blob: 819624cfc8da4c7bad401ee26611d69b7ff38f98 [file] [log] [blame]
Boaz Harroshba9e5e92008-10-28 16:11:41 +02001/*
2 * Copyright (C) 2005, 2006
Boaz Harrosh27d2e142009-06-14 17:23:09 +03003 * Avishay Traeger (avishay@gmail.com)
Boaz Harroshba9e5e92008-10-28 16:11:41 +02004 * Copyright (C) 2008, 2009
Boaz Harroshaa281ac2014-10-19 19:38:58 +03005 * Boaz Harrosh <ooo@electrozaur.com>
Boaz Harroshba9e5e92008-10-28 16:11:41 +02006 *
7 * Copyrights for code taken from ext2:
8 * Copyright (C) 1992, 1993, 1994, 1995
9 * Remy Card (card@masi.ibp.fr)
10 * Laboratoire MASI - Institut Blaise Pascal
11 * Universite Pierre et Marie Curie (Paris VI)
12 * from
13 * linux/fs/minix/inode.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * This file is part of exofs.
17 *
18 * exofs is free software; you can redistribute it and/or modify
19 * it under the terms of the GNU General Public License as published by
20 * the Free Software Foundation. Since it is based on ext2, and the only
21 * valid version of GPL for the Linux kernel is version 2, the only valid
22 * version of GPL for exofs is version 2.
23 *
24 * exofs is distributed in the hope that it will be useful,
25 * but WITHOUT ANY WARRANTY; without even the implied warranty of
26 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
27 * GNU General Public License for more details.
28 *
29 * You should have received a copy of the GNU General Public License
30 * along with exofs; if not, write to the Free Software
31 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
32 */
33
34#include <linux/string.h>
35#include <linux/parser.h>
36#include <linux/vfs.h>
37#include <linux/random.h>
Paul Gortmaker143cb492011-07-01 14:23:34 -040038#include <linux/module.h>
Boaz Harrosh8cf74b32009-03-22 12:47:26 +020039#include <linux/exportfs.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090040#include <linux/slab.h>
Boaz Harroshba9e5e92008-10-28 16:11:41 +020041
42#include "exofs.h"
43
Boaz Harrosh85e44df2011-05-16 15:26:47 +030044#define EXOFS_DBGMSG2(M...) do {} while (0)
45
Boaz Harroshba9e5e92008-10-28 16:11:41 +020046/******************************************************************************
47 * MOUNT OPTIONS
48 *****************************************************************************/
49
50/*
51 * struct to hold what we get from mount options
52 */
53struct exofs_mountopt {
Boaz Harrosh9ed96482011-01-31 14:32:14 +020054 bool is_osdname;
Boaz Harroshba9e5e92008-10-28 16:11:41 +020055 const char *dev_name;
56 uint64_t pid;
57 int timeout;
58};
59
60/*
61 * exofs-specific mount-time options.
62 */
Boaz Harrosh9ed96482011-01-31 14:32:14 +020063enum { Opt_name, Opt_pid, Opt_to, Opt_err };
Boaz Harroshba9e5e92008-10-28 16:11:41 +020064
65/*
66 * Our mount-time options. These should ideally be 64-bit unsigned, but the
67 * kernel's parsing functions do not currently support that. 32-bit should be
68 * sufficient for most applications now.
69 */
70static match_table_t tokens = {
Boaz Harrosh9ed96482011-01-31 14:32:14 +020071 {Opt_name, "osdname=%s"},
Boaz Harroshba9e5e92008-10-28 16:11:41 +020072 {Opt_pid, "pid=%u"},
73 {Opt_to, "to=%u"},
74 {Opt_err, NULL}
75};
76
77/*
78 * The main option parsing method. Also makes sure that all of the mandatory
79 * mount options were set.
80 */
81static int parse_options(char *options, struct exofs_mountopt *opts)
82{
83 char *p;
84 substring_t args[MAX_OPT_ARGS];
85 int option;
86 bool s_pid = false;
87
88 EXOFS_DBGMSG("parse_options %s\n", options);
89 /* defaults */
90 memset(opts, 0, sizeof(*opts));
91 opts->timeout = BLK_DEFAULT_SG_TIMEOUT;
92
93 while ((p = strsep(&options, ",")) != NULL) {
94 int token;
95 char str[32];
96
97 if (!*p)
98 continue;
99
100 token = match_token(p, tokens, args);
101 switch (token) {
Boaz Harrosh9ed96482011-01-31 14:32:14 +0200102 case Opt_name:
103 opts->dev_name = match_strdup(&args[0]);
104 if (unlikely(!opts->dev_name)) {
105 EXOFS_ERR("Error allocating dev_name");
106 return -ENOMEM;
107 }
108 opts->is_osdname = true;
109 break;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200110 case Opt_pid:
111 if (0 == match_strlcpy(str, &args[0], sizeof(str)))
112 return -EINVAL;
113 opts->pid = simple_strtoull(str, NULL, 0);
114 if (opts->pid < EXOFS_MIN_PID) {
115 EXOFS_ERR("Partition ID must be >= %u",
116 EXOFS_MIN_PID);
117 return -EINVAL;
118 }
119 s_pid = 1;
120 break;
121 case Opt_to:
122 if (match_int(&args[0], &option))
123 return -EINVAL;
124 if (option <= 0) {
Masanari Iidac19ca6c2016-02-08 20:53:12 +0900125 EXOFS_ERR("Timeout must be > 0");
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200126 return -EINVAL;
127 }
128 opts->timeout = option * HZ;
129 break;
130 }
131 }
132
133 if (!s_pid) {
134 EXOFS_ERR("Need to specify the following options:\n");
135 EXOFS_ERR(" -o pid=pid_no_to_use\n");
136 return -EINVAL;
137 }
138
139 return 0;
140}
141
142/******************************************************************************
143 * INODE CACHE
144 *****************************************************************************/
145
146/*
147 * Our inode cache. Isn't it pretty?
