Whamcloud - gitweb
LU-4563 Fix unsafe userspace access in many proc files
[fs/lustre-release.git] / lustre / llite / lproc_llite.c
1 /*
2  * GPL HEADER START
3  *
4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
5  *
6  * This program is free software; you can redistribute it and/or modify
7  * it under the terms of the GNU General Public License version 2 only,
8  * as published by the Free Software Foundation.
9  *
10  * This program is distributed in the hope that it will be useful, but
11  * WITHOUT ANY WARRANTY; without even the implied warranty of
12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
13  * General Public License version 2 for more details (a copy is included
14  * in the LICENSE file that accompanied this code).
15  *
16  * You should have received a copy of the GNU General Public License
17  * version 2 along with this program; If not, see
18  * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
19  *
20  * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
21  * CA 95054 USA or visit www.sun.com if you need additional information or
22  * have any questions.
23  *
24  * GPL HEADER END
25  */
26 /*
27  * Copyright (c) 2002, 2010, Oracle and/or its affiliates. All rights reserved.
28  * Use is subject to license terms.
29  *
30  * Copyright (c) 2012, 2013, Intel Corporation.
31  */
32 /*
33  * This file is part of Lustre, http://www.lustre.org/
34  * Lustre is a trademark of Sun Microsystems, Inc.
35  */
36 #define DEBUG_SUBSYSTEM S_LLITE
37
38 #include <linux/version.h>
39 #include <lustre_lite.h>
40 #include <lustre_param.h>
41 #include <lprocfs_status.h>
42 #include <obd_support.h>
43
44 #include "llite_internal.h"
45 #include "vvp_internal.h"
46
47 struct proc_dir_entry *proc_lustre_fs_root;
48
49 #ifdef LPROCFS
50 /* /proc/lustre/llite mount point registration */
51 struct file_operations ll_rw_extents_stats_fops;
52 struct file_operations ll_rw_extents_stats_pp_fops;
53 struct file_operations ll_rw_offset_stats_fops;
54
55 static int ll_blksize_seq_show(struct seq_file *m, void *v)
56 {
57         struct super_block *sb = m->private;
58         struct obd_statfs osfs;
59         int rc;
60
61         LASSERT(sb != NULL);
62         rc = ll_statfs_internal(sb, &osfs,
63                                 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
64                                 OBD_STATFS_NODELAY);
65         if (!rc)
66                 rc = seq_printf(m, "%u\n", osfs.os_bsize);
67         return rc;
68 }
69 LPROC_SEQ_FOPS_RO(ll_blksize);
70
71 static int ll_kbytestotal_seq_show(struct seq_file *m, void *v)
72 {
73         struct super_block *sb = m->private;
74         struct obd_statfs osfs;
75         int rc;
76
77         LASSERT(sb != NULL);
78         rc = ll_statfs_internal(sb, &osfs,
79                                 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
80                                 OBD_STATFS_NODELAY);
81         if (!rc) {
82                 __u32 blk_size = osfs.os_bsize >> 10;
83                 __u64 result = osfs.os_blocks;
84
85                 while (blk_size >>= 1)
86                         result <<= 1;
87
88                 rc = seq_printf(m, LPU64"\n", result);
89         }
90         return rc;
91 }
92 LPROC_SEQ_FOPS_RO(ll_kbytestotal);
93
94 static int ll_kbytesfree_seq_show(struct seq_file *m, void *v)
95 {
96         struct super_block *sb = m->private;
97         struct obd_statfs osfs;
98         int rc;
99
100         LASSERT(sb != NULL);
101         rc = ll_statfs_internal(sb, &osfs,
102                                 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
103                                 OBD_STATFS_NODELAY);
104         if (!rc) {
105                 __u32 blk_size = osfs.os_bsize >> 10;
106                 __u64 result = osfs.os_bfree;
107
108                 while (blk_size >>= 1)
109                         result <<= 1;
110
111                 rc = seq_printf(m, LPU64"\n", result);
112         }
113         return rc;
114 }
115 LPROC_SEQ_FOPS_RO(ll_kbytesfree);
116
117 static int ll_kbytesavail_seq_show(struct seq_file *m, void *v)
118 {
119         struct super_block *sb = m->private;
120         struct obd_statfs osfs;
121         int rc;
122
123         LASSERT(sb != NULL);
124         rc = ll_statfs_internal(sb, &osfs,
125                                 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
126                                 OBD_STATFS_NODELAY);
127         if (!rc) {
128                 __u32 blk_size = osfs.os_bsize >> 10;
129                 __u64 result = osfs.os_bavail;
130
131                 while (blk_size >>= 1)
132                         result <<= 1;
133
134                 rc = seq_printf(m, LPU64"\n", result);
135         }
136         return rc;
137 }
138 LPROC_SEQ_FOPS_RO(ll_kbytesavail);
139
140 static int ll_filestotal_seq_show(struct seq_file *m, void *v)
141 {
142         struct super_block *sb = m->private;
143         struct obd_statfs osfs;
144         int rc;
145
146         LASSERT(sb != NULL);
147         rc = ll_statfs_internal(sb, &osfs,
148                                 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
149                                 OBD_STATFS_NODELAY);
150         if (!rc)
151                 rc = seq_printf(m, LPU64"\n", osfs.os_files);
152         return rc;
153 }
154 LPROC_SEQ_FOPS_RO(ll_filestotal);
155
156 static int ll_filesfree_seq_show(struct seq_file *m, void *v)
157 {
158         struct super_block *sb = m->private;
159         struct obd_statfs osfs;
160         int rc;
161
162         LASSERT(sb != NULL);
163         rc = ll_statfs_internal(sb, &osfs,
164                                 cfs_time_shift_64(-OBD_STATFS_CACHE_SECONDS),
165                                 OBD_STATFS_NODELAY);
166         if (!rc)
167                 rc = seq_printf(m, LPU64"\n", osfs.os_ffree);
168         return rc;
169 }
170 LPROC_SEQ_FOPS_RO(ll_filesfree);
171
172 static int ll_client_type_seq_show(struct seq_file *m, void *v)
173 {
174         struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)m->private);
175         int rc;
176
177         LASSERT(sbi != NULL);
178
179         if (sbi->ll_flags & LL_SBI_RMT_CLIENT)
180                 rc = seq_printf(m, "remote client\n");
181         else
182                 rc = seq_printf(m, "local client\n");
183         return rc;
184 }
185 LPROC_SEQ_FOPS_RO(ll_client_type);
186
187 static int ll_fstype_seq_show(struct seq_file *m, void *v)
188 {
189         struct super_block *sb = m->private;
190
191         LASSERT(sb != NULL);
192         return seq_printf(m, "%s\n", sb->s_type->name);
193 }
194 LPROC_SEQ_FOPS_RO(ll_fstype);
195
196 static int ll_sb_uuid_seq_show(struct seq_file *m, void *v)
197 {
198         struct super_block *sb = m->private;
199
200         LASSERT(sb != NULL);
201         return seq_printf(m, "%s\n", ll_s2sbi(sb)->ll_sb_uuid.uuid);
202 }
203 LPROC_SEQ_FOPS_RO(ll_sb_uuid);
204
205 static int ll_xattr_cache_seq_show(struct seq_file *m, void *v)
206 {
207         struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)m->private);
208
209         return seq_printf(m, "%u\n", sbi->ll_xattr_cache_enabled);
210 }
211
212 static ssize_t ll_xattr_cache_seq_write(struct file *file, const char *buffer,
213                                         size_t count, loff_t *off)
214 {
215         struct seq_file *m = file->private_data;
216         struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)m->private);
217         int val, rc;
218
219         rc = lprocfs_write_helper(buffer, count, &val);
220         if (rc)
221                 return rc;
222
223         if (val != 0 && val != 1)
224                 return -ERANGE;
225
226         if (val == 1 && !(sbi->ll_flags & LL_SBI_XATTR_CACHE))
227                 return -ENOTSUPP;
228
229         sbi->ll_xattr_cache_enabled = val;
230
231         return count;
232 }
233 LPROC_SEQ_FOPS(ll_xattr_cache);
234
235 static int ll_site_stats_seq_show(struct seq_file *m, void *v)
236 {
237         struct super_block *sb = m->private;
238
239         /*
240          * See description of statistical counters in struct cl_site, and
241          * struct lu_site.
