1 /*
2 * CDDL HEADER START
3 *
4 * The contents of this file are subject to the terms of the
5 * Common Development and Distribution License (the "License").
6 * You may not use this file except in compliance with the License.
7 *
8 * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
9 * or http://www.opensolaris.org/os/licensing.
10 * See the License for the specific language governing permissions
11 * and limitations under the License.
12 *
13 * When distributing Covered Code, include this CDDL HEADER in each
14 * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
15 * If applicable, add the following below this CDDL HEADER, with the
16 * fields enclosed by brackets "[]" replaced with your own identifying
17 * information: Portions Copyright [yyyy] [name of copyright owner]
18 *
19 * CDDL HEADER END
20 */
21 /*
22 * Copyright 2010 Sun Microsystems, Inc. All rights reserved.
23 * Use is subject to license terms.
24 */
25
26 /*
27 * Copyright (c) 2012, Joyent, Inc. All rights reserved.
28 */
29
30 /*
31 * lxproc -- a loosely Linux-compatible /proc
32 *
33 * The aspiration here is to provide something that sufficiently approximates
34 * the Linux /proc implementation for purposes of offering some compatibility
35 * for simple Linux /proc readers (e.g., ps/top/htop). However, it is not
36 * intended to exactly mimic Linux semantics; when choosing between offering
37 * compatibility and telling the truth, we emphatically pick the truth. A
38 * particular glaring example of this is the Linux notion of "tasks" (that is,
39 * threads), which -- due to historical misadventures on Linux -- allocate their
40 * identifiers from the process identifier space. (That is, each thread has in
41 * effect a pid.) Some Linux /proc readers have come to depend on this
42 * attribute, and become confused when threads appear with proper identifiers,
43 * so we simply opt for the pre-2.6 behavior, and do not present the tasks
44 * directory at all. Similarly, when choosing between offering compatibility
45 * and remaining consistent with our broader security model, we (obviously)
46 * choose security over compatibility. In short, this is meant to be a best
47 * effort -- no more.
48 */
49
50 #include <sys/cpupart.h>
51 #include <sys/cpuvar.h>
52 #include <sys/session.h>
53 #include <sys/vmparam.h>
54 #include <sys/mman.h>
55 #include <vm/rm.h>
56 #include <vm/seg_vn.h>
57 #include <sys/sdt.h>
58 #include <sys/strlog.h>
59 #include <sys/stropts.h>
60 #include <sys/cmn_err.h>
61 #include <sys/x86_archext.h>
62 #include <sys/archsystm.h>
63 #include <sys/fp.h>
64 #include <sys/pool_pset.h>
65 #include <sys/pset.h>
66 #include <sys/zone.h>
67 #include <sys/pghw.h>
68 #include <sys/vfs_opreg.h>
69
70 /* Dependent on procfs */
71 extern kthread_t *prchoose(proc_t *);
72
73 #include "lxproc.h"
74
75 extern pgcnt_t swapfs_minfree;
76 extern time_t boot_time;
77
78 /*
79 * Pointer to the vnode ops vector for this fs.
