1 /* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*-
2 * vim:expandtab:shiftwidth=8:tabstop=8:
6 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License version 2 only,
10 * as published by the Free Software Foundation.
12 * This program is distributed in the hope that it will be useful, but
13 * WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * General Public License version 2 for more details (a copy is included
16 * in the LICENSE file that accompanied this code).
18 * You should have received a copy of the GNU General Public License
19 * version 2 along with this program; If not, see
20 * http://www.sun.com/software/products/lustre/docs/GPLv2.pdf
22 * Please contact Sun Microsystems, Inc., 4150 Network Circle, Santa Clara,
23 * CA 95054 USA or visit www.sun.com if you need additional information or
29 * Copyright 2008 Sun Microsystems, Inc. All rights reserved
30 * Use is subject to license terms.
33 * This file is part of Lustre, http://www.lustre.org/
34 * Lustre is a trademark of Sun Microsystems, Inc.
36 #define DEBUG_SUBSYSTEM S_CLASS
38 #include <linux/version.h>
39 #include <lprocfs_status.h>
41 #include <linux/seq_file.h>
42 #include <linux/version.h>
44 #include "filter_internal.h"
47 static int lprocfs_filter_rd_groups(char *page, char **start, off_t off,
48 int count, int *eof, void *data)
50 struct obd_device *obd = (struct obd_device *)data;
52 return snprintf(page, count, "%u\n", obd->u.filter.fo_group_count);
55 static int lprocfs_filter_rd_tot_dirty(char *page, char **start, off_t off,
56 int count, int *eof, void *data)
58 struct obd_device *obd = (struct obd_device *)data;
62 return snprintf(page, count, LPU64"\n", obd->u.filter.fo_tot_dirty);
65 static int lprocfs_filter_rd_tot_granted(char *page, char **start, off_t off,
66 int count, int *eof, void *data)
68 struct obd_device *obd = (struct obd_device *)data;
72 return snprintf(page, count, LPU64"\n", obd->u.filter.fo_tot_granted);
75 static int lprocfs_filter_rd_tot_pending(char *page, char **start, off_t off,
76 int count, int *eof, void *data)
78 struct obd_device *obd = (struct obd_device *)data;
82 return snprintf(page, count, LPU64"\n", obd->u.filter.fo_tot_pending);
85 static int lprocfs_filter_rd_mntdev(char *page, char **start, off_t off,
86 int count, int *eof, void *data)
88 struct obd_device *obd = (struct obd_device *)data;
91 LASSERT(obd->u.obt.obt_vfsmnt->mnt_devname);
93 return snprintf(page, count, "%s\n",
94 obd->u.obt.obt_vfsmnt->mnt_devname);
97 static int lprocfs_filter_rd_last_id(char *page, char **start, off_t off,
98 int count, int *eof, void *data)
100 struct obd_device *obd = data;
101 struct filter_obd *filter = &obd->u.filter;
102 int retval = 0, rc, i;
104 if (obd == NULL || !obd->obd_set_up || obd->obd_stopping)
106 rc = snprintf(page, count, LPU64"\n",filter_last_id(filter, 0));
113 for (i = FID_SEQ_OST_MDT1; i < filter->fo_group_count; i++) {
114 rc = snprintf(page, count, LPU64"\n",filter_last_id(filter, i));
126 int lprocfs_filter_rd_readcache(char *page, char **start, off_t off, int count,