148 */
149static struct kmem_cache *exofs_inode_cachep;
150
151/*
152 * Allocate an inode in the cache
153 */
154static struct inode *exofs_alloc_inode(struct super_block *sb)
155{
156 struct exofs_i_info *oi;
157
158 oi = kmem_cache_alloc(exofs_inode_cachep, GFP_KERNEL);
159 if (!oi)
160 return NULL;
161
162 oi->vfs_inode.i_version = 1;
163 return &oi->vfs_inode;
164}
165
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100166static void exofs_i_callback(struct rcu_head *head)
167{
168 struct inode *inode = container_of(head, struct inode, i_rcu);
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100169 kmem_cache_free(exofs_inode_cachep, exofs_i(inode));
170}
171
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200172/*
173 * Remove an inode from the cache
174 */
175static void exofs_destroy_inode(struct inode *inode)
176{
Nick Pigginfa0d7e3d2011-01-07 17:49:49 +1100177 call_rcu(&inode->i_rcu, exofs_i_callback);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200178}
179
180/*
181 * Initialize the inode
182 */
183static void exofs_init_once(void *foo)
184{
185 struct exofs_i_info *oi = foo;
186
187 inode_init_once(&oi->vfs_inode);
188}
189
190/*
191 * Create and initialize the inode cache
192 */
193static int init_inodecache(void)
194{
195 exofs_inode_cachep = kmem_cache_create("exofs_inode_cache",
196 sizeof(struct exofs_i_info), 0,
Vladimir Davydov5d097052016-01-14 15:18:21 -0800197 SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD |
198 SLAB_ACCOUNT, exofs_init_once);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200199 if (exofs_inode_cachep == NULL)
200 return -ENOMEM;
201 return 0;
202}
203
204/*
205 * Destroy the inode cache
206 */
207static void destroy_inodecache(void)
208{
Kirill A. Shutemov8c0a8532012-09-26 11:33:07 +1000209 /*
210 * Make sure all delayed rcu free inodes are flushed before we
211 * destroy cache.
212 */
213 rcu_barrier();
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200214 kmem_cache_destroy(exofs_inode_cachep);
215}
216
217/******************************************************************************
Boaz Harrosh85e44df2011-05-16 15:26:47 +0300218 * Some osd helpers
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200219 *****************************************************************************/
Boaz Harrosh85e44df2011-05-16 15:26:47 +0300220void exofs_make_credential(u8 cred_a[OSD_CAP_LEN], const struct osd_obj_id *obj)
221{
222 osd_sec_init_nosec_doall_caps(cred_a, obj, false, true);
223}
224
225static int exofs_read_kern(struct osd_dev *od, u8 *cred, struct osd_obj_id *obj,
226 u64 offset, void *p, unsigned length)
227{
228 struct osd_request *or = osd_start_request(od, GFP_KERNEL);
229/* struct osd_sense_info osi = {.key = 0};*/
230 int ret;
231
232 if (unlikely(!or)) {
233 EXOFS_DBGMSG("%s: osd_start_request failed.\n", __func__);
234 return -ENOMEM;
235 }
236 ret = osd_req_read_kern(or, obj, offset, p, length);
237 if (unlikely(ret)) {
238 EXOFS_DBGMSG("%s: osd_req_read_kern failed.\n", __func__);
239 goto out;
240 }
241
242 ret = osd_finalize_request(or, 0, cred, NULL);
243 if (unlikely(ret)) {
244 EXOFS_DBGMSG("Failed to osd_finalize_request() => %d\n", ret);
245 goto out;
246 }
247
248 ret = osd_execute_request(or);
249 if (unlikely(ret))
250 EXOFS_DBGMSG("osd_execute_request() => %d\n", ret);
251 /* osd_req_decode_sense(or, ret); */
252
253out:
254 osd_end_request(or);
255 EXOFS_DBGMSG2("read_kern(0x%llx) offset=0x%llx "
256 "length=0x%llx dev=%p ret=>%d\n",
257 _LLU(obj->id), _LLU(offset), _LLU(length), od, ret);
258 return ret;
259}
260
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200261static const struct osd_attr g_attr_sb_stats = ATTR_DEF(
262 EXOFS_APAGE_SB_DATA,
263 EXOFS_ATTR_SB_STATS,
264 sizeof(struct exofs_sb_stats));
265
266static int __sbi_read_stats(struct exofs_sb_info *sbi)
267{
268 struct osd_attr attrs[] = {
269 [0] = g_attr_sb_stats,
270 };
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700271 struct ore_io_state *ios;
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200272 int ret;
273
Boaz Harrosh5bf696d2011-09-28 11:39:59 +0300274 ret = ore_get_io_state(&sbi->layout, &sbi->oc, &ios);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200275 if (unlikely(ret)) {
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700276 EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200277 return ret;
278 }
279
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200280 ios->in_attr = attrs;
281 ios->in_attr_len = ARRAY_SIZE(attrs);
282
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700283 ret = ore_read(ios);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200284 if (unlikely(ret)) {
285 EXOFS_ERR("Error reading super_block stats => %d\n", ret);
286 goto out;
287 }
288
289 ret = extract_attr_from_ios(ios, &attrs[0]);
290 if (ret) {
291 EXOFS_ERR("%s: extract_attr of sb_stats failed\n", __func__);
292 goto out;
293 }
294 if (attrs[0].len) {
295 struct exofs_sb_stats *ess;
296
297 if (unlikely(attrs[0].len != sizeof(*ess))) {
298 EXOFS_ERR("%s: Wrong version of exofs_sb_stats "
299 "size(%d) != expected(%zd)\n",
300 __func__, attrs[0].len, sizeof(*ess));
301 goto out;
302 }
303
304 ess = attrs[0].val_ptr;
305 sbi->s_nextid = le64_to_cpu(ess->s_nextid);
306 sbi->s_numfiles = le32_to_cpu(ess->s_numfiles);
307 }
308
309out:
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700310 ore_put_io_state(ios);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200311 return ret;
312}
313
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700314static void stats_done(struct ore_io_state *ios, void *p)
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200315{
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700316 ore_put_io_state(ios);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200317 /* Good thanks nothing to do anymore */
318}
319
320/* Asynchronously write the stats attribute */
321int exofs_sbi_write_stats(struct exofs_sb_info *sbi)
322{
323 struct osd_attr attrs[] = {
324 [0] = g_attr_sb_stats,
325 };
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700326 struct ore_io_state *ios;
