1 /* 2 drbd_proc.c 3 4 This file is part of DRBD by Philipp Reisner and Lars Ellenberg. 5 6 Copyright (C) 2001-2008, LINBIT Information Technologies GmbH. 7 Copyright (C) 1999-2008, Philipp Reisner <philipp.reisner@linbit.com>. 8 Copyright (C) 2002-2008, Lars Ellenberg <lars.ellenberg@linbit.com>. 9 10 drbd is free software; you can redistribute it and/or modify 11 it under the terms of the GNU General Public License as published by 12 the Free Software Foundation; either version 2, or (at your option) 13 any later version. 14 15 drbd is distributed in the hope that it will be useful, 16 but WITHOUT ANY WARRANTY; without even the implied warranty of 17 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 18 GNU General Public License for more details. 19 20 You should have received a copy of the GNU General Public License 21 along with drbd; see the file COPYING. If not, write to 22 the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139, USA. 23 24 */ 25 26 #include <linux/module.h> 27 28 #include <asm/uaccess.h> 29 #include <linux/fs.h> 30 #include <linux/file.h> 31 #include <linux/proc_fs.h> 32 #include <linux/seq_file.h> 33 #include <linux/drbd.h> 34 #include "drbd_int.h" 35 36 static int drbd_proc_open(struct inode *inode, struct file *file); 37 static int drbd_proc_release(struct inode *inode, struct file *file); 38 39 40 struct proc_dir_entry *drbd_proc; 41 const struct file_operations drbd_proc_fops = { 42 .owner = THIS_MODULE, 43 .open = drbd_proc_open, 44 .read = seq_read, 45 .llseek = seq_lseek, 46 .release = drbd_proc_release, 47 }; 48 49 static void seq_printf_with_thousands_grouping(struct seq_file *seq, long v) 50 { 51 /* v is in kB/sec. We don't expect TiByte/sec yet. */ 52 if (unlikely(v >= 1000000)) { 53 /* cool: > GiByte/s */ 54 seq_printf(seq, "%ld,", v / 1000000); 55 v %= 1000000; 56 seq_printf(seq, "%03ld,%03ld", v/1000, v % 1000); 57 } else if (likely(v >= 1000)) 58 seq_printf(seq, "%ld,%03ld", v/1000, v % 1000); 59 else 60 seq_printf(seq, "%ld", v); 61 } 62 63 static void drbd_get_syncer_progress(struct drbd_device *device, 64 union drbd_dev_state state, unsigned long *rs_total, 65 unsigned long *bits_left, unsigned int *per_mil_done) 66 { 67 /* this is to break it at compile time when we change that, in case we 68 * want to support more than (1<<32) bits on a 32bit arch. */ 69 typecheck(unsigned long, device->rs_total); 70 *rs_total = device->rs_total; 71 72 /* note: both rs_total and rs_left are in bits, i.e. in 73 * units of BM_BLOCK_SIZE. 74 * for the percentage, we don't care. */ 75 76 if (state.conn == C_VERIFY_S || state.conn == C_VERIFY_T) 77 *bits_left = device->ov_left; 78 else 79 *bits_left = drbd_bm_total_weight(device) - device->rs_failed; 80 /* >> 10 to prevent overflow, 81 * +1 to prevent division by zero */ 82 if (*bits_left > *rs_total) { 83 /* D'oh. Maybe a logic bug somewhere. More likely just a race 84 * between state change and reset of rs_total. 85 */ 86 *bits_left = *rs_total; 87 *per_mil_done = *rs_total ? 0 : 1000; 88 } else { 89 /* Make sure the division happens in long context. 90 * We allow up to one petabyte storage right now, 91 * at a granularity of 4k per bit that is 2**38 bits. 92 * After shift right and multiplication by 1000, 93 * this should still fit easily into a 32bit long, 94 * so we don't need a 64bit division on 32bit arch. 95 * Note: currently we don't support such large bitmaps on 32bit 96 * arch anyways, but no harm done to be prepared for it here. 97 */ 98 unsigned int shift = *rs_total > UINT_MAX ? 16 : 10; 99 unsigned long left = *bits_left >> shift; 100 unsigned long total = 1UL + (*rs_total >> shift); 101 unsigned long tmp = 1000UL - left * 1000UL/total; 102 *per_mil_done = tmp; 103 } 104 } 105 106 107 /*lge 108 * progress bars shamelessly adapted from driver/md/md.c 109 * output looks like 110 * [=====>..............] 