127 int *eof, void *data)
129 struct obd_device *obd = data;
132 rc = snprintf(page, count, LPU64"\n",
133 obd->u.filter.fo_readcache_max_filesize);
137 int lprocfs_filter_wr_readcache(struct file *file, const char *buffer,
138 unsigned long count, void *data)
140 struct obd_device *obd = data;
144 rc = lprocfs_write_u64_helper(buffer, count, &val);
148 obd->u.filter.fo_readcache_max_filesize = val;
152 int lprocfs_filter_rd_fmd_max_num(char *page, char **start, off_t off,
153 int count, int *eof, void *data)
155 struct obd_device *obd = data;
158 rc = snprintf(page, count, "%u\n", obd->u.filter.fo_fmd_max_num);
162 int lprocfs_filter_wr_fmd_max_num(struct file *file, const char *buffer,
163 unsigned long count, void *data)
165 struct obd_device *obd = data;
169 rc = lprocfs_write_helper(buffer, count, &val);
173 if (val > 65536 || val < 1)
176 obd->u.filter.fo_fmd_max_num = val;
180 int lprocfs_filter_rd_fmd_max_age(char *page, char **start, off_t off,
181 int count, int *eof, void *data)
183 struct obd_device *obd = data;
186 rc = snprintf(page, count, "%u\n",
187 obd->u.filter.fo_fmd_max_age / CFS_HZ);
191 int lprocfs_filter_wr_fmd_max_age(struct file *file, const char *buffer,
192 unsigned long count, void *data)
194 struct obd_device *obd = data;
198 rc = lprocfs_write_helper(buffer, count, &val);
202 if (val > 65536 || val < 1)
205 obd->u.filter.fo_fmd_max_age = val * CFS_HZ;
209 static int lprocfs_filter_rd_capa(char *page, char **start, off_t off,
210 int count, int *eof, void *data)
212 struct obd_device *obd = data;
215 rc = snprintf(page, count, "capability on: %s\n",
216 obd->u.filter.fo_fl_oss_capa ? "oss" : "");
220 static int lprocfs_filter_wr_capa(struct file *file, const char *buffer,
221 unsigned long count, void *data)
223 struct obd_device *obd = data;
226 rc = lprocfs_write_helper(buffer, count, &val);
231 CERROR("invalid capability mode, only 0/1 are accepted.\n"
232 " 1: enable oss fid capability\n"
233 " 0: disable oss fid capability\n");
237 obd->u.filter.fo_fl_oss_capa = val;
238 LCONSOLE_INFO("OSS %s %s fid capability.\n", obd->obd_name,
239 val ? "enabled" : "disabled");
243 static int lprocfs_filter_rd_capa_count(char *page, char **start, off_t off,
244 int count, int *eof, void *data)
246 return snprintf(page, count, "%d %d\n",
247 capa_count[CAPA_SITE_CLIENT],
248 capa_count[CAPA_SITE_SERVER]);
251 static int lprocfs_rd_sec_level(char *page, char **start, off_t off,
252 int count, int *eof, void *data)
254 struct obd_device *obd = data;
256 return snprintf(page, count, "%d\n", obd->u.filter.fo_sec_level);
259 static int lprocfs_wr_sec_level(struct file *file, const char *buffer,
260 unsigned long count, void *data)
262 struct obd_device *obd = data;
265 rc = lprocfs_write_helper(buffer, count, &val);
269 if (val > LUSTRE_SEC_ALL || val < LUSTRE_SEC_NONE)
272 if (val == LUSTRE_SEC_SPECIFY) {
273 CWARN("security level %d will be supported in future.\n",
278 obd->u.filter.fo_sec_level = val;
282 static int lprocfs_filter_rd_cache(char *page, char **start, off_t off,
283 int count, int *eof, void *data)