242          */
243         return cl_site_stats_print(lu2cl_site(ll_s2sbi(sb)->ll_site), m);
244 }
245 LPROC_SEQ_FOPS_RO(ll_site_stats);
246
247 static int ll_max_readahead_mb_seq_show(struct seq_file *m, void *v)
248 {
249         struct super_block *sb = m->private;
250         struct ll_sb_info *sbi = ll_s2sbi(sb);
251         long pages_number;
252         int mult;
253
254         spin_lock(&sbi->ll_lock);
255         pages_number = sbi->ll_ra_info.ra_max_pages;
256         spin_unlock(&sbi->ll_lock);
257
258         mult = 1 << (20 - PAGE_CACHE_SHIFT);
259         return lprocfs_seq_read_frac_helper(m, pages_number, mult);
260 }
261
262 static ssize_t
263 ll_max_readahead_mb_seq_write(struct file *file, const char *buffer,
264                               size_t count, loff_t *off)
265 {
266         struct seq_file *m = file->private_data;
267         struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)m->private);
268         int mult, rc, pages_number;
269
270         mult = 1 << (20 - PAGE_CACHE_SHIFT);
271         rc = lprocfs_write_frac_helper(buffer, count, &pages_number, mult);
272         if (rc)
273                 return rc;
274
275         if (pages_number < 0 || pages_number > totalram_pages / 2) {
276                 /* 1/2 of RAM */
277                 CERROR("can't set file readahead more than %lu MB\n",
278                        totalram_pages >> (20 - PAGE_CACHE_SHIFT + 1));
279                 return -ERANGE;
280         }
281
282         spin_lock(&sbi->ll_lock);
283         sbi->ll_ra_info.ra_max_pages = pages_number;
284         spin_unlock(&sbi->ll_lock);
285         return count;
286 }
287 LPROC_SEQ_FOPS(ll_max_readahead_mb);
288
289 static int ll_max_readahead_per_file_mb_seq_show(struct seq_file *m, void *v)
290 {
291         struct super_block *sb = m->private;
292         struct ll_sb_info *sbi = ll_s2sbi(sb);
293         long pages_number;
294         int mult;
295
296         spin_lock(&sbi->ll_lock);
297         pages_number = sbi->ll_ra_info.ra_max_pages_per_file;
298         spin_unlock(&sbi->ll_lock);
299
300         mult = 1 << (20 - PAGE_CACHE_SHIFT);
301         return lprocfs_seq_read_frac_helper(m, pages_number, mult);
302 }
303
304 static ssize_t
305 ll_max_readahead_per_file_mb_seq_write(struct file *file, const char *buffer,
306                                        size_t count, loff_t *off)
307 {
308         struct seq_file *m = file->private_data;
309         struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)m->private);
310         int mult, rc, pages_number;
311
312         mult = 1 << (20 - PAGE_CACHE_SHIFT);
313         rc = lprocfs_write_frac_helper(buffer, count, &pages_number, mult);
314         if (rc)
315                 return rc;
316
317         if (pages_number < 0 ||
318             pages_number > sbi->ll_ra_info.ra_max_pages) {
319                 CERROR("can't set file readahead more than"
320                        "max_read_ahead_mb %lu MB\n",
321                        sbi->ll_ra_info.ra_max_pages);
322                 return -ERANGE;
323         }
324
325         spin_lock(&sbi->ll_lock);
326         sbi->ll_ra_info.ra_max_pages_per_file = pages_number;
327         spin_unlock(&sbi->ll_lock);
328         return count;
329 }
330 LPROC_SEQ_FOPS(ll_max_readahead_per_file_mb);
331
332 static int ll_max_read_ahead_whole_mb_seq_show(struct seq_file *m, void *v)
333 {
334         struct super_block *sb = m->private;
335         struct ll_sb_info *sbi = ll_s2sbi(sb);
336         long pages_number;
337         int mult;
338
339         spin_lock(&sbi->ll_lock);
340         pages_number = sbi->ll_ra_info.ra_max_read_ahead_whole_pages;
341         spin_unlock(&sbi->ll_lock);
342
343         mult = 1 << (20 - PAGE_CACHE_SHIFT);
344         return lprocfs_seq_read_frac_helper(m, pages_number, mult);
345 }
346
347 static ssize_t
348 ll_max_read_ahead_whole_mb_seq_write(struct file *file, const char *buffer,
349                                      size_t count, loff_t *off)
350 {
351         struct seq_file *m = file->private_data;
352         struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)m->private);
353         int mult, rc, pages_number;
354
355         mult = 1 << (20 - PAGE_CACHE_SHIFT);
356         rc = lprocfs_write_frac_helper(buffer, count, &pages_number, mult);
357         if (rc)
358                 return rc;
359
360         /* Cap this at the current max readahead window size, the readahead
361          * algorithm does this anyway so it's pointless to set it larger. */
362         if (pages_number < 0 ||
363             pages_number > sbi->ll_ra_info.ra_max_pages_per_file) {
364                 CERROR("can't set max_read_ahead_whole_mb more than "
365                        "max_read_ahead_per_file_mb: %lu\n",
366                         sbi->ll_ra_info.ra_max_pages_per_file >>
367                         (20 - PAGE_CACHE_SHIFT));
368                 return -ERANGE;
369         }
370
371         spin_lock(&sbi->ll_lock);
372         sbi->ll_ra_info.ra_max_read_ahead_whole_pages = pages_number;
373         spin_unlock(&sbi->ll_lock);
374         return count;
375 }
376 LPROC_SEQ_FOPS(ll_max_read_ahead_whole_mb);
377
378 static int ll_max_cached_mb_seq_show(struct seq_file *m, void *v)
379 {
380         struct super_block     *sb    = m->private;
381         struct ll_sb_info      *sbi   = ll_s2sbi(sb);
382         struct cl_client_cache *cache = &sbi->ll_cache;
383         int shift = 20 - PAGE_CACHE_SHIFT;
384         int max_cached_mb;
385         int unused_mb;
386
387         max_cached_mb = cache->ccc_lru_max >> shift;
388         unused_mb = atomic_read(&cache->ccc_lru_left) >> shift;
389         return seq_printf(m,
390                         "users: %d\n"
391                         "max_cached_mb: %d\n"
392                         "used_mb: %d\n"
393                         "unused_mb: %d\n"
394                         "reclaim_count: %u\n",
395                         atomic_read(&cache->ccc_users),
396                         max_cached_mb,
397                         max_cached_mb - unused_mb,
398                         unused_mb,
399                         cache->ccc_lru_shrinkers);
400 }
401
402 static ssize_t
403 ll_max_cached_mb_seq_write(struct file *file, const char __user *buffer,
404                            size_t count, loff_t *off)
405 {
406         struct seq_file *m = file->private_data;
407         struct super_block *sb = m->private;
408         struct ll_sb_info *sbi = ll_s2sbi(sb);
409         struct cl_client_cache *cache = &sbi->ll_cache;
410         struct lu_env *env;
411         int refcheck;
412         int mult, rc, pages_number;
413         int diff = 0;
414         int nrpages = 0;
415         char kernbuf[128];
416         ENTRY;
417
418         if (count >= sizeof(kernbuf))
419                 RETURN(-EINVAL);
420
421         if (copy_from_user(kernbuf, buffer, count))
422                 RETURN(-EFAULT);
423         kernbuf[count] = 0;
424
425         mult = 1 << (20 - PAGE_CACHE_SHIFT);
426         buffer += lprocfs_find_named_value(kernbuf, "max_cached_mb:", &count) -
427                   kernbuf;
428         rc = lprocfs_write_frac_helper(buffer, count, &pages_number, mult);
429         if (rc)
430                 RETURN(rc);
431
432         if (pages_number < 0 || pages_number > totalram_pages) {
433                 CERROR("%s: can't set max cache more than %lu MB\n",
434                        ll_get_fsname(sb, NULL, 0),
435                        totalram_pages >> (20 - PAGE_CACHE_SHIFT));
436                 RETURN(-ERANGE);
437         }
438
439         if (sbi->ll_dt_exp == NULL) /* being initialized */
440                 GOTO(out, rc = 0);
441
442         spin_lock(&sbi->ll_lock);
443         diff = pages_number - cache->ccc_lru_max;
444         spin_unlock(&sbi->ll_lock);
