPCALL_PATHLEN] = "/sbin/osd_login"; | |
654 | ||
655 | module_param_string(osd_login_prog, osd_login_prog, sizeof(osd_login_prog), | |
656 | 0600); | |
657 | MODULE_PARM_DESC(osd_login_prog, "Path to the osd_login upcall program"); | |
658 | ||
659 | struct __auto_login { | |
660 | char uri[OBJLAYOUT_MAX_URI_LEN]; | |
661 | char osdname[OBJLAYOUT_MAX_OSDNAME_LEN]; | |
662 | char systemid_hex[OBJLAYOUT_MAX_SYSID_HEX_LEN]; | |
663 | }; | |
664 | ||
665 | static int __objlayout_upcall(struct __auto_login *login) | |
666 | { | |
667 | static char *envp[] = { "HOME=/", | |
668 | "TERM=linux", | |
669 | "PATH=/sbin:/usr/sbin:/bin:/usr/bin", | |
670 | NULL | |
671 | }; | |
672 | char *argv[8]; | |
673 | int ret; | |
674 | ||
675 | if (unlikely(!osd_login_prog[0])) { | |
676 | dprintk("%s: osd_login_prog is disabled\n", __func__); | |
677 | return -EACCES; | |
678 | } | |
679 | ||
680 | dprintk("%s uri: %s\n", __func__, login->uri); | |
681 | dprintk("%s osdname %s\n", __func__, login->osdname); | |
682 | dprintk("%s systemid_hex %s\n", __func__, login->systemid_hex); | |
683 | ||
684 | argv[0] = (char *)osd_login_prog; | |
685 | argv[1] = "-u"; | |
686 | argv[2] = login->uri; | |
687 | argv[3] = "-o"; | |
688 | argv[4] = login->osdname; | |
689 | argv[5] = "-s"; | |
690 | argv[6] = login->systemid_hex; | |
691 | argv[7] = NULL; | |
692 | ||
693 | ret = call_usermodehelper(argv[0], argv, envp, UMH_WAIT_PROC); | |
694 | /* | |
695 | * Disable the upcall mechanism if we're getting an ENOENT or | |
696 | * EACCES error. The admin can re-enable it on the fly by using | |
697 | * sysfs to set the objlayoutdriver.osd_login_prog module parameter once | |
698 | * the problem has been fixed. | |
699 | */ | |
700 | if (ret == -ENOENT || ret == -EACCES) { | |
701 | printk(KERN_ERR "PNFS-OBJ: %s was not found please set " | |
702 | "objlayoutdriver.osd_login_prog kernel parameter!\n", | |
703 | osd_login_prog); | |
704 | osd_login_prog[0] = '\0'; | |
705 | } | |
706 | dprintk("%s %s return value: %d\n", __func__, osd_login_prog, ret); | |
707 | ||
708 | return ret; | |
709 | } | |
710 | ||
711 | /* Assume dest is all zeros */ | |
712 | static void __copy_nfsS_and_zero_terminate(struct nfs4_string s, | |
713 | char *dest, int max_len, | |
714 | const char *var_name) | |
715 | { | |
716 | if (!s.len) | |
717 | return; | |
718 | ||
719 | if (s.len >= max_len) { | |
720 | pr_warn_ratelimited( | |
721 | "objlayout_autologin: %s: s.len(%d) >= max_len(%d)", | |
722 | var_name, s.len, max_len); | |
723 | s.len = max_len - 1; /* space for null terminator */ | |
724 | } | |
725 | ||
726 | memcpy(dest, s.data, s.len); | |
727 | } | |
728 | ||
729 | /* Assume sysid is all zeros */ | |
730 | static void _sysid_2_hex(struct nfs4_string s, | |
731 | char sysid[OBJLAYOUT_MAX_SYSID_HEX_LEN]) | |
732 | { | |
733 | int i; | |
734 | char *cur; | |
735 | ||
736 | if (!s.len) | |
737 | return; | |
738 | ||
739 | if (s.len != OSD_SYSTEMID_LEN) { | |
740 | pr_warn_ratelimited( | |
741 | "objlayout_autologin: systemid_len(%d) != OSD_SYSTEMID_LEN", | |
742 | s.len); | |
743 | if (s.len > OSD_SYSTEMID_LEN) | |
744 | s.len = OSD_SYSTEMID_LEN; | |
745 | } | |
746 | ||
747 | cur = sysid; | |
748 | for (i = 0; i < s.len; i++) | |
749 | cur = hex_byte_pack(cur, s.data[i]); | |
750 | } | |
751 | ||
752 | int objlayout_autologin(struct pnfs_osd_deviceaddr *deviceaddr) | |
753 | { | |
754 | int rc; | |
755 | struct __auto_login login; | |
756 | ||
757 | if (!deviceaddr->oda_targetaddr.ota_netaddr.r_addr.len) | |
758 | return -ENODEV; | |
759 | ||
760 | memset(&login, 0, sizeof(login)); | |
761 | __copy_nfsS_and_zero_terminate( | |
762 | deviceaddr->oda_targetaddr.ota_netaddr.r_addr, | |
763 | login.uri, sizeof(login.uri), "URI"); | |
764 | ||
765 | __copy_nfsS_and_zero_terminate( | |
766 | deviceaddr->oda_osdname, | |
767 | login.osdname, sizeof(login.osdname), "OSDNAME"); | |
768 | ||
769 | _sysid_2_hex(deviceaddr->oda_systemid, login.systemid_hex); | |
770 | ||
771 | rc = __objlayout_upcall(&login); | |
772 | if (rc > 0) /* script returns positive values */ | |
773 | rc = -ENODEV; | |
774 | ||
775 | return rc; | |
776 | } |