285 struct obd_device *obd = (struct obd_device *)data;
286 LASSERT(obd != NULL);
288 return snprintf(page, count, "%u\n", obd->u.filter.fo_read_cache);
291 static int lprocfs_filter_wr_cache(struct file *file, const char *buffer,
292 unsigned long count, void *data)
294 struct obd_device *obd = (struct obd_device *)data;
296 LASSERT(obd != NULL);
298 rc = lprocfs_write_helper(buffer, count, &val);
303 cfs_spin_lock_bh(&obd->obd_processing_task_lock);
304 obd->u.filter.fo_read_cache = val;
305 cfs_spin_unlock_bh(&obd->obd_processing_task_lock);
309 static int lprocfs_filter_rd_wcache(char *page, char **start, off_t off,
310 int count, int *eof, void *data)
312 struct obd_device *obd = (struct obd_device *)data;
313 LASSERT(obd != NULL);
315 return snprintf(page, count, "%u\n", obd->u.filter.fo_writethrough_cache);
318 static int lprocfs_filter_wr_wcache(struct file *file, const char *buffer,
319 unsigned long count, void *data)
321 struct obd_device *obd = (struct obd_device *)data;
323 LASSERT(obd != NULL);
325 rc = lprocfs_write_helper(buffer, count, &val);
330 cfs_spin_lock_bh(&obd->obd_processing_task_lock);
331 obd->u.filter.fo_writethrough_cache = val;
332 cfs_spin_unlock_bh(&obd->obd_processing_task_lock);
336 static int lprocfs_filter_rd_mds_sync(char *page, char **start, off_t off,
337 int count, int *eof, void *data)
339 struct obd_device *obd = (struct obd_device *)data;
340 LASSERT(obd != NULL);
342 return snprintf(page, count, "%u\n", obd->u.filter.fo_mds_ost_sync);
345 int lprocfs_filter_rd_degraded(char *page, char **start, off_t off,
346 int count, int *eof, void *data)
348 struct obd_device *obd = data;
350 return snprintf(page, count, "%u\n", obd->u.filter.fo_raid_degraded);
353 int lprocfs_filter_wr_degraded(struct file *file, const char *buffer,
354 unsigned long count, void *data)
356 struct obd_device *obd = data;
359 rc = lprocfs_write_helper(buffer, count, &val);
363 cfs_spin_lock(&obd->obd_osfs_lock);
364 obd->u.filter.fo_raid_degraded = !!val;
365 cfs_spin_unlock(&obd->obd_osfs_lock);
369 static struct lprocfs_vars lprocfs_filter_obd_vars[] = {
370 { "uuid", lprocfs_rd_uuid, 0, 0 },
371 { "blocksize", lprocfs_rd_blksize, 0, 0 },
372 { "kbytestotal", lprocfs_rd_kbytestotal, 0, 0 },
373 { "kbytesfree", lprocfs_rd_kbytesfree, 0, 0 },
374 { "kbytesavail", lprocfs_rd_kbytesavail, 0, 0 },
375 { "filestotal", lprocfs_rd_filestotal, 0, 0 },
376 { "filesfree", lprocfs_rd_filesfree, 0, 0 },
377 { "filegroups", lprocfs_filter_rd_groups, 0, 0 },
378 { "fstype", lprocfs_rd_fstype, 0, 0 },
379 { "mntdev", lprocfs_filter_rd_mntdev, 0, 0 },
380 { "last_id", lprocfs_filter_rd_last_id,0, 0 },
381 { "tot_dirty", lprocfs_filter_rd_tot_dirty, 0, 0 },
382 { "tot_pending", lprocfs_filter_rd_tot_pending, 0, 0 },
383 { "tot_granted", lprocfs_filter_rd_tot_granted, 0, 0 },
384 { "hash_stats", lprocfs_obd_rd_hash, 0, 0 },
385 { "recovery_status", lprocfs_obd_rd_recovery_status, 0, 0 },
386 { "recovery_time_soft", lprocfs_obd_rd_recovery_time_soft,
387 lprocfs_obd_wr_recovery_time_soft, 0},