445
446         /* easy - add more LRU slots. */
447         if (diff >= 0) {
448                 atomic_add(diff, &cache->ccc_lru_left);
449                 GOTO(out, rc = 0);
450         }
451
452         env = cl_env_get(&refcheck);
453         if (IS_ERR(env))
454                 RETURN(rc);
455
456         diff = -diff;
457         while (diff > 0) {
458                 int tmp;
459
460                 /* reduce LRU budget from free slots. */
461                 do {
462                         int ov, nv;
463
464                         ov = atomic_read(&cache->ccc_lru_left);
465                         if (ov == 0)
466                                 break;
467
468                         nv = ov > diff ? ov - diff : 0;
469                         rc = atomic_cmpxchg(&cache->ccc_lru_left, ov, nv);
470                         if (likely(ov == rc)) {
471                                 diff -= ov - nv;
472                                 nrpages += ov - nv;
473                                 break;
474                         }
475                 } while (1);
476
477                 if (diff <= 0)
478                         break;
479
480                 /* difficult - have to ask OSCs to drop LRU slots. */
481                 tmp = diff << 1;
482                 rc = obd_set_info_async(env, sbi->ll_dt_exp,
483                                 sizeof(KEY_CACHE_LRU_SHRINK),
484                                 KEY_CACHE_LRU_SHRINK,
485                                 sizeof(tmp), &tmp, NULL);
486                 if (rc < 0)
487                         break;
488         }
489         cl_env_put(env, &refcheck);
490
491 out:
492         if (rc >= 0) {
493                 spin_lock(&sbi->ll_lock);
494                 cache->ccc_lru_max = pages_number;
495                 spin_unlock(&sbi->ll_lock);
496                 rc = count;
497         } else {
498                 atomic_add(nrpages, &cache->ccc_lru_left);
499         }
500         return rc;
501 }
502 LPROC_SEQ_FOPS(ll_max_cached_mb);
503
504 static int ll_checksum_seq_show(struct seq_file *m, void *v)
505 {
506         struct super_block *sb = m->private;
507         struct ll_sb_info *sbi = ll_s2sbi(sb);
508
509         return seq_printf(m, "%u\n", (sbi->ll_flags & LL_SBI_CHECKSUM) ? 1 : 0);
510 }
511
512 static ssize_t ll_checksum_seq_write(struct file *file, const char *buffer,
513                                      size_t count, loff_t *off)
514 {
515         struct seq_file *m = file->private_data;
516         struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)m->private);
517         int val, rc;
518
519         if (!sbi->ll_dt_exp)
520                 /* Not set up yet */
521                 return -EAGAIN;
522
523         rc = lprocfs_write_helper(buffer, count, &val);
524         if (rc)
525                 return rc;
526         if (val)
527                 sbi->ll_flags |= LL_SBI_CHECKSUM;
528         else
529                 sbi->ll_flags &= ~LL_SBI_CHECKSUM;
530
531         rc = obd_set_info_async(NULL, sbi->ll_dt_exp, sizeof(KEY_CHECKSUM),
532                                 KEY_CHECKSUM, sizeof(val), &val, NULL);
533         if (rc)
534                 CWARN("Failed to set OSC checksum flags: %d\n", rc);
535
536         return count;
537 }
538 LPROC_SEQ_FOPS(ll_checksum);
539
540 static int ll_max_rw_chunk_seq_show(struct seq_file *m, void *v)
541 {
542         struct super_block *sb = m->private;
543
544         return seq_printf(m, "%lu\n", ll_s2sbi(sb)->ll_max_rw_chunk);
545 }
546
547 static ssize_t ll_max_rw_chunk_seq_write(struct file *file, const char *buffer,
548                                          size_t count, loff_t *off)
549 {
550         struct seq_file *m = file->private_data;
551         struct super_block *sb = m->private;
552         int rc, val;
553
554         rc = lprocfs_write_helper(buffer, count, &val);
555         if (rc)
556                 return rc;
557         ll_s2sbi(sb)->ll_max_rw_chunk = val;
558         return count;
559 }
560 LPROC_SEQ_FOPS(ll_max_rw_chunk);
561
562 static int ll_rd_track_id(struct seq_file *m, enum stats_track_type type)
563 {
564         struct super_block *sb = m->private;
565
566         if (ll_s2sbi(sb)->ll_stats_track_type == type) {
567                 return seq_printf(m, "%d\n",
568                                   ll_s2sbi(sb)->ll_stats_track_id);
569         } else if (ll_s2sbi(sb)->ll_stats_track_type == STATS_TRACK_ALL) {
570                 return seq_printf(m, "0 (all)\n");
571         } else {
572                 return seq_printf(m, "untracked\n");
573         }
574 }
575
576 static int ll_wr_track_id(const char *buffer, unsigned long count, void *data,
577                           enum stats_track_type type)
578 {
579         struct super_block *sb = data;
580         int rc, pid;
581
582         rc = lprocfs_write_helper(buffer, count, &pid);
583         if (rc)
584                 return rc;
585         ll_s2sbi(sb)->ll_stats_track_id = pid;
586         if (pid == 0)
587                 ll_s2sbi(sb)->ll_stats_track_type = STATS_TRACK_ALL;
588         else
589                 ll_s2sbi(sb)->ll_stats_track_type = type;
590         lprocfs_clear_stats(ll_s2sbi(sb)->ll_stats);
591         return count;
592 }
593
594 static int ll_track_pid_seq_show(struct seq_file *m, void *v)
595 {
596         return ll_rd_track_id(m, STATS_TRACK_PID);
597 }
598
599 static ssize_t ll_track_pid_seq_write(struct file *file, const char *buffer,
600                                       size_t count, loff_t *off)
601 {
602         struct seq_file *seq = file->private_data;
603         return ll_wr_track_id(buffer, count, seq->private, STATS_TRACK_PID);
604 }
605 LPROC_SEQ_FOPS(ll_track_pid);
606
607 static int ll_track_ppid_seq_show(struct seq_file *m, void *v)
608 {
609         return ll_rd_track_id(m, STATS_TRACK_PPID);
610 }
611
612 static ssize_t ll_track_ppid_seq_write(struct file *file, const char *buffer,
613                                        size_t count, loff_t *off)
614 {
615         struct seq_file *seq = file->private_data;
616         return ll_wr_track_id(buffer, count, seq->private, STATS_TRACK_PPID);
617 }
618 LPROC_SEQ_FOPS(ll_track_ppid);
619
620 static int ll_track_gid_seq_show(struct seq_file *m, void *v)
621 {
622         return ll_rd_track_id(m, STATS_TRACK_GID);
623 }
624
625 static ssize_t ll_track_gid_seq_write(struct file *file, const char *buffer,
626                                       size_t count, loff_t *off)
627 {
628         struct seq_file *seq = file->private_data;
629         return ll_wr_track_id(buffer, count, seq->private, STATS_TRACK_GID);
630 }
631 LPROC_SEQ_FOPS(ll_track_gid);
632
633 static int ll_statahead_max_seq_show(struct seq_file *m, void *v)
634 {
635         struct super_block *sb = m->private;
636         struct ll_sb_info *sbi = ll_s2sbi(sb);
637
638         return seq_printf(m, "%u\n", sbi->ll_sa_max);
639 }
640
641 static ssize_t ll_statahead_max_seq_write(struct file *file, const char *buffer,
642                                           size_t count, loff_t *off)
643 {
644         struct seq_file *m = file->private_data;
645         struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)m->private);
646         int val, rc;
647
648         rc = lprocfs_write_helper(buffer, count, &val);
649         if (rc)
650                 return rc;
651
652         if (val >= 0 && val <= LL_SA_RPC_MAX)
653                 sbi->ll_sa_max = val;
654         else
655                 CERROR("Bad statahead_max value %d. Valid values are in the "
656                        "range [0, %d]\n", val, LL_SA_RPC_MAX);
657
658         return count;
659 }
660 LPROC_SEQ_FOPS(ll_statahead_max);
661
662 static int ll_statahead_agl_seq_show(struct seq_file *m, void *v)
663 {
664         struct super_block *sb = m->private;