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200327 int ret;
328
Boaz Harrosh5bf696d2011-09-28 11:39:59 +0300329 ret = ore_get_io_state(&sbi->layout, &sbi->oc, &ios);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200330 if (unlikely(ret)) {
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700331 EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200332 return ret;
333 }
334
335 sbi->s_ess.s_nextid = cpu_to_le64(sbi->s_nextid);
336 sbi->s_ess.s_numfiles = cpu_to_le64(sbi->s_numfiles);
337 attrs[0].val_ptr = &sbi->s_ess;
338
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700339
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200340 ios->done = stats_done;
341 ios->private = sbi;
342 ios->out_attr = attrs;
343 ios->out_attr_len = ARRAY_SIZE(attrs);
344
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700345 ret = ore_write(ios);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200346 if (unlikely(ret)) {
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700347 EXOFS_ERR("%s: ore_write failed.\n", __func__);
348 ore_put_io_state(ios);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200349 }
350
351 return ret;
352}
353
Boaz Harrosh85e44df2011-05-16 15:26:47 +0300354/******************************************************************************
355 * SUPERBLOCK FUNCTIONS
356 *****************************************************************************/
357static const struct super_operations exofs_sops;
358static const struct export_operations exofs_export_ops;
359
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200360/*
361 * Write the superblock to the OSD
362 */
H Hartley Sweetende74b052011-09-23 13:46:51 -0700363static int exofs_sync_fs(struct super_block *sb, int wait)
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200364{
365 struct exofs_sb_info *sbi;
366 struct exofs_fscb *fscb;
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700367 struct ore_comp one_comp;
Boaz Harrosh5bf696d2011-09-28 11:39:59 +0300368 struct ore_components oc;
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700369 struct ore_io_state *ios;
Christoph Hellwig80e09fb2009-06-08 10:03:58 +0200370 int ret = -ENOMEM;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200371
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200372 fscb = kmalloc(sizeof(*fscb), GFP_KERNEL);
373 if (unlikely(!fscb))
374 return -ENOMEM;
Boaz Harrosh06886a52009-11-08 14:54:08 +0200375
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200376 sbi = sb->s_fs_info;
377
378 /* NOTE: We no longer dirty the super_block anywhere in exofs. The
379 * reason we write the fscb here on unmount is so we can stay backwards
380 * compatible with fscb->s_version == 1. (What we are not compatible
381 * with is if a new version FS crashed and then we try to mount an old
382 * version). Otherwise the exofs_fscb is read-only from mkfs time. All
383 * the writeable info is set in exofs_sbi_write_stats() above.
384 */
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700385
Boaz Harrosh5bf696d2011-09-28 11:39:59 +0300386 exofs_init_comps(&oc, &one_comp, sbi, EXOFS_SUPER_ID);
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700387
Boaz Harrosh5bf696d2011-09-28 11:39:59 +0300388 ret = ore_get_io_state(&sbi->layout, &oc, &ios);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200389 if (unlikely(ret))
Boaz Harrosh06886a52009-11-08 14:54:08 +0200390 goto out;
391
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200392 ios->length = offsetof(struct exofs_fscb, s_dev_table_oid);
Boaz Harrosh06886a52009-11-08 14:54:08 +0200393 memset(fscb, 0, ios->length);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200394 fscb->s_nextid = cpu_to_le64(sbi->s_nextid);
Dan Carpenterb6d1f2d2011-11-21 17:43:11 +0300395 fscb->s_numfiles = cpu_to_le64(sbi->s_numfiles);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200396 fscb->s_magic = cpu_to_le16(sb->s_magic);
397 fscb->s_newfs = 0;
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200398 fscb->s_version = EXOFS_FSCB_VER;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200399
Boaz Harrosh06886a52009-11-08 14:54:08 +0200400 ios->offset = 0;
401 ios->kern_buff = fscb;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200402
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700403 ret = ore_write(ios);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200404 if (unlikely(ret))
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700405 EXOFS_ERR("%s: ore_write failed.\n", __func__);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200406
407out:
Boaz Harrosh06886a52009-11-08 14:54:08 +0200408 EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret);
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700409 ore_put_io_state(ios);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200410 kfree(fscb);
Christoph Hellwig80e09fb2009-06-08 10:03:58 +0200411 return ret;
412}
413
Boaz Harrosh19fe2942009-09-03 20:38:02 +0300414static void _exofs_print_device(const char *msg, const char *dev_path,
415 struct osd_dev *od, u64 pid)
416{
417 const struct osd_dev_info *odi = osduld_device_info(od);
418
419 printk(KERN_NOTICE "exofs: %s %s osd_name-%s pid-0x%llx\n",
420 msg, dev_path ?: "", odi->osdname, _LLU(pid));
421}
422
H Hartley Sweetende74b052011-09-23 13:46:51 -0700423static void exofs_free_sbi(struct exofs_sb_info *sbi)
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200424{
Boaz Harroshd866d872011-09-28 14:43:09 +0300425 unsigned numdevs = sbi->oc.numdevs;
426
427 while (numdevs) {
428 unsigned i = --numdevs;
429 struct osd_dev *od = ore_comp_dev(&sbi->oc, i);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200430
431 if (od) {
Boaz Harroshd866d872011-09-28 14:43:09 +0300432 ore_comp_set_dev(&sbi->oc, i, NULL);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200433 osduld_put_device(od);
434 }
435 }
Boaz Harroshd866d872011-09-28 14:43:09 +0300436 kfree(sbi->oc.ods);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200437 kfree(sbi);
438}
439
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200440/*
441 * This function is called when the vfs is freeing the superblock. We just
442 * need to free our own part.
