1/*- 2 * SPDX-License-Identifier: BSD-3-Clause 3 * 4 * Copyright (c) 1982, 1986, 1990, 1991, 1993 5 * The Regents of the University of California. All rights reserved. 6 * (c) UNIX System Laboratories, Inc. 7 * All or some portions of this file are derived from material licensed 8 * to the University of California by American Telephone and Telegraph 9 * Co. or Unix System Laboratories, Inc. and are reproduced herein with 10 * the permission of UNIX System Laboratories, Inc. 11 * 12 * Copyright (c) 2002 Networks Associates Technologies, Inc. 13 * All rights reserved. 14 * 15 * Portions of this software were developed for the FreeBSD Project by 16 * ThinkSec AS and NAI Labs, the Security Research Division of Network 17 * Associates, Inc. under DARPA/SPAWAR contract N66001-01-C-8035 18 * ("CBOSS"), as part of the DARPA CHATS research program. 19 * 20 * Redistribution and use in source and binary forms, with or without 21 * modification, are permitted provided that the following conditions 22 * are met: 23 * 1. Redistributions of source code must retain the above copyright 24 * notice, this list of conditions and the following disclaimer. 25 * 2. Redistributions in binary form must reproduce the above copyright 26 * notice, this list of conditions and the following disclaimer in the 27 * documentation and/or other materials provided with the distribution. 28 * 3. Neither the name of the University nor the names of its contributors 29 * may be used to endorse or promote products derived from this software 30 * without specific prior written permission. 31 * 32 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND 33 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 34 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE 35 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE 36 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL 37 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS 38 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 39 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT 40 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY 41 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF 42 * SUCH DAMAGE. 43 */ 44 45#include <sys/cdefs.h> 46__FBSDID("$FreeBSD$"); 47 48#include "opt_stack.h" 49 50#include <sys/param.h> 51#include <sys/cons.h> 52#include <sys/kdb.h> 53#include <sys/lock.h> 54#include <sys/malloc.h> 55#include <sys/mutex.h> 56#include <sys/proc.h> 57#include <sys/resourcevar.h> 58#include <sys/sbuf.h> 59#include <sys/sched.h> 60#include <sys/stack.h> 61#include <sys/sysctl.h> 62#include <sys/systm.h> 63#include <sys/tty.h> 64 65#include <vm/vm.h> 66#include <vm/pmap.h> 67#include <vm/vm_map.h> 68 69/* 70 * Returns 1 if p2 is "better" than p1 71 * 72 * The algorithm for picking the "interesting" process is thus: 73 * 74 * 1) Only foreground processes are eligible - implied. 75 * 2) Runnable processes are favored over anything else. The runner 76 * with the highest cpu utilization is picked (p_estcpu). Ties are 77 * broken by picking the highest pid. 78 * 3) The sleeper with the shortest sleep time is next. With ties, 79 * we pick out just "short-term" sleepers (P_SINTR == 0). 