4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
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.
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).
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.gnu.org/licenses/gpl-2.0.html
23 * Copyright (c) 2008, 2010, Oracle and/or its affiliates. All rights reserved.
24 * Use is subject to license terms.
26 * Copyright (c) 2012, 2016, Intel Corporation.
29 * This file is part of Lustre, http://www.lustre.org/
30 * Lustre is a trademark of Sun Microsystems, Inc.
32 * lustre/osd-zfs/osd_lproc.c
34 * Author: Alex Zhuravlev <bzzz@whamcloud.com>
35 * Author: Mike Pershin <tappro@whamcloud.com>
38 #define DEBUG_SUBSYSTEM S_OSD
41 #include <obd_class.h>
42 #include <lprocfs_status.h>
44 #include "osd_internal.h"
48 #define pct(a, b) (b ? a * 100 / b : 0)
50 static void display_brw_stats(struct seq_file *seq, char *name, char *units,
51 struct obd_histogram *read,
52 struct obd_histogram *write, int scale)
54 unsigned long read_tot, write_tot, r, w, read_cum = 0, write_cum = 0;
57 seq_printf(seq, "\n%26s read | write\n", " ");
58 seq_printf(seq, "%-22s %-5s %% cum %% | %-11s %% cum %%\n",
61 read_tot = lprocfs_oh_sum(read);
62 write_tot = lprocfs_oh_sum(write);
63 for (i = 0; i < OBD_HIST_MAX; i++) {
64 r = read->oh_buckets[i];
65 w = write->oh_buckets[i];
68 if (read_cum == 0 && write_cum == 0)
72 seq_printf(seq, "%u", i);
74 seq_printf(seq, "%u", scale << i);
76 seq_printf(seq, "%uK", scale << (i-10));
78 seq_printf(seq, "%uM", scale << (i-20));
80 seq_printf(seq, ":\t\t%10lu %3lu %3lu | %4lu %3lu %3lu\n",
81 r, pct(r, read_tot), pct(read_cum, read_tot),
82 w, pct(w, write_tot), pct(write_cum, write_tot));
84 if (read_cum == read_tot && write_cum == write_tot)
89 static void brw_stats_show(struct seq_file *seq, struct brw_stats *brw_stats)
91 struct timespec64 now;
93 /* this sampling races with updates */
94 ktime_get_real_ts64(&now);
95 seq_printf(seq, "snapshot_time: %llu.%09lu (secs.nsecs)\n",
96 (s64)now.tv_sec, now.tv_nsec);
98 display_brw_stats(seq, "pages per bulk r/w", "rpcs",
99 &brw_stats->hist[BRW_R_PAGES],
100 &brw_stats->hist[BRW_W_PAGES], 1);
101 display_brw_stats(seq, "discontiguous pages", "rpcs",
102 &brw_stats->hist[BRW_R_DISCONT_PAGES],
103 &brw_stats->hist[BRW_W_DISCONT_PAGES], 0);
105 display_brw_stats(seq, "discontiguous blocks", "rpcs",
106 &brw_stats->hist[BRW_R_DISCONT_BLOCKS],
107 &brw_stats->hist[BRW_W_DISCONT_BLOCKS], 0);
109 display_brw_stats(seq, "disk fragmented I/Os", "ios",
110 &brw_stats->hist[BRW_R_DIO_FRAGS],
111 &brw_stats->hist[BRW_W_DIO_FRAGS], 0);
113 display_brw_stats(seq, "disk I/Os in flight", "ios",
114 &brw_stats->hist[BRW_R_RPC_HIST],
115 &brw_stats->hist[BRW_W_RPC_HIST], 0);
117 display_brw_stats(seq, "I/O time (1/1000s)", "ios",
118 &brw_stats->hist[BRW_R_IO_TIME],
119 &brw_stats->hist[BRW_W_IO_TIME], 1000 / HZ);
121 display_brw_stats(seq, "disk I/O size", "ios",
122 &brw_stats->hist[BRW_R_DISK_IOSIZE],
123 &brw_stats->hist[BRW_W_DISK_IOSIZE], 1);
128 static int osd_brw_stats_seq_show(struct seq_file *seq, void *v)
130 struct osd_device *osd = seq->private;
132 brw_stats_show(seq, &osd->od_brw_stats);
137 static ssize_t osd_brw_stats_seq_write(struct file *file,
138 const char __user *buf,
139 size_t len, loff_t *off)
141 struct seq_file *seq = file->private_data;
142 struct osd_device *osd = seq->private;
145 for (i = 0; i < BRW_LAST; i++)
146 lprocfs_oh_clear(&osd->od_brw_stats.hist[i]);
151 LPROC_SEQ_FOPS(osd_brw_stats);
153 static int osd_stats_init(struct osd_device *osd)
158 for (i = 0; i < BRW_LAST; i++)
159 spin_lock_init(&osd->od_brw_stats.hist[i].oh_lock);
161 osd->od_stats = lprocfs_alloc_stats(LPROC_OSD_LAST, 0);
162 if (osd->od_stats != NULL) {
163 result = lprocfs_register_stats(osd->od_proc_entry, "stats",
