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.filter.fo_vfsmnt->mnt_devname);
93 return snprintf(page, count, "%s\n",
94 obd->u.filter.fo_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;
107 for (i = FILTER_GROUP_MDS0; i < filter->fo_group_count; i++) {
108 rc = snprintf(page, count, LPU64"\n",filter_last_id(filter, i));
120 int lprocfs_filter_rd_readcache(char *page, char **start, off_t off, int count,
121 int *eof, void *data)
123 struct obd_device *obd = data;
126 rc = snprintf(page, count, LPU64"\n",
127 obd->u.filter.fo_readcache_max_filesize);
131 int lprocfs_filter_wr_readcache(struct file *file, const char *buffer,
132 unsigned long count, void *data)
134 struct obd_device *obd = data;
138 rc = lprocfs_write_u64_helper(buffer, count, &val);
142 obd->u.filter.fo_readcache_max_filesize = val;
147 int lprocfs_filter_rd_fmd_max_num(char *page, char **start, off_t off,
148 int count, int *eof, void *data)
150 struct obd_device *obd = data;
153 rc = snprintf(page, count, "%u\n", obd->u.filter.fo_fmd_max_num);
157 int lprocfs_filter_wr_fmd_max_num(struct file *file, const char *buffer,
158 unsigned long count, void *data)
160 struct obd_device *obd = data;
164 rc = lprocfs_write_helper(buffer, count, &val);
168 if (val > 65536 || val < 1)
171 obd->u.filter.fo_fmd_max_num = val;
175 int lprocfs_filter_rd_fmd_max_age(char *page, char **start, off_t off,
176 int count, int *eof, void *data)
178 struct obd_device *obd = data;
181 rc = snprintf(page, count, "%u\n", obd->u.filter.fo_fmd_max_age / HZ);
185 int lprocfs_filter_wr_fmd_max_age(struct file *file, const char *buffer,
186 unsigned long count, void *data)
188 struct obd_device *obd = data;
192 rc = lprocfs_write_helper(buffer, count, &val);
196 if (val > 65536 || val < 1)
199 obd->u.filter.fo_fmd_max_age = val * HZ;
203 static int lprocfs_filter_rd_capa(char *page, char **start, off_t off,
204 int count, int *eof, void *data)
206 struct obd_device *obd = data;
209 rc = snprintf(page, count, "capability on: %s\n",
210 obd->u.filter.fo_fl_oss_capa ? "oss" : "");
214 static int lprocfs_filter_wr_capa(struct file *file, const char *buffer,
215 unsigned long count, void *data)
217 struct obd_device *obd = data;
220 rc = lprocfs_write_helper(buffer, count, &val);
225 CERROR("invalid capability mode, only 0/1 are accepted.\n"
226 " 1: enable oss fid capability\n"
227 " 0: disable oss fid capability\n");
231 obd->u.filter.fo_fl_oss_capa = val;
232 LCONSOLE_INFO("OSS %s %s fid capability.\n", obd->obd_name,
233 val ? "enabled" : "disabled");
237 static int lprocfs_filter_rd_capa_count(char *page, char **start, off_t off,
238 int count, int *eof, void *data)
240 return snprintf(page, count, "%d %d\n",
241 capa_count[CAPA_SITE_CLIENT],
242 capa_count[CAPA_SITE_SERVER]);
245 static int lprocfs_filter_rd_cache(char *page, char **start, off_t off,
246 int count, int *eof, void *data)
248 struct obd_device *obd = (struct obd_device *)data;