665         struct ll_sb_info *sbi = ll_s2sbi(sb);
666
667         return seq_printf(m, "%u\n",
668                           sbi->ll_flags & LL_SBI_AGL_ENABLED ? 1 : 0);
669 }
670
671 static ssize_t ll_statahead_agl_seq_write(struct file *file, const char *buffer,
672                                           size_t count, loff_t *off)
673 {
674         struct seq_file *m = file->private_data;
675         struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)m->private);
676         int val, rc;
677
678         rc = lprocfs_write_helper(buffer, count, &val);
679         if (rc)
680                 return rc;
681
682         if (val)
683                 sbi->ll_flags |= LL_SBI_AGL_ENABLED;
684         else
685                 sbi->ll_flags &= ~LL_SBI_AGL_ENABLED;
686
687         return count;
688 }
689 LPROC_SEQ_FOPS(ll_statahead_agl);
690
691 static int ll_statahead_stats_seq_show(struct seq_file *m, void *v)
692 {
693         struct super_block *sb = m->private;
694         struct ll_sb_info *sbi = ll_s2sbi(sb);
695
696         return seq_printf(m,
697                         "statahead total: %u\n"
698                         "statahead wrong: %u\n"
699                         "agl total: %u\n",
700                         atomic_read(&sbi->ll_sa_total),
701                         atomic_read(&sbi->ll_sa_wrong),
702                         atomic_read(&sbi->ll_agl_total));
703 }
704 LPROC_SEQ_FOPS_RO(ll_statahead_stats);
705
706 static int ll_lazystatfs_seq_show(struct seq_file *m, void *v)
707 {
708         struct super_block *sb = m->private;
709         struct ll_sb_info *sbi = ll_s2sbi(sb);
710
711         return seq_printf(m, "%u\n",
712                           (sbi->ll_flags & LL_SBI_LAZYSTATFS) ? 1 : 0);
713 }
714
715 static ssize_t ll_lazystatfs_seq_write(struct file *file, const char *buffer,
716                                         size_t count, loff_t *off)
717 {
718         struct seq_file *m = file->private_data;
719         struct ll_sb_info *sbi = ll_s2sbi((struct super_block *)m->private);
720         int val, rc;
721
722         rc = lprocfs_write_helper(buffer, count, &val);
723         if (rc)
724                 return rc;
725
726         if (val)
727                 sbi->ll_flags |= LL_SBI_LAZYSTATFS;
728         else
729                 sbi->ll_flags &= ~LL_SBI_LAZYSTATFS;
730
731         return count;
732 }
733 LPROC_SEQ_FOPS(ll_lazystatfs);
734
735 static int ll_max_easize_seq_show(struct seq_file *m, void *v)
736 {
737         struct super_block *sb = m->private;
738         struct ll_sb_info *sbi = ll_s2sbi(sb);
739         unsigned int ealen;
740         int rc;
741
742         rc = ll_get_max_mdsize(sbi, &ealen);
743         if (rc)
744                 return rc;
745
746         return seq_printf(m, "%u\n", ealen);
747 }
748 LPROC_SEQ_FOPS_RO(ll_max_easize);
749
750 static int ll_defult_easize_seq_show(struct seq_file *m, void *v)
751 {
752         struct super_block *sb = m->private;
753         struct ll_sb_info *sbi = ll_s2sbi(sb);
754         unsigned int ealen;
755         int rc;
756
757         rc = ll_get_default_mdsize(sbi, &ealen);
758         if (rc)
759                 return rc;
760
761         return seq_printf(m, "%u\n", ealen);
762 }
763 LPROC_SEQ_FOPS_RO(ll_defult_easize);
764
765 static int ll_max_cookiesize_seq_show(struct seq_file *m, void *v)
766 {
767         struct super_block *sb = m->private;
768         struct ll_sb_info *sbi = ll_s2sbi(sb);
769         unsigned int cookielen;
770         int rc;
771
772         rc = ll_get_max_cookiesize(sbi, &cookielen);
773         if (rc)
774                 return rc;
775
776         return seq_printf(m, "%u\n", cookielen);
777 }
778 LPROC_SEQ_FOPS_RO(ll_max_cookiesize);
779
780 static int ll_defult_cookiesize_seq_show(struct seq_file *m, void *v)
781 {
782         struct super_block *sb = m->private;
783         struct ll_sb_info *sbi = ll_s2sbi(sb);
784         unsigned int cookielen;
785         int rc;
786
787         rc = ll_get_default_cookiesize(sbi, &cookielen);
788         if (rc)
789                 return rc;
790
791         return seq_printf(m, "%u\n", cookielen);
792 }
793 LPROC_SEQ_FOPS_RO(ll_defult_cookiesize);
794
795 static int ll_sbi_flags_seq_show(struct seq_file *m, void *v)
796 {
797         const char *str[] = LL_SBI_FLAGS;
798         struct super_block *sb = m->private;
799         int flags = ll_s2sbi(sb)->ll_flags;
800         int i = 0;
801
802         while (flags != 0) {
803                 if (ARRAY_SIZE(str) <= i) {
804                         CERROR("%s: Revise array LL_SBI_FLAGS to match sbi "
805                                 "flags please.\n", ll_get_fsname(sb, NULL, 0));
806                         return -EINVAL;
807                 }
808
809                 if (flags & 0x1)
810                         seq_printf(m, "%s ", str[i]);
811                 flags >>= 1;
812                 ++i;
813         }
814         seq_printf(m, "\b\n");
815         return 0;
816 }
817 LPROC_SEQ_FOPS_RO(ll_sbi_flags);
818
819 static int ll_unstable_stats_seq_show(struct seq_file *m, void *v)
820 {
821         struct super_block      *sb    = m->private;
822         struct ll_sb_info       *sbi   = ll_s2sbi(sb);
823         struct cl_client_cache  *cache = &sbi->ll_cache;
824         int pages, mb;
825
826         pages = atomic_read(&cache->ccc_unstable_nr);
827         mb    = (pages * PAGE_CACHE_SIZE) >> 20;
828
829         return seq_printf(m, "unstable_pages: %8d\n"
830                                 "unstable_mb:    %8d\n", pages, mb);
831 }
832 LPROC_SEQ_FOPS_RO(ll_unstable_stats);
833
834 struct lprocfs_seq_vars lprocfs_llite_obd_vars[] = {
835         { .name =       "uuid",
836           .fops =       &ll_sb_uuid_fops                        },
837         { .name =       "fstype",
838           .fops =       &ll_fstype_fops                         },
839         { .name =       "site",
840           .fops =       &ll_site_stats_fops                     },
841         { .name =       "blocksize",
842           .fops =       &ll_blksize_fops                        },
843         { .name =       "kbytestotal",
844           .fops =       &ll_kbytestotal_fops                    },
845         { .name =       "kbytesfree",
846           .fops =       &ll_kbytesfree_fops                     },
847         { .name =       "kbytesavail",
848           .fops =       &ll_kbytesavail_fops                    },
849         { .name =       "filestotal",
850           .fops =       &ll_filestotal_fops                     },
851         { .name =       "filesfree",
852           .fops =       &ll_filesfree_fops                      },
853         { .name =       "client_type",
854           .fops =       &ll_client_type_fops                    },
855         { .name =       "max_read_ahead_mb",
856           .fops =       &ll_max_readahead_mb_fops               },
857         { .name =       "max_read_ahead_per_file_mb",
858           .fops =       &ll_max_readahead_per_file_mb_fops      },
859         { .name =       "max_read_ahead_whole_mb",
860           .fops =       &ll_max_read_ahead_whole_mb_fops        },
861         { .name =       "max_cached_mb",
862           .fops =       &ll_max_cached_mb_fops                  },
863         { .name =       "checksum_pages",
864           .fops =       &ll_checksum_fops                       },
865         { .name =       "max_rw_chunk",
866           .fops =       &ll_max_rw_chunk_fops                   },
867         { .name =       "stats_track_pid",
868           .fops =       &ll_track_pid_fops                      },
869         { .name =       "stats_track_ppid",
870           .fops =       &ll_track_ppid_fops                     },
871         { .name =       "stats_track_gid",
872           .fops =       &ll_track_gid_fops                      },