443 */
444static void exofs_put_super(struct super_block *sb)
445{
446 int num_pend;
447 struct exofs_sb_info *sbi = sb->s_fs_info;
448
449 /* make sure there are no pending commands */
450 for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0;
451 num_pend = atomic_read(&sbi->s_curr_pending)) {
452 wait_queue_head_t wq;
Boaz Harrosha49fb4c2011-02-07 18:12:15 +0200453
454 printk(KERN_NOTICE "%s: !!Pending operations in flight. "
455 "This is a BUG. please report to osd-dev@open-osd.org\n",
456 __func__);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200457 init_waitqueue_head(&wq);
458 wait_event_timeout(wq,
459 (atomic_read(&sbi->s_curr_pending) == 0),
460 msecs_to_jiffies(100));
461 }
462
Boaz Harroshd866d872011-09-28 14:43:09 +0300463 _exofs_print_device("Unmounting", NULL, ore_comp_dev(&sbi->oc, 0),
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700464 sbi->one_comp.obj.partition);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200465
Sachin Bhamare8b56a302012-03-14 18:01:45 -0700466 exofs_sysfs_sb_del(sbi);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200467 exofs_free_sbi(sbi);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200468 sb->s_fs_info = NULL;
469}
470
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200471static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs,
472 struct exofs_device_table *dt)
473{
Boaz Harrosh5a51c0c2011-09-28 13:18:45 +0300474 int ret;
Boaz Harrosh5d952b82010-02-01 13:35:51 +0200475
Boaz Harrosh8d2d83a2011-08-10 14:15:02 -0700476 sbi->layout.stripe_unit =
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200477 le64_to_cpu(dt->dt_data_map.cb_stripe_unit);
Boaz Harrosh8d2d83a2011-08-10 14:15:02 -0700478 sbi->layout.group_width =
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200479 le32_to_cpu(dt->dt_data_map.cb_group_width);
Boaz Harrosh8d2d83a2011-08-10 14:15:02 -0700480 sbi->layout.group_depth =
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200481 le32_to_cpu(dt->dt_data_map.cb_group_depth);
Boaz Harrosh8d2d83a2011-08-10 14:15:02 -0700482 sbi->layout.mirrors_p1 =
483 le32_to_cpu(dt->dt_data_map.cb_mirror_cnt) + 1;
484 sbi->layout.raid_algorithm =
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200485 le32_to_cpu(dt->dt_data_map.cb_raid_algorithm);
486
Boaz Harrosh5a51c0c2011-09-28 13:18:45 +0300487 ret = ore_verify_layout(numdevs, &sbi->layout);
Boaz Harrosh5d952b82010-02-01 13:35:51 +0200488
Boaz Harrosh6d4073e2011-07-27 17:51:53 -0700489 EXOFS_DBGMSG("exofs: layout: "
490 "num_comps=%u stripe_unit=0x%x group_width=%u "
491 "group_depth=0x%llx mirrors_p1=%u raid_algorithm=%u\n",
492 numdevs,
493 sbi->layout.stripe_unit,
494 sbi->layout.group_width,
495 _LLU(sbi->layout.group_depth),
496 sbi->layout.mirrors_p1,
Boaz Harrosh8d2d83a2011-08-10 14:15:02 -0700497 sbi->layout.raid_algorithm);
Boaz Harrosh5a51c0c2011-09-28 13:18:45 +0300498 return ret;
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200499}
500
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700501static unsigned __ra_pages(struct ore_layout *layout)
bharrosh@panasas.com66cd6ca2010-10-07 14:28:18 -0400502{
503 const unsigned _MIN_RA = 32; /* min 128K read-ahead */
504 unsigned ra_pages = layout->group_width * layout->stripe_unit /
505 PAGE_SIZE;
506 unsigned max_io_pages = exofs_max_io_pages(layout, ~0);
507
508 ra_pages *= 2; /* two stripes */
509 if (ra_pages < _MIN_RA)
510 ra_pages = roundup(_MIN_RA, ra_pages / 2);
511
512 if (ra_pages > max_io_pages)
513 ra_pages = max_io_pages;
514
515 return ra_pages;
516}
517
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200518/* @odi is valid only as long as @fscb_dev is valid */
519static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev,
520 struct osd_dev_info *odi)
521{
522 odi->systemid_len = le32_to_cpu(dt_dev->systemid_len);
Dan Carpenter72749a22012-01-30 10:59:49 +0300523 if (likely(odi->systemid_len))
524 memcpy(odi->systemid, dt_dev->systemid, OSD_SYSTEMID_LEN);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200525
526 odi->osdname_len = le32_to_cpu(dt_dev->osdname_len);
527 odi->osdname = dt_dev->osdname;
528
529 /* FIXME support long names. Will need a _put function */
530 if (dt_dev->long_name_offset)
531 return -EINVAL;
532
533 /* Make sure osdname is printable!
534 * mkexofs should give us space for a null-terminator else the
535 * device-table is invalid.