80 * 4) Further ties are broken by picking the highest pid. 81 */ 82 83#define TESTAB(a, b) ((a)<<1 | (b)) 84#define ONLYA 2 85#define ONLYB 1 86#define BOTH 3 87 88static int 89proc_sum(struct proc *p, fixpt_t *estcpup) 90{ 91 struct thread *td; 92 int estcpu; 93 int val; 94 95 val = 0; 96 estcpu = 0; 97 FOREACH_THREAD_IN_PROC(p, td) { 98 thread_lock(td); 99 if (TD_ON_RUNQ(td) || 100 TD_IS_RUNNING(td)) 101 val = 1; 102 estcpu += sched_pctcpu(td); 103 thread_unlock(td); 104 } 105 *estcpup = estcpu; 106 107 return (val); 108} 109 110static int 111thread_compare(struct thread *td, struct thread *td2) 112{ 113 int runa, runb; 114 int slpa, slpb; 115 fixpt_t esta, estb; 116 117 if (td == NULL) 118 return (1); 119 120 /* 121 * Fetch running stats, pctcpu usage, and interruptable flag. 122 */ 123 thread_lock(td); 124 runa = TD_IS_RUNNING(td) | TD_ON_RUNQ(td); 125 slpa = td->td_flags & TDF_SINTR; 126 esta = sched_pctcpu(td); 127 thread_unlock(td); 128 thread_lock(td2); 129 runb = TD_IS_RUNNING(td2) | TD_ON_RUNQ(td2); 130 estb = sched_pctcpu(td2); 131 slpb = td2->td_flags & TDF_SINTR; 132 thread_unlock(td2); 133 /* 134 * see if at least one of them is runnable 135 */ 136 switch (TESTAB(runa, runb)) { 137 case ONLYA: 138 return (0); 139 case ONLYB: 140 return (1); 141 case BOTH: 142 break; 143 } 144 /* 145 * favor one with highest recent cpu utilization 146 */ 147 if (estb > esta) 148 return (1); 149 if (esta > estb) 150 return (0); 151 /* 152 * favor one sleeping in a non-interruptible sleep 153 */ 154 switch (TESTAB(slpa, slpb)) { 155 case ONLYA: 156 return (0); 157 case ONLYB: 158 return (1); 159 case BOTH: 160 break; 161 } 162 163 return (td < td2); 164} 165 166static int 167proc_compare(struct proc *p1, struct proc *p2) 168{ 169 170 int runa, runb; 171 fixpt_t esta, estb; 172 173 if (p1 == NULL) 174 return (1); 175 176 /* 177 * Fetch various stats about these processes. After we drop the 178 * lock the information could be stale but the race is unimportant. 179 */ 180 PROC_LOCK(p1); 181 runa = proc_sum(p1, &esta); 182 PROC_UNLOCK(p1); 183 PROC_LOCK(p2); 184 runb = proc_sum(p2, &estb); 185 PROC_UNLOCK(p2); 186 187 /* 188 * see if at least one of them is runnable 189 */ 190 switch (TESTAB(runa, runb)) { 191 case ONLYA: 192 return (0); 193 case ONLYB: 194 return (1); 195 case BOTH: 196 break; 197 } 198 /* 199 * favor one with highest recent cpu utilization 200 */ 201 if (estb > esta) 202 return (1); 203 if (esta > estb) 204 return (0); 205 /* 206 * weed out zombies 207 */ 208 switch (TESTAB(p1->p_state == PRS_ZOMBIE, p2->p_state == PRS_ZOMBIE)) { 209 case ONLYA: 210 return (1); 211 case ONLYB: 212 return (0); 213 case BOTH: 214 break; 215 } 216 217 return (p2->p_pid > p1->p_pid); /* tie - return highest pid */ 218} 219 220static int 221sbuf_tty_drain(void *a, const char *d, int len) 222{ 223 struct tty *tp; 224 int rc; 225 226 tp = a; 227 228 if (kdb_active) { 229 cnputsn(d, len); 230 return (len); 231 } 232 if (tp != NULL && !KERNEL_PANICKED()) { 233 rc = tty_putstrn(tp, d, len); 234 if (rc != 0) 235 return (-ENXIO); 236 return (len); 237 } 238 return (-ENXIO); 239} 240 241#ifdef STACK 242static int tty_info_kstacks = STACK_SBUF_FMT_COMPACT; 243 244static int 245sysctl_tty_info_kstacks(SYSCTL_HANDLER_ARGS) 246{ 247 enum stack_sbuf_fmt val; 248 int error; 249 250 val = tty_info_kstacks; 251 error = sysctl_handle_int(oidp, &val, 0, req); 252 if (error != 0 || req->newptr == NULL) 253 return (error); 254 255 switch (val) { 256 case STACK_SBUF_FMT_NONE: 257 case STACK_SBUF_FMT_LONG: 258 case STACK_SBUF_FMT_COMPACT: 259 tty_info_kstacks = val; 260 break; 261 default: 262 error = EINVAL; 263 } 264 265 return (error); 266} 267SYSCTL_PROC(_kern, OID_AUTO, tty_info_kstacks, 268 CTLFLAG_RWTUN | CTLFLAG_MPSAFE | CTLTYPE_INT, NULL, 0, 269 sysctl_tty_info_kstacks, "I", 270 "Adjust format of kernel stack(9) traces on ^T (tty info): " 271 "0 - disabled; 1 - long; 2 - compact"); 272#endif 273 274/* 275 * Report on state of foreground process group. 