873         { .name =       "statahead_max",
874           .fops =       &ll_statahead_max_fops                  },
875         { .name =       "statahead_agl",
876           .fops =       &ll_statahead_agl_fops                  },
877         { .name =       "statahead_stats",
878           .fops =       &ll_statahead_stats_fops                },
879         { .name =       "lazystatfs",
880           .fops =       &ll_lazystatfs_fops                     },
881         { .name =       "max_easize",
882           .fops =       &ll_max_easize_fops                     },
883         { .name =       "default_easize",
884           .fops =       &ll_defult_easize_fops                  },
885         { .name =       "max_cookiesize",
886           .fops =       &ll_max_cookiesize_fops                 },
887         { .name =       "default_cookiesize",
888           .fops =       &ll_defult_cookiesize_fops              },
889         { .name =       "sbi_flags",
890           .fops =       &ll_sbi_flags_fops                      },
891         { .name =       "xattr_cache",
892           .fops =       &ll_xattr_cache_fops                    },
893         { .name =       "unstable_stats",
894           .fops =       &ll_unstable_stats_fops                 },
895         { 0 }
896 };
897
898 #define MAX_STRING_SIZE 128
899
900 static const struct llite_file_opcode {
901         __u32       opcode;
902         __u32       type;
903         const char *opname;
904 } llite_opcode_table[LPROC_LL_FILE_OPCODES] = {
905         /* file operation */
906         { LPROC_LL_DIRTY_HITS,     LPROCFS_TYPE_REGS, "dirty_pages_hits" },
907         { LPROC_LL_DIRTY_MISSES,   LPROCFS_TYPE_REGS, "dirty_pages_misses" },
908         { LPROC_LL_READ_BYTES,     LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_BYTES,
909                                    "read_bytes" },
910         { LPROC_LL_WRITE_BYTES,    LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_BYTES,
911                                    "write_bytes" },
912         { LPROC_LL_BRW_READ,       LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
913                                    "brw_read" },
914         { LPROC_LL_BRW_WRITE,      LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_PAGES,
915                                    "brw_write" },
916         { LPROC_LL_OSC_READ,       LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_BYTES,
917                                    "osc_read" },
918         { LPROC_LL_OSC_WRITE,      LPROCFS_CNTR_AVGMINMAX|LPROCFS_TYPE_BYTES,
919                                    "osc_write" },
920         { LPROC_LL_IOCTL,          LPROCFS_TYPE_REGS, "ioctl" },
921         { LPROC_LL_OPEN,           LPROCFS_TYPE_REGS, "open" },
922         { LPROC_LL_RELEASE,        LPROCFS_TYPE_REGS, "close" },
923         { LPROC_LL_MAP,            LPROCFS_TYPE_REGS, "mmap" },
924         { LPROC_LL_LLSEEK,         LPROCFS_TYPE_REGS, "seek" },
925         { LPROC_LL_FSYNC,          LPROCFS_TYPE_REGS, "fsync" },
926         { LPROC_LL_READDIR,        LPROCFS_TYPE_REGS, "readdir" },
927         /* inode operation */
928         { LPROC_LL_SETATTR,        LPROCFS_TYPE_REGS, "setattr" },
929         { LPROC_LL_TRUNC,          LPROCFS_TYPE_REGS, "truncate" },
930         { LPROC_LL_FLOCK,          LPROCFS_TYPE_REGS, "flock" },
931         { LPROC_LL_GETATTR,        LPROCFS_TYPE_REGS, "getattr" },
932         /* dir inode operation */
933         { LPROC_LL_CREATE,         LPROCFS_TYPE_REGS, "create" },
934         { LPROC_LL_LINK,           LPROCFS_TYPE_REGS, "link" },
935         { LPROC_LL_UNLINK,         LPROCFS_TYPE_REGS, "unlink" },
936         { LPROC_LL_SYMLINK,        LPROCFS_TYPE_REGS, "symlink" },
937         { LPROC_LL_MKDIR,          LPROCFS_TYPE_REGS, "mkdir" },
938         { LPROC_LL_RMDIR,          LPROCFS_TYPE_REGS, "rmdir" },
939         { LPROC_LL_MKNOD,          LPROCFS_TYPE_REGS, "mknod" },
940         { LPROC_LL_RENAME,         LPROCFS_TYPE_REGS, "rename" },
941         /* special inode operation */
942         { LPROC_LL_STAFS,          LPROCFS_TYPE_REGS, "statfs" },
943         { LPROC_LL_ALLOC_INODE,    LPROCFS_TYPE_REGS, "alloc_inode" },
944         { LPROC_LL_SETXATTR,       LPROCFS_TYPE_REGS, "setxattr" },
945         { LPROC_LL_GETXATTR,       LPROCFS_TYPE_REGS, "getxattr" },
946         { LPROC_LL_GETXATTR_HITS,  LPROCFS_TYPE_REGS, "getxattr_hits" },
947         { LPROC_LL_LISTXATTR,      LPROCFS_TYPE_REGS, "listxattr" },
948         { LPROC_LL_REMOVEXATTR,    LPROCFS_TYPE_REGS, "removexattr" },
949         { LPROC_LL_INODE_PERM,     LPROCFS_TYPE_REGS, "inode_permission" },
950 };
951
952 void ll_stats_ops_tally(struct ll_sb_info *sbi, int op, int count)
953 {
954         if (!sbi->ll_stats)
955                 return;
956         if (sbi->ll_stats_track_type == STATS_TRACK_ALL)
957                 lprocfs_counter_add(sbi->ll_stats, op, count);
958         else if (sbi->ll_stats_track_type == STATS_TRACK_PID &&
959                  sbi->ll_stats_track_id == current->pid)
960                 lprocfs_counter_add(sbi->ll_stats, op, count);
961         else if (sbi->ll_stats_track_type == STATS_TRACK_PPID &&
962                  sbi->ll_stats_track_id == current->parent->pid)
963                 lprocfs_counter_add(sbi->ll_stats, op, count);
964         else if (sbi->ll_stats_track_type == STATS_TRACK_GID &&
965                  sbi->ll_stats_track_id ==
966                         from_kgid(&init_user_ns, current_gid()))
967                 lprocfs_counter_add(sbi->ll_stats, op, count);
968 }
969 EXPORT_SYMBOL(ll_stats_ops_tally);
970
971 static const char *ra_stat_string[] = {
972         [RA_STAT_HIT] = "hits",
973         [RA_STAT_MISS] = "misses",
974         [RA_STAT_DISTANT_READPAGE] = "readpage not consecutive",
975         [RA_STAT_MISS_IN_WINDOW] = "miss inside window",
976         [RA_STAT_FAILED_GRAB_PAGE] = "failed grab_cache_page",
977         [RA_STAT_FAILED_MATCH] = "failed lock match",
978         [RA_STAT_DISCARDED] = "read but discarded",
979         [RA_STAT_ZERO_LEN] = "zero length file",
980         [RA_STAT_ZERO_WINDOW] = "zero size window",
981         [RA_STAT_EOF] = "read-ahead to EOF",
982         [RA_STAT_MAX_IN_FLIGHT] = "hit max r-a issue",
983         [RA_STAT_WRONG_GRAB_PAGE] = "wrong page from grab_cache_page",
984         [RA_STAT_FAILED_REACH_END] = "failed to reach end"
985 };
986
987 LPROC_SEQ_FOPS_RO_TYPE(llite, name);
988 LPROC_SEQ_FOPS_RO_TYPE(llite, uuid);
989
990 int lprocfs_register_mountpoint(struct proc_dir_entry *parent,
991                                 struct super_block *sb, char *osc, char *mdc)
992 {
993         struct lprocfs_seq_vars lvars[2];
994         struct lustre_sb_info *lsi = s2lsi(sb);
995         struct ll_sb_info *sbi = ll_s2sbi(sb);
996         struct obd_device *obd;
997         struct proc_dir_entry *dir;
998         char name[MAX_STRING_SIZE + 1], *ptr;
999         int err, id, len, rc;
1000         ENTRY;
1001
1002         memset(lvars, 0, sizeof(lvars));
1003
1004         name[MAX_STRING_SIZE] = '\0';
1005         lvars[0].name = name;
1006
1007         LASSERT(sbi != NULL);
1008         LASSERT(mdc != NULL);
1009         LASSERT(osc != NULL);
1010
1011         /* Get fsname */
1012         len = strlen(lsi->lsi_lmd->lmd_profile);
1013         ptr = strrchr(lsi->lsi_lmd->lmd_profile, '-');
1014         if (ptr && (strcmp(ptr, "-client") == 0))
1015                 len -= 7;
1016
1017         /* Mount info */
1018         snprintf(name, MAX_STRING_SIZE, "%.*s-%p", len,
1019                  lsi->lsi_lmd->lmd_profile, sb);
1020
1021         sbi->ll_proc_root = lprocfs_seq_register(name, parent, NULL, NULL);
1022         if (IS_ERR(sbi->ll_proc_root)) {
1023                 err = PTR_ERR(sbi->ll_proc_root);