536 */
537 if (unlikely(odi->osdname_len >= sizeof(dt_dev->osdname)))
538 odi->osdname_len = sizeof(dt_dev->osdname) - 1;
539 dt_dev->osdname[odi->osdname_len] = 0;
540
541 /* If it's all zeros something is bad we read past end-of-obj */
542 return !(odi->systemid_len || odi->osdname_len);
543}
544
Rashika Kheriaaf8d0cc2014-04-02 17:23:29 +0300545static int __alloc_dev_table(struct exofs_sb_info *sbi, unsigned numdevs,
Boaz Harroshd866d872011-09-28 14:43:09 +0300546 struct exofs_dev **peds)
547{
548 struct __alloc_ore_devs_and_exofs_devs {
549 /* Twice bigger table: See exofs_init_comps() and comment at
550 * exofs_read_lookup_dev_table()
551 */
552 struct ore_dev *oreds[numdevs * 2 - 1];
553 struct exofs_dev eds[numdevs];
554 } *aoded;
555 struct exofs_dev *eds;
556 unsigned i;
557
558 aoded = kzalloc(sizeof(*aoded), GFP_KERNEL);
559 if (unlikely(!aoded)) {
Masanari Iida3e576382012-02-01 23:53:46 +0900560 EXOFS_ERR("ERROR: failed allocating Device array[%d]\n",
Boaz Harroshd866d872011-09-28 14:43:09 +0300561 numdevs);
562 return -ENOMEM;
563 }
564
565 sbi->oc.ods = aoded->oreds;
566 *peds = eds = aoded->eds;
567 for (i = 0; i < numdevs; ++i)
568 aoded->oreds[i] = &eds[i].ored;
569 return 0;
570}
571
Boaz Harrosh6d4073e2011-07-27 17:51:53 -0700572static int exofs_read_lookup_dev_table(struct exofs_sb_info *sbi,
573 struct osd_dev *fscb_od,
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200574 unsigned table_count)
575{
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700576 struct ore_comp comp;
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200577 struct exofs_device_table *dt;
Boaz Harroshd866d872011-09-28 14:43:09 +0300578 struct exofs_dev *eds;
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200579 unsigned table_bytes = table_count * sizeof(dt->dt_dev_table[0]) +
580 sizeof(*dt);
581 unsigned numdevs, i;
582 int ret;
583
584 dt = kmalloc(table_bytes, GFP_KERNEL);
585 if (unlikely(!dt)) {
586 EXOFS_ERR("ERROR: allocating %x bytes for device table\n",
587 table_bytes);
588 return -ENOMEM;
589 }
590
Boaz Harrosh5bf696d2011-09-28 11:39:59 +0300591 sbi->oc.numdevs = 0;
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700592
593 comp.obj.partition = sbi->one_comp.obj.partition;
594 comp.obj.id = EXOFS_DEVTABLE_ID;
595 exofs_make_credential(comp.cred, &comp.obj);
596
597 ret = exofs_read_kern(fscb_od, comp.cred, &comp.obj, 0, dt,
598 table_bytes);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200599 if (unlikely(ret)) {
600 EXOFS_ERR("ERROR: reading device table\n");
601 goto out;
602 }
603
604 numdevs = le64_to_cpu(dt->dt_num_devices);
605 if (unlikely(!numdevs)) {
606 ret = -EINVAL;
607 goto out;
608 }
609 WARN_ON(table_count != numdevs);
610
611 ret = _read_and_match_data_map(sbi, numdevs, dt);
612 if (unlikely(ret))
613 goto out;
614
Boaz Harroshd866d872011-09-28 14:43:09 +0300615 ret = __alloc_dev_table(sbi, numdevs, &eds);
616 if (unlikely(ret))
617 goto out;
618 /* exofs round-robins the device table view according to inode
619 * number. We hold a: twice bigger table hence inodes can point
620 * to any device and have a sequential view of the table
621 * starting at this device. See exofs_init_comps()
622 */
623 memcpy(&sbi->oc.ods[numdevs], &sbi->oc.ods[0],
624 (numdevs - 1) * sizeof(sbi->oc.ods[0]));
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200625
Sachin Bhamare8b56a302012-03-14 18:01:45 -0700626 /* create sysfs subdir under which we put the device table
627 * And cluster layout. A Superblock is identified by the string:
628 * "dev[0].osdname"_"pid"
629 */
630 exofs_sysfs_sb_add(sbi, &dt->dt_dev_table[0]);
631
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200632 for (i = 0; i < numdevs; i++) {
633 struct exofs_fscb fscb;
634 struct osd_dev_info odi;
635 struct osd_dev *od;
636
637 if (exofs_devs_2_odi(&dt->dt_dev_table[i], &odi)) {
638 EXOFS_ERR("ERROR: Read all-zeros device entry\n");
639 ret = -EINVAL;
640 goto out;
641 }
642
643 printk(KERN_NOTICE "Add device[%d]: osd_name-%s\n",
644 i, odi.osdname);
645
Boaz Harroshd866d872011-09-28 14:43:09 +0300646 /* the exofs id is currently the table index */
647 eds[i].did = i;
648
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200649 /* On all devices the device table is identical. The user can
650 * specify any one of the participating devices on the command
651 * line. We always keep them in device-table order.
652 */
653 if (fscb_od && osduld_device_same(fscb_od, &odi)) {
Boaz Harroshd866d872011-09-28 14:43:09 +0300654 eds[i].ored.od = fscb_od;
Boaz Harrosh5bf696d2011-09-28 11:39:59 +0300655 ++sbi->oc.numdevs;
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200656 fscb_od = NULL;
Sachin Bhamare8b56a302012-03-14 18:01:45 -0700657 exofs_sysfs_odev_add(&eds[i], sbi);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200658 continue;
659 }
660
661 od = osduld_info_lookup(&odi);
Tobias Klauser2c722c92010-12-09 15:55:21 +0100662 if (IS_ERR(od)) {
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200663 ret = PTR_ERR(od);
664 EXOFS_ERR("ERROR: device requested is not found "
665 "osd_name-%s =>%d\n", odi.osdname, ret);
666 goto out;
667 }
668
Boaz Harroshd866d872011-09-28 14:43:09 +0300669 eds[i].ored.od = od;
Boaz Harrosh5bf696d2011-09-28 11:39:59 +0300670 ++sbi->oc.numdevs;
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200671
672 /* Read the fscb of the other devices to make sure the FS
673 * partition is there.