276 */ 277void 278tty_info(struct tty *tp) 279{ 280 struct timeval rtime, utime, stime; 281#ifdef STACK 282 struct stack stack; 283 int sterr, kstacks_val; 284 bool print_kstacks; 285#endif 286 struct proc *p, *ppick; 287 struct thread *td, *tdpick; 288 const char *stateprefix, *state; 289 struct sbuf sb; 290 long rss; 291 int load, pctcpu; 292 pid_t pid; 293 char comm[MAXCOMLEN + 1]; 294 struct rusage ru; 295 296 tty_assert_locked(tp); 297 298 if (tty_checkoutq(tp) == 0) 299 return; 300 301 (void)sbuf_new(&sb, tp->t_prbuf, tp->t_prbufsz, SBUF_FIXEDLEN); 302 sbuf_set_drain(&sb, sbuf_tty_drain, tp); 303 304 /* Print load average. */ 305 load = (averunnable.ldavg[0] * 100 + FSCALE / 2) >> FSHIFT; 306 sbuf_printf(&sb, "%sload: %d.%02d ", tp->t_column == 0 ? "" : "\n", 307 load / 100, load % 100); 308 309 if (tp->t_session == NULL) { 310 sbuf_printf(&sb, "not a controlling terminal\n"); 311 goto out; 312 } 313 if (tp->t_pgrp == NULL) { 314 sbuf_printf(&sb, "no foreground process group\n"); 315 goto out; 316 } 317 PGRP_LOCK(tp->t_pgrp); 318 if (LIST_EMPTY(&tp->t_pgrp->pg_members)) { 319 PGRP_UNLOCK(tp->t_pgrp); 320 sbuf_printf(&sb, "empty foreground process group\n"); 321 goto out; 322 } 323 324 /* 325 * Pick the most interesting process and copy some of its 326 * state for printing later. This operation could rely on stale 327 * data as we can't hold the proc slock or thread locks over the 328 * whole list. However, we're guaranteed not to reference an exited 329 * thread or proc since we hold the tty locked. 330 */ 331 p = NULL; 332 LIST_FOREACH(ppick, &tp->t_pgrp->pg_members, p_pglist) 333 if (proc_compare(p, ppick)) 334 p = ppick; 335 336 PROC_LOCK(p); 337 PGRP_UNLOCK(tp->t_pgrp); 338 td = NULL; 339 FOREACH_THREAD_IN_PROC(p, tdpick) 340 if (thread_compare(td, tdpick)) 341 td = tdpick; 342 stateprefix = ""; 343 thread_lock(td); 344 if (TD_IS_RUNNING(td)) 345 state = "running"; 346 else if (TD_ON_RUNQ(td) || TD_CAN_RUN(td)) 347 state = "runnable"; 348 else if (TD_IS_SLEEPING(td)) { 349 /* XXX: If we're sleeping, are we ever not in a queue? */ 350 if (TD_ON_SLEEPQ(td)) 351 state = td->td_wmesg; 352 else 353 state = "sleeping without queue"; 354 } else if (TD_ON_LOCK(td)) { 355 state = td->td_lockname; 356 stateprefix = "*"; 357 } else if (TD_IS_SUSPENDED(td)) 358 state = "suspended"; 359 else if (TD_AWAITING_INTR(td)) 360 state = "intrwait"; 361 else if (p->p_state == PRS_ZOMBIE) 362 state = "zombie"; 363 else 364 state = "unknown"; 365 pctcpu = (sched_pctcpu(td) * 10000 + FSCALE / 2) >> FSHIFT; 366#ifdef STACK 367 kstacks_val = atomic_load_int(&tty_info_kstacks); 368 print_kstacks = (kstacks_val != STACK_SBUF_FMT_NONE); 369 370 if (print_kstacks) { 371 if (TD_IS_SWAPPED(td)) 372 sterr = ENOENT; 373 else 374 sterr = stack_save_td(&stack, td); 375 } 376#endif 377 thread_unlock(td); 378 if (p->p_state == PRS_NEW || p->p_state == PRS_ZOMBIE) 379 rss = 0; 380 else 381 rss = pgtok(vmspace_resident_count(p->p_vmspace)); 382 microuptime(&rtime); 383 timevalsub(&rtime, &p->p_stats->p_start); 384 rufetchcalc(p, &ru, &utime, &stime); 385 pid = p->p_pid; 386 strlcpy(comm, p->p_comm, sizeof comm); 387 PROC_UNLOCK(p); 388 389 /* Print command, pid, state, rtime, utime, stime, %cpu, and rss. */ 390 sbuf_printf(&sb, 391 " cmd: %s %d [%s%s] %ld.%02ldr %ld.%02ldu %ld.%02lds %d%% %ldk\n", 392 comm, pid, stateprefix, state, 393 (long)rtime.tv_sec, rtime.tv_usec / 10000, 394 (long)utime.tv_sec, utime.tv_usec / 10000, 395 (long)stime.tv_sec, stime.tv_usec / 10000, 396 pctcpu / 100, rss); 397 398#ifdef STACK 399 if (print_kstacks && sterr == 0) 400 stack_sbuf_print_flags(&sb, &stack, M_NOWAIT, kstacks_val); 401#endif 402 403out: 404 sbuf_finish(&sb); 405 sbuf_delete(&sb); 406} 407