1024                 sbi->ll_proc_root = NULL;
1025                 RETURN(err);
1026         }
1027
1028         rc = lprocfs_seq_create(sbi->ll_proc_root, "dump_page_cache", 0444,
1029                                 &vvp_dump_pgcache_file_ops, sbi);
1030         if (rc)
1031                 CWARN("Error adding the dump_page_cache file\n");
1032
1033         rc = lprocfs_seq_create(sbi->ll_proc_root, "extents_stats", 0644,
1034                                 &ll_rw_extents_stats_fops, sbi);
1035         if (rc)
1036                 CWARN("Error adding the extent_stats file\n");
1037
1038         rc = lprocfs_seq_create(sbi->ll_proc_root, "extents_stats_per_process",
1039                                 0644, &ll_rw_extents_stats_pp_fops, sbi);
1040         if (rc)
1041                 CWARN("Error adding the extents_stats_per_process file\n");
1042
1043         rc = lprocfs_seq_create(sbi->ll_proc_root, "offset_stats", 0644,
1044                                 &ll_rw_offset_stats_fops, sbi);
1045         if (rc)
1046                 CWARN("Error adding the offset_stats file\n");
1047
1048         /* File operations stats */
1049         sbi->ll_stats = lprocfs_alloc_stats(LPROC_LL_FILE_OPCODES,
1050                                             LPROCFS_STATS_FLAG_NONE);
1051         if (sbi->ll_stats == NULL)
1052                 GOTO(out, err = -ENOMEM);
1053         /* do counter init */
1054         for (id = 0; id < LPROC_LL_FILE_OPCODES; id++) {
1055                 __u32 type = llite_opcode_table[id].type;
1056                 void *ptr = NULL;
1057                 if (type & LPROCFS_TYPE_REGS)
1058                         ptr = "regs";
1059                 else if (type & LPROCFS_TYPE_BYTES)
1060                         ptr = "bytes";
1061                 else if (type & LPROCFS_TYPE_PAGES)
1062                         ptr = "pages";
1063                 lprocfs_counter_init(sbi->ll_stats,
1064                                      llite_opcode_table[id].opcode,
1065                                      (type & LPROCFS_CNTR_AVGMINMAX),
1066                                      llite_opcode_table[id].opname, ptr);
1067         }
1068         err = lprocfs_register_stats(sbi->ll_proc_root, "stats", sbi->ll_stats);
1069         if (err)
1070                 GOTO(out, err);
1071
1072         sbi->ll_ra_stats = lprocfs_alloc_stats(ARRAY_SIZE(ra_stat_string),
1073                                                LPROCFS_STATS_FLAG_NONE);
1074         if (sbi->ll_ra_stats == NULL)
1075                 GOTO(out, err = -ENOMEM);
1076
1077         for (id = 0; id < ARRAY_SIZE(ra_stat_string); id++)
1078                 lprocfs_counter_init(sbi->ll_ra_stats, id, 0,
1079                                      ra_stat_string[id], "pages");
1080         err = lprocfs_register_stats(sbi->ll_proc_root, "read_ahead_stats",
1081                                      sbi->ll_ra_stats);
1082         if (err)
1083                 GOTO(out, err);
1084
1085
1086         err = lprocfs_seq_add_vars(sbi->ll_proc_root, lprocfs_llite_obd_vars, sb);
1087         if (err)
1088                 GOTO(out, err);
1089
1090         /* MDC info */
1091         obd = class_name2obd(mdc);
1092
1093         LASSERT(obd != NULL);
1094         LASSERT(obd->obd_magic == OBD_DEVICE_MAGIC);
1095         LASSERT(obd->obd_type->typ_name != NULL);
1096
1097         dir = proc_mkdir(obd->obd_type->typ_name, sbi->ll_proc_root);
1098         if (dir == NULL)
1099                 GOTO(out, err = -ENOMEM);
1100
1101         snprintf(name, MAX_STRING_SIZE, "common_name");
1102         lvars[0].fops = &llite_name_fops;
1103         err = lprocfs_seq_add_vars(dir, lvars, obd);
1104         if (err)
1105                 GOTO(out, err);
1106
1107         snprintf(name, MAX_STRING_SIZE, "uuid");
1108         lvars[0].fops = &llite_uuid_fops;
1109         err = lprocfs_seq_add_vars(dir, lvars, obd);
1110         if (err)
1111                 GOTO(out, err);
1112
1113         /* OSC */
1114         obd = class_name2obd(osc);
1115
1116         LASSERT(obd != NULL);
1117         LASSERT(obd->obd_magic == OBD_DEVICE_MAGIC);
1118         LASSERT(obd->obd_type->typ_name != NULL);
1119
1120         dir = proc_mkdir(obd->obd_type->typ_name, sbi->ll_proc_root);
1121         if (dir == NULL)
1122                 GOTO(out, err = -ENOMEM);
1123
1124         snprintf(name, MAX_STRING_SIZE, "common_name");
1125         lvars[0].fops = &llite_name_fops;
1126         err = lprocfs_seq_add_vars(dir, lvars, obd);
1127         if (err)
1128                 GOTO(out, err);
1129
1130         snprintf(name, MAX_STRING_SIZE, "uuid");
1131         lvars[0].fops = &llite_uuid_fops;
1132         err = lprocfs_seq_add_vars(dir, lvars, obd);
1133 out:
1134         if (err) {
1135                 lprocfs_remove(&sbi->ll_proc_root);
1136                 lprocfs_free_stats(&sbi->ll_ra_stats);
1137                 lprocfs_free_stats(&sbi->ll_stats);
1138         }
1139         RETURN(err);
1140 }
1141
1142 void lprocfs_unregister_mountpoint(struct ll_sb_info *sbi)
1143 {
1144         if (sbi->ll_proc_root) {
1145                 lprocfs_remove(&sbi->ll_proc_root);
1146                 lprocfs_free_stats(&sbi->ll_ra_stats);
1147                 lprocfs_free_stats(&sbi->ll_stats);
1148         }
1149 }
1150 #undef MAX_STRING_SIZE
1151
1152 #define pct(a,b) (b ? a * 100 / b : 0)
1153
1154 static void ll_display_extents_info(struct ll_rw_extents_info *io_extents,
1155                                    struct seq_file *seq, int which)
1156 {
1157         unsigned long read_tot = 0, write_tot = 0, read_cum, write_cum;
1158         unsigned long start, end, r, w;
1159         char *unitp = "KMGTPEZY";
1160         int i, units = 10;
1161         struct per_process_info *pp_info = &io_extents->pp_extents[which];
1162
1163         read_cum = 0;
1164         write_cum = 0;
1165         start = 0;
1166
1167         for(i = 0; i < LL_HIST_MAX; i++) {
1168                 read_tot += pp_info->pp_r_hist.oh_buckets[i];
1169                 write_tot += pp_info->pp_w_hist.oh_buckets[i];
1170         }
1171
1172         for(i = 0; i < LL_HIST_MAX; i++) {
1173                 r = pp_info->pp_r_hist.oh_buckets[i];
1174                 w = pp_info->pp_w_hist.oh_buckets[i];
1175                 read_cum += r;
1176                 write_cum += w;
1177                 end = 1 << (i + LL_HIST_START - units);
1178                 seq_printf(seq, "%4lu%c - %4lu%c%c: %14lu %4lu %4lu  | "
1179                            "%14lu %4lu %4lu\n", start, *unitp, end, *unitp,
1180                            (i == LL_HIST_MAX - 1) ? '+' : ' ',
1181                            r, pct(r, read_tot), pct(read_cum, read_tot),
1182                            w, pct(w, write_tot), pct(write_cum, write_tot));
1183                 start = end;
1184                 if (start == 1<<10) {
1185                         start = 1;
1186                         units += 10;
1187                         unitp++;
1188                 }
1189                 if (read_cum == read_tot && write_cum == write_tot)
1190                         break;
1191         }
1192 }
1193
1194 static int ll_rw_extents_stats_pp_seq_show(struct seq_file *seq, void *v)
1195 {
1196         struct timeval now;
1197         struct ll_sb_info *sbi = seq->private;
1198         struct ll_rw_extents_info *io_extents = &sbi->ll_rw_extents_info;
1199         int k;
1200
1201         do_gettimeofday(&now);
1202
1203         if (!sbi->ll_rw_stats_on) {
1204                 seq_printf(seq, "disabled\n"
1205                                 "write anything in this file to activate, "
1206                                 "then 0 or \"[D/d]isabled\" to deactivate\n");
1207                 return 0;
1208         }
1209         seq_printf(seq, "snapshot_time:         %lu.%lu (secs.usecs)\n",