674 */
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700675 ret = exofs_read_kern(od, comp.cred, &comp.obj, 0, &fscb,
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200676 sizeof(fscb));
677 if (unlikely(ret)) {
678 EXOFS_ERR("ERROR: Malformed participating device "
679 "error reading fscb osd_name-%s\n",
680 odi.osdname);
681 goto out;
682 }
Sachin Bhamare8b56a302012-03-14 18:01:45 -0700683 exofs_sysfs_odev_add(&eds[i], sbi);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200684
685 /* TODO: verify other information is correct and FS-uuid
686 * matches. Benny what did you say about device table
687 * generation and old devices?
688 */
689 }
690
691out:
692 kfree(dt);
Boaz Harroshd866d872011-09-28 14:43:09 +0300693 if (unlikely(fscb_od && !ret)) {
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700694 EXOFS_ERR("ERROR: Bad device-table container device not present\n");
695 osduld_put_device(fscb_od);
696 return -EINVAL;
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700697 }
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200698 return ret;
699}
700
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200701/*
702 * Read the superblock from the OSD and fill in the fields
703 */
704static int exofs_fill_super(struct super_block *sb, void *data, int silent)
705{
706 struct inode *root;
707 struct exofs_mountopt *opts = data;
708 struct exofs_sb_info *sbi; /*extended info */
Boaz Harrosh06886a52009-11-08 14:54:08 +0200709 struct osd_dev *od; /* Master device */
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200710 struct exofs_fscb fscb; /*on-disk superblock info */
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700711 struct ore_comp comp;
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200712 unsigned table_count;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200713 int ret;
714
715 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
716 if (!sbi)
717 return -ENOMEM;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200718
719 /* use mount options to fill superblock */
Boaz Harrosh9ed96482011-01-31 14:32:14 +0200720 if (opts->is_osdname) {
721 struct osd_dev_info odi = {.systemid_len = 0};
722
723 odi.osdname_len = strlen(opts->dev_name);
724 odi.osdname = (u8 *)opts->dev_name;
725 od = osduld_info_lookup(&odi);
Boaz Harrosh9ce73042011-08-03 20:18:01 -0700726 kfree(opts->dev_name);
727 opts->dev_name = NULL;
Boaz Harrosh9ed96482011-01-31 14:32:14 +0200728 } else {
729 od = osduld_path_lookup(opts->dev_name);
730 }
Boaz Harrosh06886a52009-11-08 14:54:08 +0200731 if (IS_ERR(od)) {
Boaz Harrosh9ed96482011-01-31 14:32:14 +0200732 ret = -EINVAL;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200733 goto free_sbi;
734 }
735
Boaz Harrosh45d3abc2010-01-28 11:46:16 +0200736 /* Default layout in case we do not have a device-table */
Boaz Harrosh5d952b82010-02-01 13:35:51 +0200737 sbi->layout.stripe_unit = PAGE_SIZE;
738 sbi->layout.mirrors_p1 = 1;
739 sbi->layout.group_width = 1;
Boaz Harrosh50a76fd2010-02-11 13:01:39 +0200740 sbi->layout.group_depth = -1;
741 sbi->layout.group_count = 1;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200742 sbi->s_timeout = opts->timeout;
743
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700744 sbi->one_comp.obj.partition = opts->pid;
745 sbi->one_comp.obj.id = 0;
746 exofs_make_credential(sbi->one_comp.cred, &sbi->one_comp.obj);
Boaz Harrosh5bf696d2011-09-28 11:39:59 +0300747 sbi->oc.single_comp = EC_SINGLE_COMP;
748 sbi->oc.comps = &sbi->one_comp;
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700749
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200750 /* fill in some other data by hand */
751 memset(sb->s_id, 0, sizeof(sb->s_id));
752 strcpy(sb->s_id, "exofs");
753 sb->s_blocksize = EXOFS_BLKSIZE;
754 sb->s_blocksize_bits = EXOFS_BLKSHIFT;
755 sb->s_maxbytes = MAX_LFS_FILESIZE;
Al Viro8de52772012-02-06 12:45:27 -0500756 sb->s_max_links = EXOFS_LINK_MAX;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200757 atomic_set(&sbi->s_curr_pending, 0);
758 sb->s_bdev = NULL;
759 sb->s_dev = 0;
760
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700761 comp.obj.partition = sbi->one_comp.obj.partition;
762 comp.obj.id = EXOFS_SUPER_ID;
763 exofs_make_credential(comp.cred, &comp.obj);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200764
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700765 ret = exofs_read_kern(od, comp.cred, &comp.obj, 0, &fscb, sizeof(fscb));
Boaz Harrosh06886a52009-11-08 14:54:08 +0200766 if (unlikely(ret))
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200767 goto free_sbi;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200768
769 sb->s_magic = le16_to_cpu(fscb.s_magic);
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200770 /* NOTE: we read below to be backward compatible with old versions */