80 * This is instantiated in lxprinit() in lxpr_vfsops.c
81 */
82 vnodeops_t *lxpr_vnodeops;
83
84 static int lxpr_open(vnode_t **, int, cred_t *, caller_context_t *);
85 static int lxpr_close(vnode_t *, int, int, offset_t, cred_t *,
86 caller_context_t *);
87 static int lxpr_read(vnode_t *, uio_t *, int, cred_t *, caller_context_t *);
88 static int lxpr_getattr(vnode_t *, vattr_t *, int, cred_t *,
89 caller_context_t *);
90 static int lxpr_access(vnode_t *, int, int, cred_t *, caller_context_t *);
91 static int lxpr_lookup(vnode_t *, char *, vnode_t **,
92 pathname_t *, int, vnode_t *, cred_t *, caller_context_t *, int *,
93 pathname_t *);
94 static int lxpr_readdir(vnode_t *, uio_t *, cred_t *, int *,
95 caller_context_t *, int);
96 static int lxpr_readlink(vnode_t *, uio_t *, cred_t *, caller_context_t *);
97 static int lxpr_cmp(vnode_t *, vnode_t *, caller_context_t *);
98 static int lxpr_realvp(vnode_t *, vnode_t **, caller_context_t *);
99 static int lxpr_sync(void);
100 static void lxpr_inactive(vnode_t *, cred_t *, caller_context_t *);
101
102 static vnode_t *lxpr_lookup_procdir(vnode_t *, char *);
103 static vnode_t *lxpr_lookup_piddir(vnode_t *, char *);
104 static vnode_t *lxpr_lookup_not_a_dir(vnode_t *, char *);
105 static vnode_t *lxpr_lookup_fddir(vnode_t *, char *);
106 static vnode_t *lxpr_lookup_netdir(vnode_t *, char *);
107
108 static int lxpr_readdir_procdir(lxpr_node_t *, uio_t *, int *);
109 static int lxpr_readdir_piddir(lxpr_node_t *, uio_t *, int *);
110 static int lxpr_readdir_not_a_dir(lxpr_node_t *, uio_t *, int *);
111 static int lxpr_readdir_fddir(lxpr_node_t *, uio_t *, int *);
112 static int lxpr_readdir_netdir(lxpr_node_t *, uio_t *, int *);
113
114 static void lxpr_read_invalid(lxpr_node_t *, lxpr_uiobuf_t *);
115 static void lxpr_read_empty(lxpr_node_t *, lxpr_uiobuf_t *);
116 static void lxpr_read_cpuinfo(lxpr_node_t *, lxpr_uiobuf_t *);
117 static void lxpr_read_isdir(lxpr_node_t *, lxpr_uiobuf_t *);
118 static void lxpr_read_fd(lxpr_node_t *, lxpr_uiobuf_t *);
119 static void lxpr_read_kmsg(lxpr_node_t *, lxpr_uiobuf_t *);
120 static void lxpr_read_loadavg(lxpr_node_t *, lxpr_uiobuf_t *);
121 static void lxpr_read_meminfo(lxpr_node_t *, lxpr_uiobuf_t *);
122 static void lxpr_read_mounts(lxpr_node_t *, lxpr_uiobuf_t *);
123 static void lxpr_read_partitions(lxpr_node_t *, lxpr_uiobuf_t *);
124 static void lxpr_read_stat(lxpr_node_t *, lxpr_uiobuf_t *);
125 static void lxpr_read_uptime(lxpr_node_t *, lxpr_uiobuf_t *);
126 static void lxpr_read_version(lxpr_node_t *, lxpr_uiobuf_t *);
127
128 static void lxpr_read_pid_cmdline(lxpr_node_t *, lxpr_uiobuf_t *);
129 static void lxpr_read_pid_maps(lxpr_node_t *, lxpr_uiobuf_t *);
130 static void lxpr_read_pid_stat(lxpr_node_t *, lxpr_uiobuf_t *);
131 static void lxpr_read_pid_statm(lxpr_node_t *, lxpr_uiobuf_t *);
132 static void lxpr_read_pid_status(lxpr_node_t *, lxpr_uiobuf_t *);
133