1210                    now.tv_sec, now.tv_usec);
1211         seq_printf(seq, "%15s %19s       | %20s\n", " ", "read", "write");
1212         seq_printf(seq, "%13s   %14s %4s %4s  | %14s %4s %4s\n",
1213                    "extents", "calls", "%", "cum%",
1214                    "calls", "%", "cum%");
1215         spin_lock(&sbi->ll_pp_extent_lock);
1216         for (k = 0; k < LL_PROCESS_HIST_MAX; k++) {
1217                 if (io_extents->pp_extents[k].pid != 0) {
1218                         seq_printf(seq, "\nPID: %d\n",
1219                                    io_extents->pp_extents[k].pid);
1220                         ll_display_extents_info(io_extents, seq, k);
1221                 }
1222         }
1223         spin_unlock(&sbi->ll_pp_extent_lock);
1224         return 0;
1225 }
1226
1227 static ssize_t ll_rw_extents_stats_pp_seq_write(struct file *file,
1228                                                 const char __user *buf,
1229                                                 size_t len,
1230                                                 loff_t *off)
1231 {
1232         struct seq_file *seq = file->private_data;
1233         struct ll_sb_info *sbi = seq->private;
1234         struct ll_rw_extents_info *io_extents = &sbi->ll_rw_extents_info;
1235         int i;
1236         int value = 1, rc = 0;
1237
1238         if (len == 0)
1239                 return -EINVAL;
1240
1241         rc = lprocfs_write_helper(buf, len, &value);
1242         if (rc < 0 && len < 16) {
1243                 char kernbuf[16];
1244
1245                 if (copy_from_user(kernbuf, buf, len))
1246                         return -EFAULT;
1247                 kernbuf[len] = 0;
1248
1249                 if (kernbuf[len - 1] == '\n')
1250                         kernbuf[len - 1] = 0;
1251
1252                 if (strcmp(kernbuf, "disabled") == 0 ||
1253                     strcmp(kernbuf, "Disabled") == 0)
1254                         value = 0;
1255         }
1256
1257         if (value == 0)
1258                 sbi->ll_rw_stats_on = 0;
1259         else
1260                 sbi->ll_rw_stats_on = 1;
1261
1262         spin_lock(&sbi->ll_pp_extent_lock);
1263         for (i = 0; i < LL_PROCESS_HIST_MAX; i++) {
1264                 io_extents->pp_extents[i].pid = 0;
1265                 lprocfs_oh_clear(&io_extents->pp_extents[i].pp_r_hist);
1266                 lprocfs_oh_clear(&io_extents->pp_extents[i].pp_w_hist);
1267         }
1268         spin_unlock(&sbi->ll_pp_extent_lock);
1269         return len;
1270 }
1271
1272 LPROC_SEQ_FOPS(ll_rw_extents_stats_pp);
1273
1274 static int ll_rw_extents_stats_seq_show(struct seq_file *seq, void *v)
1275 {
1276         struct timeval now;
1277         struct ll_sb_info *sbi = seq->private;
1278         struct ll_rw_extents_info *io_extents = &sbi->ll_rw_extents_info;
1279
1280         do_gettimeofday(&now);
1281
1282         if (!sbi->ll_rw_stats_on) {
1283                 seq_printf(seq, "disabled\n"
1284                                 "write anything in this file to activate, "
1285                                 "then 0 or \"[D/d]isabled\" to deactivate\n");
1286                 return 0;
1287         }
1288         seq_printf(seq, "snapshot_time:         %lu.%lu (secs.usecs)\n",
1289                    now.tv_sec, now.tv_usec);
1290
1291         seq_printf(seq, "%15s %19s       | %20s\n", " ", "read", "write");
1292         seq_printf(seq, "%13s   %14s %4s %4s  | %14s %4s %4s\n",
1293                    "extents", "calls", "%", "cum%",
1294                    "calls", "%", "cum%");
1295         spin_lock(&sbi->ll_lock);
1296         ll_display_extents_info(io_extents, seq, LL_PROCESS_HIST_MAX);
1297         spin_unlock(&sbi->ll_lock);
1298
1299         return 0;
1300 }
1301
1302 static ssize_t ll_rw_extents_stats_seq_write(struct file *file,
1303                                              const char __user *buf,
1304                                              size_t len, loff_t *off)
1305 {
1306         struct seq_file *seq = file->private_data;
1307         struct ll_sb_info *sbi = seq->private;
1308         struct ll_rw_extents_info *io_extents = &sbi->ll_rw_extents_info;
1309         int i;
1310         int value = 1, rc = 0;
1311
1312         if (len == 0)
1313                 return -EINVAL;
1314
1315         rc = lprocfs_write_helper(buf, len, &value);
1316         if (rc < 0 && len < 16) {
1317                 char kernbuf[16];
1318
1319                 if (copy_from_user(kernbuf, buf, len))
1320                         return -EFAULT;
1321                 kernbuf[len] = 0;
1322
1323                 if (kernbuf[len - 1] == '\n')
1324                         kernbuf[len - 1] = 0;
1325
1326                 if (strcmp(kernbuf, "disabled") == 0 ||
1327                     strcmp(kernbuf, "Disabled") == 0)
1328                         value = 0;
1329         }
1330
1331         if (value == 0)
1332                 sbi->ll_rw_stats_on = 0;
1333         else
1334                 sbi->ll_rw_stats_on = 1;
1335
1336         spin_lock(&sbi->ll_pp_extent_lock);
1337         for (i = 0; i <= LL_PROCESS_HIST_MAX; i++) {
1338                 io_extents->pp_extents[i].pid = 0;
1339                 lprocfs_oh_clear(&io_extents->pp_extents[i].pp_r_hist);
1340                 lprocfs_oh_clear(&io_extents->pp_extents[i].pp_w_hist);
1341         }
1342         spin_unlock(&sbi->ll_pp_extent_lock);
1343
1344         return len;
1345 }
1346 LPROC_SEQ_FOPS(ll_rw_extents_stats);
1347
1348 void ll_rw_stats_tally(struct ll_sb_info *sbi, pid_t pid,
1349                        struct ll_file_data *file, loff_t pos,
1350                        size_t count, int rw)
1351 {
1352         int i, cur = -1;
1353         struct ll_rw_process_info *process;
1354         struct ll_rw_process_info *offset;
1355         int *off_count = &sbi->ll_rw_offset_entry_count;
1356         int *process_count = &sbi->ll_offset_process_count;
1357         struct ll_rw_extents_info *io_extents = &sbi->ll_rw_extents_info;
1358
1359         if(!sbi->ll_rw_stats_on)
1360                 return;
1361         process = sbi->ll_rw_process_info;
1362         offset = sbi->ll_rw_offset_info;
1363
1364         spin_lock(&sbi->ll_pp_extent_lock);
1365         /* Extent statistics */
1366         for(i = 0; i < LL_PROCESS_HIST_MAX; i++) {
1367                 if(io_extents->pp_extents[i].pid == pid) {
1368                         cur = i;
1369                         break;
1370                 }
1371         }
1372
1373         if (cur == -1) {
1374                 /* new process */
1375                 sbi->ll_extent_process_count =
1376                         (sbi->ll_extent_process_count + 1) % LL_PROCESS_HIST_MAX;
1377                 cur = sbi->ll_extent_process_count;
1378                 io_extents->pp_extents[cur].pid = pid;
1379                 lprocfs_oh_clear(&io_extents->pp_extents[cur].pp_r_hist);
1380                 lprocfs_oh_clear(&io_extents->pp_extents[cur].pp_w_hist);
1381         }
1382
1383         for(i = 0; (count >= (1 << LL_HIST_START << i)) &&
1384              (i < (LL_HIST_MAX - 1)); i++);
1385         if (rw == 0) {
1386                 io_extents->pp_extents[cur].pp_r_hist.oh_buckets[i]++;
1387                 io_extents->pp_extents[LL_PROCESS_HIST_MAX].pp_r_hist.oh_buckets[i]++;
1388         } else {
1389                 io_extents->pp_extents[cur].pp_w_hist.oh_buckets[i]++;
1390                 io_extents->pp_extents[LL_PROCESS_HIST_MAX].pp_w_hist.oh_buckets[i]++;
1391         }
1392         spin_unlock(&sbi->ll_pp_extent_lock);
1393
1394         spin_lock(&sbi->ll_process_lock);
1395         /* Offset statistics */
1396         for (i = 0; i < LL_PROCESS_HIST_MAX; i++) {
1397                 if (process[i].rw_pid == pid) {
1398                         if (process[i].rw_last_file != file) {