388 { "recovery_time_hard", lprocfs_obd_rd_recovery_time_hard,
389 lprocfs_obd_wr_recovery_time_hard, 0},
390 { "evict_client", 0, lprocfs_wr_evict_client, 0,
391 &lprocfs_evict_client_fops},
392 { "num_exports", lprocfs_rd_num_exports, 0, 0 },
393 { "readcache_max_filesize",
394 lprocfs_filter_rd_readcache,
395 lprocfs_filter_wr_readcache, 0 },
396 #ifdef HAVE_QUOTA_SUPPORT
397 { "quota_type", lprocfs_quota_rd_type,
398 lprocfs_quota_wr_type, 0},
400 { "client_cache_count", lprocfs_filter_rd_fmd_max_num,
401 lprocfs_filter_wr_fmd_max_num, 0 },
402 { "client_cache_seconds", lprocfs_filter_rd_fmd_max_age,
403 lprocfs_filter_wr_fmd_max_age, 0 },
404 { "capa", lprocfs_filter_rd_capa,
405 lprocfs_filter_wr_capa, 0 },
406 { "capa_count", lprocfs_filter_rd_capa_count, 0, 0 },
407 { "sec_level", lprocfs_rd_sec_level,
408 lprocfs_wr_sec_level, 0 },
409 { "read_cache_enable", lprocfs_filter_rd_cache, lprocfs_filter_wr_cache, 0},
410 { "writethrough_cache_enable", lprocfs_filter_rd_wcache,
411 lprocfs_filter_wr_wcache, 0},
412 { "mds_sync", lprocfs_filter_rd_mds_sync, 0, 0},
413 { "degraded", lprocfs_filter_rd_degraded,
414 lprocfs_filter_wr_degraded, 0 },
418 static struct lprocfs_vars lprocfs_filter_module_vars[] = {
419 { "num_refs", lprocfs_rd_numrefs, 0, 0 },
423 void filter_tally(struct obd_export *exp, struct page **pages, int nr_pages,
424 unsigned long *blocks, int blocks_per_page, int wr)
426 struct filter_obd *filter = &exp->exp_obd->u.filter;
427 struct page *last_page = NULL;
428 unsigned long *last_block = NULL;
429 unsigned long discont_pages = 0;
430 unsigned long discont_blocks = 0;
436 lprocfs_oh_tally_log2(&filter->fo_filter_stats.hist[BRW_R_PAGES + wr],
438 if (exp->exp_nid_stats && exp->exp_nid_stats->nid_brw_stats)
439 lprocfs_oh_tally_log2(&exp->exp_nid_stats->nid_brw_stats->
440 hist[BRW_R_PAGES + wr], nr_pages);
442 while (nr_pages-- > 0) {
443 if (last_page && (*pages)->index != (last_page->index + 1))
447 for (i = 0; i < blocks_per_page; i++) {
448 if (last_block && *blocks != (*last_block + 1))
450 last_block = blocks++;
454 lprocfs_oh_tally(&filter->fo_filter_stats.hist[BRW_R_DISCONT_PAGES +wr],
456 lprocfs_oh_tally(&filter->fo_filter_stats.hist[BRW_R_DISCONT_BLOCKS+wr],
459 if (exp->exp_nid_stats && exp->exp_nid_stats->nid_brw_stats) {
460 lprocfs_oh_tally_log2(&exp->exp_nid_stats->nid_brw_stats->
461 hist[BRW_R_DISCONT_PAGES + wr],
463 lprocfs_oh_tally_log2(&exp->exp_nid_stats->nid_brw_stats->
464 hist[BRW_R_DISCONT_BLOCKS + wr],
469 #define pct(a,b) (b ? a * 100 / b : 0)
471 static void display_brw_stats(struct seq_file *seq, char *name, char *units,
472 struct obd_histogram *read, struct obd_histogram *write, int log2)
474 unsigned long read_tot, write_tot, r, w, read_cum = 0, write_cum = 0;
477 seq_printf(seq, "\n%26s read | write\n", " ");
478 seq_printf(seq, "%-22s %-5s %% cum %% | %-5s %% cum %%\n",
481 read_tot = lprocfs_oh_sum(read);
482 write_tot = lprocfs_oh_sum(write);
483 for (i = 0; i < OBD_HIST_MAX; i++) {
484 r = read->oh_buckets[i];
485 w = write->oh_buckets[i];
488 if (read_cum == 0 && write_cum == 0)