249 LASSERT(obd != NULL);
251 return snprintf(page, count, "%u\n", obd->u.filter.fo_read_cache);
254 static int lprocfs_filter_wr_cache(struct file *file, const char *buffer,
255 unsigned long count, void *data)
257 struct obd_device *obd = (struct obd_device *)data;
259 LASSERT(obd != NULL);
261 rc = lprocfs_write_helper(buffer, count, &val);
266 obd->u.filter.fo_read_cache = val;
270 static int lprocfs_filter_rd_wcache(char *page, char **start, off_t off,
271 int count, int *eof, void *data)
273 struct obd_device *obd = (struct obd_device *)data;
274 LASSERT(obd != NULL);
276 return snprintf(page, count, "%u\n", obd->u.filter.fo_writethrough_cache);
279 static int lprocfs_filter_wr_wcache(struct file *file, const char *buffer,
280 unsigned long count, void *data)
282 struct obd_device *obd = (struct obd_device *)data;
284 LASSERT(obd != NULL);
286 rc = lprocfs_write_helper(buffer, count, &val);
291 obd->u.filter.fo_writethrough_cache = val;
295 static struct lprocfs_vars lprocfs_filter_obd_vars[] = {
296 { "uuid", lprocfs_rd_uuid, 0, 0 },
297 { "blocksize", lprocfs_rd_blksize, 0, 0 },
298 { "kbytestotal", lprocfs_rd_kbytestotal, 0, 0 },
299 { "kbytesfree", lprocfs_rd_kbytesfree, 0, 0 },
300 { "kbytesavail", lprocfs_rd_kbytesavail, 0, 0 },
301 { "filestotal", lprocfs_rd_filestotal, 0, 0 },
302 { "filesfree", lprocfs_rd_filesfree, 0, 0 },
303 { "filegroups", lprocfs_filter_rd_groups, 0, 0 },
304 { "fstype", lprocfs_rd_fstype, 0, 0 },
305 { "mntdev", lprocfs_filter_rd_mntdev, 0, 0 },
306 { "last_id", lprocfs_filter_rd_last_id,0, 0 },
307 { "tot_dirty", lprocfs_filter_rd_tot_dirty, 0, 0 },
308 { "tot_pending", lprocfs_filter_rd_tot_pending, 0, 0 },
309 { "tot_granted", lprocfs_filter_rd_tot_granted, 0, 0 },
310 { "hash_stats", lprocfs_obd_rd_hash, 0, 0 },
311 { "recovery_status", lprocfs_obd_rd_recovery_status, 0, 0 },
312 { "recovery_maxtime", lprocfs_obd_rd_recovery_maxtime,
313 lprocfs_obd_wr_recovery_maxtime, 0},
314 { "evict_client", 0, lprocfs_wr_evict_client, 0,
315 &lprocfs_evict_client_fops},
316 { "num_exports", lprocfs_rd_num_exports, 0, 0 },
317 { "readcache_max_filesize",
318 lprocfs_filter_rd_readcache,
319 lprocfs_filter_wr_readcache, 0 },
320 #ifdef HAVE_QUOTA_SUPPORT
321 { "quota_bunit_sz", lprocfs_rd_bunit, lprocfs_wr_bunit, 0},
322 { "quota_btune_sz", lprocfs_rd_btune, lprocfs_wr_btune, 0},
323 { "quota_iunit_sz", lprocfs_rd_iunit, lprocfs_wr_iunit, 0},
324 { "quota_itune_sz", lprocfs_rd_itune, lprocfs_wr_itune, 0},
325 { "quota_type", lprocfs_rd_type, lprocfs_wr_type, 0},
327 { "client_cache_count", lprocfs_filter_rd_fmd_max_num,
328 lprocfs_filter_wr_fmd_max_num, 0 },
329 { "client_cache_seconds", lprocfs_filter_rd_fmd_max_age,
330 lprocfs_filter_wr_fmd_max_age, 0 },
331 { "capa", lprocfs_filter_rd_capa,
332 lprocfs_filter_wr_capa, 0 },
333 { "capa_count", lprocfs_filter_rd_capa_count, 0, 0 },
334 { "read_cache_enable", lprocfs_filter_rd_cache, lprocfs_filter_wr_cache, 0},