1399                                 process[i].rw_range_start = pos;
1400                                 process[i].rw_last_file_pos = pos + count;
1401                                 process[i].rw_smallest_extent = count;
1402                                 process[i].rw_largest_extent = count;
1403                                 process[i].rw_offset = 0;
1404                                 process[i].rw_last_file = file;
1405                                 spin_unlock(&sbi->ll_process_lock);
1406                                 return;
1407                         }
1408                         if (process[i].rw_last_file_pos != pos) {
1409                                 *off_count =
1410                                     (*off_count + 1) % LL_OFFSET_HIST_MAX;
1411                                 offset[*off_count].rw_op = process[i].rw_op;
1412                                 offset[*off_count].rw_pid = pid;
1413                                 offset[*off_count].rw_range_start =
1414                                         process[i].rw_range_start;
1415                                 offset[*off_count].rw_range_end =
1416                                         process[i].rw_last_file_pos;
1417                                 offset[*off_count].rw_smallest_extent =
1418                                         process[i].rw_smallest_extent;
1419                                 offset[*off_count].rw_largest_extent =
1420                                         process[i].rw_largest_extent;
1421                                 offset[*off_count].rw_offset =
1422                                         process[i].rw_offset;
1423                                 process[i].rw_op = rw;
1424                                 process[i].rw_range_start = pos;
1425                                 process[i].rw_smallest_extent = count;
1426                                 process[i].rw_largest_extent = count;
1427                                 process[i].rw_offset = pos -
1428                                         process[i].rw_last_file_pos;
1429                         }
1430                         if(process[i].rw_smallest_extent > count)
1431                                 process[i].rw_smallest_extent = count;
1432                         if(process[i].rw_largest_extent < count)
1433                                 process[i].rw_largest_extent = count;
1434                         process[i].rw_last_file_pos = pos + count;
1435                         spin_unlock(&sbi->ll_process_lock);
1436                         return;
1437                 }
1438         }
1439         *process_count = (*process_count + 1) % LL_PROCESS_HIST_MAX;
1440         process[*process_count].rw_pid = pid;
1441         process[*process_count].rw_op = rw;
1442         process[*process_count].rw_range_start = pos;
1443         process[*process_count].rw_last_file_pos = pos + count;
1444         process[*process_count].rw_smallest_extent = count;
1445         process[*process_count].rw_largest_extent = count;
1446         process[*process_count].rw_offset = 0;
1447         process[*process_count].rw_last_file = file;
1448         spin_unlock(&sbi->ll_process_lock);
1449 }
1450
1451 static int ll_rw_offset_stats_seq_show(struct seq_file *seq, void *v)
1452 {
1453         struct timeval now;
1454         struct ll_sb_info *sbi = seq->private;
1455         struct ll_rw_process_info *offset = sbi->ll_rw_offset_info;
1456         struct ll_rw_process_info *process = sbi->ll_rw_process_info;
1457         int i;
1458
1459         do_gettimeofday(&now);
1460
1461         if (!sbi->ll_rw_stats_on) {
1462                 seq_printf(seq, "disabled\n"
1463                                 "write anything in this file to activate, "
1464                                 "then 0 or \"[D/d]isabled\" to deactivate\n");
1465                 return 0;
1466         }
1467         spin_lock(&sbi->ll_process_lock);
1468
1469         seq_printf(seq, "snapshot_time:         %lu.%lu (secs.usecs)\n",
1470                    now.tv_sec, now.tv_usec);
1471         seq_printf(seq, "%3s %10s %14s %14s %17s %17s %14s\n",
1472                    "R/W", "PID", "RANGE START", "RANGE END",
1473                    "SMALLEST EXTENT", "LARGEST EXTENT", "OFFSET");
1474
1475         /* We stored the discontiguous offsets here; print them first */
1476         for (i = 0; i < LL_OFFSET_HIST_MAX; i++) {
1477                 if (offset[i].rw_pid != 0)
1478                         seq_printf(seq,
1479                                    "%3c %10d %14Lu %14Lu %17lu %17lu %14Lu",
1480                                    offset[i].rw_op == READ ? 'R' : 'W',
1481                                    offset[i].rw_pid,
1482                                    offset[i].rw_range_start,
1483                                    offset[i].rw_range_end,
1484                                    (unsigned long)offset[i].rw_smallest_extent,
1485                                    (unsigned long)offset[i].rw_largest_extent,
1486                                    offset[i].rw_offset);
1487         }
1488
1489         /* Then print the current offsets for each process */
1490         for (i = 0; i < LL_PROCESS_HIST_MAX; i++) {
1491                 if (process[i].rw_pid != 0)
1492                         seq_printf(seq,
1493                                    "%3c %10d %14Lu %14Lu %17lu %17lu %14Lu",
1494                                    process[i].rw_op == READ ? 'R' : 'W',
1495                                    process[i].rw_pid,
1496                                    process[i].rw_range_start,
1497                                    process[i].rw_last_file_pos,
1498                                    (unsigned long)process[i].rw_smallest_extent,
1499                                    (unsigned long)process[i].rw_largest_extent,
1500                                    process[i].rw_offset);
1501         }
1502         spin_unlock(&sbi->ll_process_lock);
1503
1504         return 0;
1505 }
1506
1507 static ssize_t ll_rw_offset_stats_seq_write(struct file *file,
1508                                             const char __user *buf,
1509                                             size_t len, loff_t *off)
1510 {
1511         struct seq_file *seq = file->private_data;
1512         struct ll_sb_info *sbi = seq->private;
1513         struct ll_rw_process_info *process_info = sbi->ll_rw_process_info;
1514         struct ll_rw_process_info *offset_info = sbi->ll_rw_offset_info;
1515         int value = 1, rc = 0;
1516
1517         if (len == 0)
1518                 return -EINVAL;
1519
1520         rc = lprocfs_write_helper(buf, len, &value);
1521
1522         if (rc < 0 && len < 16) {
1523                 char kernbuf[16];
1524
1525                 if (copy_from_user(kernbuf, buf, len))
1526                         return -EFAULT;
1527                 kernbuf[len] = 0;
1528
1529                 if (kernbuf[len - 1] == '\n')
1530                         kernbuf[len - 1] = 0;
1531
1532                 if (strcmp(kernbuf, "disabled") == 0 ||
1533                     strcmp(kernbuf, "Disabled") == 0)
1534                         value = 0;
1535         }
1536
1537         if (value == 0)
1538                 sbi->ll_rw_stats_on = 0;
1539         else
1540                 sbi->ll_rw_stats_on = 1;
1541
1542         spin_lock(&sbi->ll_process_lock);
1543         sbi->ll_offset_process_count = 0;
1544         sbi->ll_rw_offset_entry_count = 0;
1545         memset(process_info, 0, sizeof(struct ll_rw_process_info) *
1546                LL_PROCESS_HIST_MAX);
1547         memset(offset_info, 0, sizeof(struct ll_rw_process_info) *
1548                LL_OFFSET_HIST_MAX);
1549         spin_unlock(&sbi->ll_process_lock);
1550
1551         return len;
1552 }
1553
1554 LPROC_SEQ_FOPS(ll_rw_offset_stats);
1555 #endif /* LPROCFS */