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200771 sbi->s_nextid = le64_to_cpu(fscb.s_nextid);
772 sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles);
773
774 /* make sure what we read from the object store is correct */
775 if (sb->s_magic != EXOFS_SUPER_MAGIC) {
776 if (!silent)
777 EXOFS_ERR("ERROR: Bad magic value\n");
778 ret = -EINVAL;
779 goto free_sbi;
780 }
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200781 if (le32_to_cpu(fscb.s_version) > EXOFS_FSCB_VER) {
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200782 EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n",
783 EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version));
784 ret = -EINVAL;
785 goto free_sbi;
786 }
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200787
788 /* start generation numbers from a random point */
789 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
790 spin_lock_init(&sbi->s_next_gen_lock);
791
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200792 table_count = le64_to_cpu(fscb.s_dev_table_count);
793 if (table_count) {
Boaz Harrosh6d4073e2011-07-27 17:51:53 -0700794 ret = exofs_read_lookup_dev_table(sbi, od, table_count);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200795 if (unlikely(ret))
796 goto free_sbi;
Boaz Harrosh6d4073e2011-07-27 17:51:53 -0700797 } else {
Boaz Harroshd866d872011-09-28 14:43:09 +0300798 struct exofs_dev *eds;
799
800 ret = __alloc_dev_table(sbi, 1, &eds);
801 if (unlikely(ret))
802 goto free_sbi;
803
804 ore_comp_set_dev(&sbi->oc, 0, od);
Boaz Harrosh6abe4a82012-05-16 14:22:21 +0300805 sbi->oc.numdevs = 1;
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200806 }
807
Boaz Harrosh1cea3122011-02-03 17:53:25 +0200808 __sbi_read_stats(sbi);
809
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200810 /* set up operation vectors */
Jan Karac7f01472017-04-12 12:24:39 +0200811 ret = super_setup_bdi(sb);
812 if (ret) {
813 EXOFS_DBGMSG("Failed to super_setup_bdi\n");
814 goto free_sbi;
815 }
816 sb->s_bdi->ra_pages = __ra_pages(&sbi->layout);
Boaz Harrosh06886a52009-11-08 14:54:08 +0200817 sb->s_fs_info = sbi;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200818 sb->s_op = &exofs_sops;
Boaz Harrosh8cf74b32009-03-22 12:47:26 +0200819 sb->s_export_op = &exofs_export_ops;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200820 root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF);
821 if (IS_ERR(root)) {
822 EXOFS_ERR("ERROR: exofs_iget failed\n");
823 ret = PTR_ERR(root);
824 goto free_sbi;
825 }
Al Viro48fde702012-01-08 22:15:13 -0500826 sb->s_root = d_make_root(root);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200827 if (!sb->s_root) {
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200828 EXOFS_ERR("ERROR: get root inode failed\n");
829 ret = -ENOMEM;
830 goto free_sbi;
831 }
832
833 if (!S_ISDIR(root->i_mode)) {
834 dput(sb->s_root);
835 sb->s_root = NULL;
836 EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n",
837 root->i_mode);
838 ret = -EINVAL;
839 goto free_sbi;
840 }
841
Sachin Bhamare8b56a302012-03-14 18:01:45 -0700842 exofs_sysfs_dbg_print();
Boaz Harroshd866d872011-09-28 14:43:09 +0300843 _exofs_print_device("Mounting", opts->dev_name,
844 ore_comp_dev(&sbi->oc, 0),
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700845 sbi->one_comp.obj.partition);
Boaz Harrosh06886a52009-11-08 14:54:08 +0200846 return 0;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200847
848free_sbi:
Boaz Harrosh06886a52009-11-08 14:54:08 +0200849 EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n",
Boaz Harrosh9e9db452011-08-05 15:06:04 -0700850 opts->dev_name, sbi->one_comp.obj.partition, ret);
Boaz Harrosh04dc1e82009-11-16 16:03:05 +0200851 exofs_free_sbi(sbi);
Boaz Harrosh06886a52009-11-08 14:54:08 +0200852 return ret;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200853}
854
855/*
856 * Set up the superblock (calls exofs_fill_super eventually)
857 */
Al Viro3c26ff62010-07-25 11:46:36 +0400858static struct dentry *exofs_mount(struct file_system_type *type,
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200859 int flags, const char *dev_name,
Al Viro3c26ff62010-07-25 11:46:36 +0400860 void *data)
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200861{
862 struct exofs_mountopt opts;
863 int ret;
864
865 ret = parse_options(data, &opts);
866 if (ret)
Al Viro3c26ff62010-07-25 11:46:36 +0400867 return ERR_PTR(ret);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200868
Boaz Harrosh9ed96482011-01-31 14:32:14 +0200869 if (!opts.dev_name)
870 opts.dev_name = dev_name;
Al Viro3c26ff62010-07-25 11:46:36 +0400871 return mount_nodev(type, flags, &opts, exofs_fill_super);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200872}
873
874/*
875 * Return information about the file system state in the buffer. This is used
876 * by the 'df' command, for example.