335 { "writethrough_cache_enable", lprocfs_filter_rd_wcache,
336 lprocfs_filter_wr_wcache, 0},
340 static struct lprocfs_vars lprocfs_filter_module_vars[] = {
341 { "num_refs", lprocfs_rd_numrefs, 0, 0 },
345 void filter_tally(struct obd_export *exp, struct page **pages, int nr_pages,
346 unsigned long *blocks, int blocks_per_page, int wr)
348 struct filter_obd *filter = &exp->exp_obd->u.filter;
349 struct filter_export_data *fed = &exp->exp_filter_data;
350 struct page *last_page = NULL;
351 unsigned long *last_block = NULL;
352 unsigned long discont_pages = 0;
353 unsigned long discont_blocks = 0;
359 lprocfs_oh_tally_log2(&filter->fo_filter_stats.hist[BRW_R_PAGES + wr],
361 lprocfs_oh_tally_log2(&fed->fed_brw_stats.hist[BRW_R_PAGES + wr],
363 if (exp->exp_nid_stats && exp->exp_nid_stats->nid_brw_stats)
364 lprocfs_oh_tally_log2(&exp->exp_nid_stats->nid_brw_stats->
365 hist[BRW_R_PAGES + wr], nr_pages);
367 while (nr_pages-- > 0) {
368 if (last_page && (*pages)->index != (last_page->index + 1))
372 for (i = 0; i < blocks_per_page; i++) {
373 if (last_block && *blocks != (*last_block + 1))
375 last_block = blocks++;
379 lprocfs_oh_tally(&filter->fo_filter_stats.hist[BRW_R_DISCONT_PAGES +wr],
381 lprocfs_oh_tally(&fed->fed_brw_stats.hist[BRW_R_DISCONT_PAGES + wr],
383 lprocfs_oh_tally(&filter->fo_filter_stats.hist[BRW_R_DISCONT_BLOCKS+wr],
385 lprocfs_oh_tally(&fed->fed_brw_stats.hist[BRW_R_DISCONT_BLOCKS + wr],
388 if (exp->exp_nid_stats && exp->exp_nid_stats->nid_brw_stats) {
389 lprocfs_oh_tally_log2(&exp->exp_nid_stats->nid_brw_stats->
390 hist[BRW_R_DISCONT_PAGES + wr],
392 lprocfs_oh_tally_log2(&exp->exp_nid_stats->nid_brw_stats->
393 hist[BRW_R_DISCONT_BLOCKS + wr],
398 #define pct(a,b) (b ? a * 100 / b : 0)
400 static void display_brw_stats(struct seq_file *seq, char *name, char *units,
401 struct obd_histogram *read, struct obd_histogram *write, int log2)
403 unsigned long read_tot, write_tot, r, w, read_cum = 0, write_cum = 0;
406 seq_printf(seq, "\n%26s read | write\n", " ");
407 seq_printf(seq, "%-22s %-5s %% cum %% | %-5s %% cum %%\n",
410 read_tot = lprocfs_oh_sum(read);
411 write_tot = lprocfs_oh_sum(write);
412 for (i = 0; i < OBD_HIST_MAX; i++) {
413 r = read->oh_buckets[i];
414 w = write->oh_buckets[i];
417 if (read_cum == 0 && write_cum == 0)
421 seq_printf(seq, "%u", i);
423 seq_printf(seq, "%u", 1<<i);
425 seq_printf(seq, "%uK", 1<<(i-10));
427 seq_printf(seq, "%uM", 1<<(i-20));
429 seq_printf(seq, ":\t\t%10lu %3lu %3lu | %4lu %3lu %3lu\n",
430 r, pct(r, read_tot), pct(read_cum, read_tot),
431 w, pct(w, write_tot), pct(write_cum, write_tot));
433 if (read_cum == read_tot && write_cum == write_tot)
438 static void brw_stats_show(struct seq_file *seq, struct brw_stats *brw_stats)
442 /* this sampling races with updates */
443 do_gettimeofday(&now);
444 seq_printf(seq, "snapshot_time: %lu.%lu (secs.usecs)\n",
445 now.tv_sec, now.tv_usec);