33.5% (23456/123456) 111 * finish: 2:20:20 speed: 6,345 (6,456) K/sec 112 */ 113 static void drbd_syncer_progress(struct drbd_device *device, struct seq_file *seq, 114 union drbd_dev_state state) 115 { 116 unsigned long db, dt, dbdt, rt, rs_total, rs_left; 117 unsigned int res; 118 int i, x, y; 119 int stalled = 0; 120 121 drbd_get_syncer_progress(device, state, &rs_total, &rs_left, &res); 122 123 x = res/50; 124 y = 20-x; 125 seq_printf(seq, "\t["); 126 for (i = 1; i < x; i++) 127 seq_printf(seq, "="); 128 seq_printf(seq, ">"); 129 for (i = 0; i < y; i++) 130 seq_printf(seq, "."); 131 seq_printf(seq, "] "); 132 133 if (state.conn == C_VERIFY_S || state.conn == C_VERIFY_T) 134 seq_printf(seq, "verified:"); 135 else 136 seq_printf(seq, "sync'ed:"); 137 seq_printf(seq, "%3u.%u%% ", res / 10, res % 10); 138 139 /* if more than a few GB, display in MB */ 140 if (rs_total > (4UL << (30 - BM_BLOCK_SHIFT))) 141 seq_printf(seq, "(%lu/%lu)M", 142 (unsigned long) Bit2KB(rs_left >> 10), 143 (unsigned long) Bit2KB(rs_total >> 10)); 144 else 145 seq_printf(seq, "(%lu/%lu)K\n\t", 146 (unsigned long) Bit2KB(rs_left), 147 (unsigned long) Bit2KB(rs_total)); 148 149 /* see drivers/md/md.c 150 * We do not want to overflow, so the order of operands and 151 * the * 100 / 100 trick are important. We do a +1 to be 152 * safe against division by zero. We only estimate anyway. 153 * 154 * dt: time from mark until now 155 * db: blocks written from mark until now 156 * rt: remaining time 157 */ 158 /* Rolling marks. last_mark+1 may just now be modified. last_mark+2 is 159 * at least (DRBD_SYNC_MARKS-2)*DRBD_SYNC_MARK_STEP old, and has at 160 * least DRBD_SYNC_MARK_STEP time before it will be modified. */ 161 /* ------------------------ ~18s average ------------------------ */ 162 i = (device->rs_last_mark + 2) % DRBD_SYNC_MARKS; 163 dt = (jiffies - device->rs_mark_time[i]) / HZ; 164 if (dt > 180) 165 stalled = 1; 166 167 if (!dt) 168 dt++; 169 db = device->rs_mark_left[i] - rs_left; 170 rt = (dt * (rs_left / (db/100+1)))/100; /* seconds */ 171 172 seq_printf(seq, "finish: %lu:%02lu:%02lu", 173 rt / 3600, (rt % 3600) / 60, rt % 60); 174 175 dbdt = Bit2KB(db/dt); 176 seq_printf(seq, " speed: "); 177 seq_printf_with_thousands_grouping(seq, dbdt); 178 seq_printf(seq, " ("); 179 /* ------------------------- ~3s average ------------------------ */ 180 if (proc_details >= 1) { 181 /* this is what drbd_rs_should_slow_down() uses */ 182 i = (device->rs_last_mark + DRBD_SYNC_MARKS-1) % DRBD_SYNC_MARKS; 183 dt = (jiffies - device->rs_mark_time[i]) / HZ; 184 if (!dt) 185 dt++; 186 db = device->rs_mark_left[i] - rs_left; 187 dbdt = Bit2KB(db/dt); 188 seq_printf_with_thousands_grouping(seq, dbdt); 189 seq_printf(seq, " -- "); 190 } 191 192 /* --------------------- long term average ---------------------- */ 193 /* mean speed since syncer started 194 * we do account for PausedSync periods */ 195 dt = (jiffies - device->rs_start - device->rs_paused) / HZ; 196 if (dt == 0) 197 dt = 1; 198 db = rs_total - rs_left; 199 dbdt = Bit2KB(db/dt); 200 seq_printf_with_thousands_grouping(seq, dbdt); 201 seq_printf(seq, ")"); 202 203 if (state.conn == C_SYNC_TARGET || 204 state.conn == C_VERIFY_S) { 205 seq_printf(seq, " want: "); 206 seq_printf_with_thousands_grouping(seq, device->c_sync_rate); 207 } 208 seq_printf(seq, " K/sec%s\n", stalled ? " (stalled)" : ""); 209 210 if (proc_details >= 1) { 211 /* 64 bit: 212 * we convert to sectors in the display below. */ 213 unsigned long bm_bits = drbd_bm_bits(device); 214 unsigned long bit_pos; 215 unsigned long long stop_sector = 0; 216 if (state.conn == C_VERIFY_S || 217 state.conn == C_VERIFY_T) { 218 bit_pos = bm_bits - device->ov_left; 219 if (verify_can_do_stop_sector(device)) 220 stop_sector = device->ov_stop_sector; 221 } else 222 bit_pos = device->bm_resync_fo; 223 /* Total sectors may be slightly off for oddly 224 * sized devices. So what. */ 225 seq_printf(seq, 226 "\t%3d%% sector pos: %llu/%llu", 227 (int)(bit_pos / (bm_bits/100+1)), 228 (unsigned long long)bit_pos * BM_SECT_PER_BIT, 229 (unsigned long long)bm_bits * BM_SECT_PER_BIT); 230 if (stop_sector != 0 && stop_sector != ULLONG_MAX) 231 seq_printf(seq, " stop sector: %llu", stop_sector); 232 seq_printf(seq, "\n"); 233 } 234 } 235 236 static int drbd_seq_show(struct seq_file *seq, void *v) 237 { 238 int i, prev_i = -1; 239 const char *sn; 240 struct drbd_device *device; 241 struct net_conf *nc; 242 union drbd_dev_state state; 243 char wp; 244 245 static char write_ordering_chars[] = { 246 [WO_none] = 'n', 247 [WO_drain_io] = 'd', 248 [WO_bdev_flush] = 'f', 249 }; 250 251 seq_printf(seq, "version: " REL_VERSION " (api:%d/proto:%d-%d)\n%s\n", 252 API_VERSION, PRO_VERSION_MIN, PRO_VERSION_MAX, drbd_buildtag()); 253 254 /* 255 cs .. connection state 256 ro .. node role (local/remote) 257 ds .. disk state (local/remote) 258 protocol 259 various flags 260 ns .. network send 261 nr .. network receive 262 dw .. disk write 263 dr .. disk read 264 al .. activity log write count 265 bm .. bitmap update write count 266 pe .. pending (waiting for ack or data reply) 267 ua .. unack'd (still need to send ack or data reply) 268 ap .. application requests accepted, but not yet completed 269 ep .. number of epochs currently "on the fly", P_BARRIER_ACK pending 270 wo .. write ordering mode currently in use 271 oos .. known out-of-sync kB 272 */ 273 274 rcu_read_lock(); 275 idr_for_each_entry(&drbd_devices, device, i) { 276 if (prev_i != i - 1) 277 seq_printf(seq, "\n"); 278 prev_i = i; 279 280 state = device->state; 281 sn = drbd_conn_str(state.conn); 282 283 if (state.conn == C_STANDALONE && 284 state.disk == D_DISKLESS && 285 state.role == R_SECONDARY) { 286 seq_printf(seq, "%2d: cs:Unconfigured\n", i); 287 } else { 288 /* reset device->congestion_reason */ 289 bdi_rw_congested(&device->rq_queue->backing_dev_info); 290 291 nc = rcu_dereference(first_peer_device(device)->connection->net_conf); 292 wp = nc ? nc->wire_protocol - DRBD_PROT_A + 'A' : ' '; 293 seq_printf(seq, 294 "%2d: cs:%s ro:%s/%s ds:%s/%s %c %c%c%c%c%c%c\n" 295 " ns:%u nr:%u dw:%u dr:%u al:%u bm:%u " 296 "lo:%d pe:%d ua:%d ap:%d ep:%d wo:%c", 297 i, sn, 298 drbd_role_str(state.role), 299 drbd_role_str(state.peer), 300 drbd_disk_str(state.disk), 301 drbd_disk_str(state.pdsk), 302 wp, 303 drbd_suspended(device) ? 's' : 'r', 304 state.aftr_isp ? 'a' : '-', 305 state.peer_isp ? 'p' : '-', 306 state.user_isp ? 'u' : '-', 307 device->congestion_reason ?: '-', 308 test_bit(AL_SUSPENDED, &device->flags) ? 's' : '-', 309 device->send_cnt/2, 310 device->recv_cnt/2, 311 device->writ_cnt/2, 312 device->read_cnt/2, 313 device->al_writ_cnt, 314 device->bm_writ_cnt, 315 atomic_read(&device->local_cnt), 316 atomic_read(&device->ap_pending_cnt) + 317 atomic_read(&device->rs_pending_cnt), 318 atomic_read(&device->unacked_cnt), 319 atomic_read(&device->ap_bio_cnt), 320 first_peer_device(device)->connection->epochs, 321 write_ordering_chars[device->resource->write_ordering] 322 ); 323 seq_printf(seq, " oos:%llu\n", 324 Bit2KB((unsigned long long) 325 drbd_bm_total_weight(device))); 326 } 327 if (state.conn == C_SYNC_SOURCE || 328 state.conn == C_SYNC_TARGET || 329 state.conn == C_VERIFY_S || 330 state.conn == C_VERIFY_T) 331 drbd_syncer_progress(device, seq, state); 332 333 if (proc_details >= 1 && get_ldev_if_state(device, D_FAILED)) { 334 lc_seq_printf_stats(seq, device->resync); 335 lc_seq_printf_stats(seq, device->act_log); 336 put_ldev(device); 337 } 338 339 if (proc_details >= 2) 340 seq_printf(seq, "\tblocked on activity log: %d\n", atomic_read(&device->ap_actlog_cnt)); 341 } 342 rcu_read_unlock(); 343 344 return 0; 345 } 346 347 static int drbd_proc_open(struct inode *inode, struct file *file) 348 { 349 int err; 350 351 if (try_module_get(THIS_MODULE)) { 352 err = single_open(file, drbd_seq_show, NULL); 353 if (err) 354 module_put(THIS_MODULE); 355 return err; 356 } 357 return -ENODEV; 358 } 359 360 static int drbd_proc_release(struct inode *inode, struct file *file) 361 { 362 module_put(THIS_MODULE); 363 return single_release(inode, file); 364 } 365 366 /* PROC FS stuff end */ 367