168 lprocfs_counter_init(osd->od_stats, LPROC_OSD_GET_PAGE,
169 LPROCFS_CNTR_AVGMINMAX|LPROCFS_CNTR_STDDEV,
171 lprocfs_counter_init(osd->od_stats, LPROC_OSD_NO_PAGE,
172 LPROCFS_CNTR_AVGMINMAX,
173 "get_page_failures", "num");
174 lprocfs_counter_init(osd->od_stats, LPROC_OSD_CACHE_ACCESS,
175 LPROCFS_CNTR_AVGMINMAX,
176 "cache_access", "pages");
177 lprocfs_counter_init(osd->od_stats, LPROC_OSD_CACHE_HIT,
178 LPROCFS_CNTR_AVGMINMAX,
179 "cache_hit", "pages");
180 lprocfs_counter_init(osd->od_stats, LPROC_OSD_CACHE_MISS,
181 LPROCFS_CNTR_AVGMINMAX,
182 "cache_miss", "pages");
183 lprocfs_counter_init(osd->od_stats, LPROC_OSD_COPY_IO,
184 LPROCFS_CNTR_AVGMINMAX,
186 lprocfs_counter_init(osd->od_stats, LPROC_OSD_ZEROCOPY_IO,
187 LPROCFS_CNTR_AVGMINMAX,
188 "zerocopy", "pages");
189 lprocfs_counter_init(osd->od_stats, LPROC_OSD_TAIL_IO,
190 LPROCFS_CNTR_AVGMINMAX,
192 #ifdef OSD_THANDLE_STATS
193 lprocfs_counter_init(osd->od_stats, LPROC_OSD_THANDLE_STARTING,
194 LPROCFS_CNTR_AVGMINMAX,
195 "thandle_starting", "usec");
196 lprocfs_counter_init(osd->od_stats, LPROC_OSD_THANDLE_OPEN,
197 LPROCFS_CNTR_AVGMINMAX,
198 "thandle_open", "usec");
199 lprocfs_counter_init(osd->od_stats, LPROC_OSD_THANDLE_CLOSING,
200 LPROCFS_CNTR_AVGMINMAX,
201 "thandle_closing", "usec");
203 result = lprocfs_seq_create(osd->od_proc_entry, "brw_stats",
204 0644, &osd_brw_stats_fops, osd);
213 static int zfs_osd_fstype_seq_show(struct seq_file *m, void *data)
215 seq_puts(m, "zfs\n");
218 LPROC_SEQ_FOPS_RO(zfs_osd_fstype);
220 static int zfs_osd_mntdev_seq_show(struct seq_file *m, void *data)
222 struct osd_device *osd = osd_dt_dev((struct dt_device *)m->private);
224 LASSERT(osd != NULL);
225 seq_printf(m, "%s\n", osd->od_mntdev);
228 LPROC_SEQ_FOPS_RO(zfs_osd_mntdev);
231 lprocfs_osd_force_sync_seq_write(struct file *file, const char __user *buffer,
232 size_t count, loff_t *off)
234 struct seq_file *m = file->private_data;
235 struct dt_device *dt = m->private;
239 rc = lu_env_init(&env, LCT_LOCAL);
242 rc = dt_sync(&env, dt);
245 return rc == 0 ? count : rc;
247 LPROC_SEQ_FOPS_WR_ONLY(zfs, osd_force_sync);
249 LPROC_SEQ_FOPS_RO_TYPE(zfs, dt_blksize);
250 LPROC_SEQ_FOPS_RO_TYPE(zfs, dt_kbytestotal);
251 LPROC_SEQ_FOPS_RO_TYPE(zfs, dt_kbytesfree);
252 LPROC_SEQ_FOPS_RO_TYPE(zfs, dt_kbytesavail);
253 LPROC_SEQ_FOPS_RO_TYPE(zfs, dt_filestotal);
254 LPROC_SEQ_FOPS_RO_TYPE(zfs, dt_filesfree);
256 struct lprocfs_vars lprocfs_osd_obd_vars[] = {
257 { .name = "blocksize",
258 .fops = &zfs_dt_blksize_fops },
259 { .name = "kbytestotal",
260 .fops = &zfs_dt_kbytestotal_fops },
261 { .name = "kbytesfree",
262 .fops = &zfs_dt_kbytesfree_fops },
263 { .name = "kbytesavail",
264 .fops = &zfs_dt_kbytesavail_fops },
265 { .name = "filestotal",
266 .fops = &zfs_dt_filestotal_fops },
267 { .name = "filesfree",
268 .fops = &zfs_dt_filesfree_fops },
270 .fops = &zfs_osd_fstype_fops },
272 .fops = &zfs_osd_mntdev_fops },
273 { .name = "force_sync",
274 .fops = &zfs_osd_force_sync_fops },
278 int osd_procfs_init(struct osd_device *osd, const char *name)
280 struct obd_type *type;
284 if (osd->od_proc_entry)
287 /* at the moment there is no linkage between lu_type
288 * and obd_type, so we lookup obd_type this way */
289 type = class_search_type(LUSTRE_OSD_ZFS_NAME);
291 LASSERT(name != NULL);
292 LASSERT(type != NULL);
294 osd->od_proc_entry = lprocfs_register(name, type->typ_procroot,
295 lprocfs_osd_obd_vars,
297 if (IS_ERR(osd->od_proc_entry)) {
298 rc = PTR_ERR(osd->od_proc_entry);
299 CERROR("Error %d setting up lprocfs for %s\n", rc, name);
300 osd->od_proc_entry = NULL;
304 rc = osd_stats_init(osd);
309 osd_procfs_fini(osd);
313 int osd_procfs_fini(struct osd_device *osd)
318 lprocfs_free_stats(&osd->od_stats);
320 if (osd->od_proc_entry) {
321 lprocfs_remove(&osd->od_proc_entry);
322 osd->od_proc_entry = NULL;