447 display_brw_stats(seq, "pages per bulk r/w", "rpcs",
448 &brw_stats->hist[BRW_R_PAGES],
449 &brw_stats->hist[BRW_W_PAGES], 1);
451 display_brw_stats(seq, "discontiguous pages", "rpcs",
452 &brw_stats->hist[BRW_R_DISCONT_PAGES],
453 &brw_stats->hist[BRW_W_DISCONT_PAGES], 0);
455 display_brw_stats(seq, "discontiguous blocks", "rpcs",
456 &brw_stats->hist[BRW_R_DISCONT_BLOCKS],
457 &brw_stats->hist[BRW_W_DISCONT_BLOCKS], 0);
459 display_brw_stats(seq, "disk fragmented I/Os", "ios",
460 &brw_stats->hist[BRW_R_DIO_FRAGS],
461 &brw_stats->hist[BRW_W_DIO_FRAGS], 0);
463 display_brw_stats(seq, "disk I/Os in flight", "ios",
464 &brw_stats->hist[BRW_R_RPC_HIST],
465 &brw_stats->hist[BRW_W_RPC_HIST], 0);
469 sprintf(title, "I/O time (1/%ds)", HZ);
470 display_brw_stats(seq, title, "ios",
471 &brw_stats->hist[BRW_R_IO_TIME],
472 &brw_stats->hist[BRW_W_IO_TIME], 1);
475 display_brw_stats(seq, "disk I/O size", "ios",
476 &brw_stats->hist[BRW_R_DISK_IOSIZE],
477 &brw_stats->hist[BRW_W_DISK_IOSIZE], 1);
482 static int filter_brw_stats_seq_show(struct seq_file *seq, void *v)
484 struct obd_device *dev = seq->private;
485 struct filter_obd *filter = &dev->u.filter;
487 brw_stats_show(seq, &filter->fo_filter_stats);
492 static ssize_t filter_brw_stats_seq_write(struct file *file, const char *buf,
493 size_t len, loff_t *off)
495 struct seq_file *seq = file->private_data;
496 struct obd_device *dev = seq->private;
497 struct filter_obd *filter = &dev->u.filter;
500 for (i = 0; i < BRW_LAST; i++)
501 lprocfs_oh_clear(&filter->fo_filter_stats.hist[i]);
506 LPROC_SEQ_FOPS(filter_brw_stats);
508 int lproc_filter_attach_seqstat(struct obd_device *dev)
510 return lprocfs_obd_seq_create(dev, "brw_stats", 0444,
511 &filter_brw_stats_fops, dev);
514 static int filter_per_export_stats_seq_show(struct seq_file *seq, void *v)
516 struct filter_export_data *fed = seq->private;
518 brw_stats_show(seq, &fed->fed_brw_stats);
523 static ssize_t filter_per_export_stats_seq_write(struct file *file,
524 const char *buf, size_t len, loff_t *off)
526 struct seq_file *seq = file->private_data;
527 struct filter_export_data *fed = seq->private;
530 for (i = 0; i < BRW_LAST; i++)
531 lprocfs_oh_clear(&fed->fed_brw_stats.hist[i]);
536 LPROC_SEQ_FOPS(filter_per_export_stats);
538 void lprocfs_filter_init_vars(struct lprocfs_static_vars *lvars)
540 lvars->module_vars = lprocfs_filter_module_vars;
541 lvars->obd_vars = lprocfs_filter_obd_vars;
544 static int filter_per_nid_stats_seq_show(struct seq_file *seq, void *v)
546 nid_stat_t *tmp = seq->private;
548 if (tmp->nid_brw_stats)
549 brw_stats_show(seq, tmp->nid_brw_stats);
554 static ssize_t filter_per_nid_stats_seq_write(struct file *file,
555 const char *buf, size_t len,
558 struct seq_file *seq = file->private_data;
559 nid_stat_t *tmp = seq->private;
562 if (tmp->nid_brw_stats)
563 for (i = 0; i < BRW_LAST; i++)
564 lprocfs_oh_clear(&tmp->nid_brw_stats->hist[i]);
569 LPROC_SEQ_FOPS(filter_per_nid_stats);