877 */
878static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf)
879{
880 struct super_block *sb = dentry->d_sb;
881 struct exofs_sb_info *sbi = sb->s_fs_info;
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700882 struct ore_io_state *ios;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200883 struct osd_attr attrs[] = {
884 ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS,
885 OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)),
886 ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION,
887 OSD_ATTR_PI_USED_CAPACITY, sizeof(__be64)),
888 };
889 uint64_t capacity = ULLONG_MAX;
890 uint64_t used = ULLONG_MAX;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200891 int ret;
892
Boaz Harrosh5bf696d2011-09-28 11:39:59 +0300893 ret = ore_get_io_state(&sbi->layout, &sbi->oc, &ios);
Boaz Harrosh06886a52009-11-08 14:54:08 +0200894 if (ret) {
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700895 EXOFS_DBGMSG("ore_get_io_state failed.\n");
Boaz Harrosh06886a52009-11-08 14:54:08 +0200896 return ret;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200897 }
898
Boaz Harrosh06886a52009-11-08 14:54:08 +0200899 ios->in_attr = attrs;
900 ios->in_attr_len = ARRAY_SIZE(attrs);
901
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700902 ret = ore_read(ios);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200903 if (unlikely(ret))
904 goto out;
905
Boaz Harrosh06886a52009-11-08 14:54:08 +0200906 ret = extract_attr_from_ios(ios, &attrs[0]);
Boaz Harroshcae012d2009-11-02 18:19:24 +0200907 if (likely(!ret)) {
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200908 capacity = get_unaligned_be64(attrs[0].val_ptr);
Boaz Harroshcae012d2009-11-02 18:19:24 +0200909 if (unlikely(!capacity))
910 capacity = ULLONG_MAX;
911 } else
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200912 EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
913
Boaz Harrosh06886a52009-11-08 14:54:08 +0200914 ret = extract_attr_from_ios(ios, &attrs[1]);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200915 if (likely(!ret))
916 used = get_unaligned_be64(attrs[1].val_ptr);
917 else
918 EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n");
919
920 /* fill in the stats buffer */
921 buf->f_type = EXOFS_SUPER_MAGIC;
922 buf->f_bsize = EXOFS_BLKSIZE;
Boaz Harroshcae012d2009-11-02 18:19:24 +0200923 buf->f_blocks = capacity >> 9;
924 buf->f_bfree = (capacity - used) >> 9;
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200925 buf->f_bavail = buf->f_bfree;
926 buf->f_files = sbi->s_numfiles;
927 buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles;
928 buf->f_namelen = EXOFS_NAME_LEN;
929
930out:
Boaz Harrosh8ff660a2011-08-06 19:26:31 -0700931 ore_put_io_state(ios);
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200932 return ret;
933}
934
935static const struct super_operations exofs_sops = {
936 .alloc_inode = exofs_alloc_inode,
937 .destroy_inode = exofs_destroy_inode,
938 .write_inode = exofs_write_inode,
Al Viro4ec70c92010-06-07 11:42:26 -0400939 .evict_inode = exofs_evict_inode,
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200940 .put_super = exofs_put_super,
Christoph Hellwig80e09fb2009-06-08 10:03:58 +0200941 .sync_fs = exofs_sync_fs,
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200942 .statfs = exofs_statfs,
943};
944
945/******************************************************************************
Boaz Harrosh8cf74b32009-03-22 12:47:26 +0200946 * EXPORT OPERATIONS
947 *****************************************************************************/
948
H Hartley Sweetende74b052011-09-23 13:46:51 -0700949static struct dentry *exofs_get_parent(struct dentry *child)
Boaz Harrosh8cf74b32009-03-22 12:47:26 +0200950{
951 unsigned long ino = exofs_parent_ino(child);
952
953 if (!ino)
Al Viroa803b802011-07-08 20:56:55 -0400954 return ERR_PTR(-ESTALE);
Boaz Harrosh8cf74b32009-03-22 12:47:26 +0200955
Al Virofc640052016-04-10 01:33:30 -0400956 return d_obtain_alias(exofs_iget(child->d_sb, ino));
Boaz Harrosh8cf74b32009-03-22 12:47:26 +0200957}
958
959static struct inode *exofs_nfs_get_inode(struct super_block *sb,
960 u64 ino, u32 generation)
961{
962 struct inode *inode;
963
964 inode = exofs_iget(sb, ino);
965 if (IS_ERR(inode))
966 return ERR_CAST(inode);
967 if (generation && inode->i_generation != generation) {
968 /* we didn't find the right inode.. */
969 iput(inode);
970 return ERR_PTR(-ESTALE);
971 }
972 return inode;
973}
974
975static struct dentry *exofs_fh_to_dentry(struct super_block *sb,
976 struct fid *fid, int fh_len, int fh_type)
977{
978 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
979 exofs_nfs_get_inode);
980}
981
982static struct dentry *exofs_fh_to_parent(struct super_block *sb,
983 struct fid *fid, int fh_len, int fh_type)
984{
985 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
986 exofs_nfs_get_inode);
987}
988
989static const struct export_operations exofs_export_ops = {
990 .fh_to_dentry = exofs_fh_to_dentry,
991 .fh_to_parent = exofs_fh_to_parent,
992 .get_parent = exofs_get_parent,
993};
994
995/******************************************************************************
Boaz Harroshba9e5e92008-10-28 16:11:41 +0200996 * INSMOD/RMMOD
997 *****************************************************************************/
998
999/*
1000 * struct that describes this file system
1001 */
1002static struct file_system_type exofs_type = {
1003 .owner = THIS_MODULE,
1004 .name = "exofs",
Al Viro3c26ff62010-07-25 11:46:36 +04001005 .mount = exofs_mount,
Boaz Harroshba9e5e92008-10-28 16:11:41 +02001006 .kill_sb = generic_shutdown_super,
1007};
Eric W. Biederman7f78e032013-03-02 19:39:14 -08001008MODULE_ALIAS_FS("exofs");
Boaz Harroshba9e5e92008-10-28 16:11:41 +02001009
1010static int __init init_exofs(void)
1011{
1012 int err;
1013
1014 err = init_inodecache();
1015 if (err)
1016 goto out;
1017
1018 err = register_filesystem(&exofs_type);
1019 if (err)
1020 goto out_d;
1021
Sachin Bhamare8b56a302012-03-14 18:01:45 -07001022 /* We don't fail if sysfs creation failed */
1023 exofs_sysfs_init();
1024
Boaz Harroshba9e5e92008-10-28 16:11:41 +02001025 return 0;
1026out_d:
1027 destroy_inodecache();
1028out:
1029 return err;
1030}
1031
1032static void __exit exit_exofs(void)
1033{
Sachin Bhamare8b56a302012-03-14 18:01:45 -07001034 exofs_sysfs_uninit();
Boaz Harroshba9e5e92008-10-28 16:11:41 +02001035 unregister_filesystem(&exofs_type);
1036 destroy_inodecache();
1037}
1038
1039MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>");
1040MODULE_DESCRIPTION("exofs");
1041MODULE_LICENSE("GPL");
1042
1043module_init(init_exofs)
1044module_exit(exit_exofs)