492 seq_printf(seq, "%u", i);
494 seq_printf(seq, "%u", 1<<i);
496 seq_printf(seq, "%uK", 1<<(i-10));
498 seq_printf(seq, "%uM", 1<<(i-20));
500 seq_printf(seq, ":\t\t%10lu %3lu %3lu | %4lu %3lu %3lu\n",
501 r, pct(r, read_tot), pct(read_cum, read_tot),
502 w, pct(w, write_tot), pct(write_cum, write_tot));
504 if (read_cum == read_tot && write_cum == write_tot)
509 static void brw_stats_show(struct seq_file *seq, struct brw_stats *brw_stats)
513 /* this sampling races with updates */
514 cfs_gettimeofday(&now);
515 seq_printf(seq, "snapshot_time: %lu.%lu (secs.usecs)\n",
516 now.tv_sec, now.tv_usec);
518 display_brw_stats(seq, "pages per bulk r/w", "rpcs",
519 &brw_stats->hist[BRW_R_PAGES],
520 &brw_stats->hist[BRW_W_PAGES], 1);
522 display_brw_stats(seq, "discontiguous pages", "rpcs",
523 &brw_stats->hist[BRW_R_DISCONT_PAGES],
524 &brw_stats->hist[BRW_W_DISCONT_PAGES], 0);
526 display_brw_stats(seq, "discontiguous blocks", "rpcs",
527 &brw_stats->hist[BRW_R_DISCONT_BLOCKS],
528 &brw_stats->hist[BRW_W_DISCONT_BLOCKS], 0);
530 display_brw_stats(seq, "disk fragmented I/Os", "ios",
531 &brw_stats->hist[BRW_R_DIO_FRAGS],
532 &brw_stats->hist[BRW_W_DIO_FRAGS], 0);
534 display_brw_stats(seq, "disk I/Os in flight", "ios",
535 &brw_stats->hist[BRW_R_RPC_HIST],
536 &brw_stats->hist[BRW_W_RPC_HIST], 0);
540 sprintf(title, "I/O time (1/%ds)", CFS_HZ);
541 display_brw_stats(seq, title, "ios",
542 &brw_stats->hist[BRW_R_IO_TIME],
543 &brw_stats->hist[BRW_W_IO_TIME], 1);
546 display_brw_stats(seq, "disk I/O size", "ios",
547 &brw_stats->hist[BRW_R_DISK_IOSIZE],
548 &brw_stats->hist[BRW_W_DISK_IOSIZE], 1);
553 static int filter_brw_stats_seq_show(struct seq_file *seq, void *v)
555 struct obd_device *dev = seq->private;
556 struct filter_obd *filter = &dev->u.filter;
558 brw_stats_show(seq, &filter->fo_filter_stats);
563 static ssize_t filter_brw_stats_seq_write(struct file *file, const char *buf,
564 size_t len, loff_t *off)
566 struct seq_file *seq = file->private_data;
567 struct obd_device *dev = seq->private;
568 struct filter_obd *filter = &dev->u.filter;
571 for (i = 0; i < BRW_LAST; i++)
572 lprocfs_oh_clear(&filter->fo_filter_stats.hist[i]);
577 LPROC_SEQ_FOPS(filter_brw_stats);
579 int lproc_filter_attach_seqstat(struct obd_device *dev)
581 return lprocfs_obd_seq_create(dev, "brw_stats", 0444,
582 &filter_brw_stats_fops, dev);
585 void lprocfs_filter_init_vars(struct lprocfs_static_vars *lvars)
587 lvars->module_vars = lprocfs_filter_module_vars;
588 lvars->obd_vars = lprocfs_filter_obd_vars;
591 static int filter_per_nid_stats_seq_show(struct seq_file *seq, void *v)
593 nid_stat_t * stat = seq->private;
595 if (stat->nid_brw_stats)
596 brw_stats_show(seq, stat->nid_brw_stats);
601 static ssize_t filter_per_nid_stats_seq_write(struct file *file,
602 const char *buf, size_t len,
605 struct seq_file *seq = file->private_data;
606 nid_stat_t *stat = seq->private;
609 if (stat->nid_brw_stats)
610 for (i = 0; i < BRW_LAST; i++)
611 lprocfs_oh_clear(&stat->nid_brw_stats->hist[i]);
616 LPROC_SEQ_FOPS(filter_per_nid_stats);