134 static void lxpr_read_net_arp(lxpr_node_t *, lxpr_uiobuf_t *);
135 static void lxpr_read_net_dev(lxpr_node_t *, lxpr_uiobuf_t *);
136 static void lxpr_read_net_dev_mcast(lxpr_node_t *, lxpr_uiobuf_t *);
137 static void lxpr_read_net_igmp(lxpr_node_t *, lxpr_uiobuf_t *);
138 static void lxpr_read_net_ip_mr_cache(lxpr_node_t *, lxpr_uiobuf_t *);
139 static void lxpr_read_net_ip_mr_vif(lxpr_node_t *, lxpr_uiobuf_t *);
140 static void lxpr_read_net_mcfilter(lxpr_node_t *, lxpr_uiobuf_t *);
141 static void lxpr_read_net_netstat(lxpr_node_t *, lxpr_uiobuf_t *);
142 static void lxpr_read_net_raw(lxpr_node_t *, lxpr_uiobuf_t *);
143 static void lxpr_read_net_route(lxpr_node_t *, lxpr_uiobuf_t *);
144 static void lxpr_read_net_rpc(lxpr_node_t *, lxpr_uiobuf_t *);
145 static void lxpr_read_net_rt_cache(lxpr_node_t *, lxpr_uiobuf_t *);
146 static void lxpr_read_net_sockstat(lxpr_node_t *, lxpr_uiobuf_t *);
147 static void lxpr_read_net_snmp(lxpr_node_t *, lxpr_uiobuf_t *);
148 static void lxpr_read_net_stat(lxpr_node_t *, lxpr_uiobuf_t *);
149 static void lxpr_read_net_tcp(lxpr_node_t *, lxpr_uiobuf_t *);
150 static void lxpr_read_net_udp(lxpr_node_t *, lxpr_uiobuf_t *);
151 static void lxpr_read_net_unix(lxpr_node_t *, lxpr_uiobuf_t *);
152
153 /*
154 * Simple conversion
155 */
156 #define btok(x) ((x) >> 10) /* bytes to kbytes */
157 #define ptok(x) ((x) << (PAGESHIFT - 10)) /* pages to kbytes */
158
159 /*
160 * The lxproc vnode operations vector
161 */
162 const fs_operation_def_t lxpr_vnodeops_template[] = {
163 VOPNAME_OPEN, { .vop_open = lxpr_open },
164 VOPNAME_CLOSE, { .vop_close = lxpr_close },
165 VOPNAME_READ, { .vop_read = lxpr_read },
166 VOPNAME_GETATTR, { .vop_getattr = lxpr_getattr },
167 VOPNAME_ACCESS, { .vop_access = lxpr_access },
168 VOPNAME_LOOKUP, { .vop_lookup = lxpr_lookup },
169 VOPNAME_READDIR, { .vop_readdir = lxpr_readdir },
170 VOPNAME_READLINK, { .vop_readlink = lxpr_readlink },
171 VOPNAME_FSYNC, { .error = lxpr_sync },
172 VOPNAME_SEEK, { .error = lxpr_sync },
173 VOPNAME_INACTIVE, { .vop_inactive = lxpr_inactive },
174 VOPNAME_CMP, { .vop_cmp = lxpr_cmp },
175 VOPNAME_REALVP, { .vop_realvp = lxpr_realvp },
176 NULL, NULL
177 };
178
179 /*
180 * file contents of an lxproc directory.
181 */
182 static lxpr_dirent_t lxpr_dir[] = {
183 { LXPR_CMDLINE, "cmdline" },
184 { LXPR_CPUINFO, "cpuinfo" },
185 { LXPR_DEVICES, "devices" },
186 { LXPR_DMA, "dma" },
187 { LXPR_FILESYSTEMS, "filesystems" },
188 { LXPR_INTERRUPTS, "interrupts" },
189 { LXPR_IOPORTS, "ioports" },
190 { LXPR_KCORE, "kcore" },
191 { LXPR_KMSG, "kmsg" },
192 { LXPR_LOADAVG, "loadavg" },
193 { LXPR_MEMINFO, "meminfo" },
194 { LXPR_MOUNTS, "mounts" },
195 { LXPR_NETDIR, "net" },
196 { LXPR_PARTITIONS, "partitions" },
197 { LXPR_SELF, "self" },
198 { LXPR_STAT, "stat" },
199 { LXPR_UPTIME, "uptime" },
200 { LXPR_VERSION, "version" }
201 };
202
203 #define PROCDIRFILES (sizeof (lxpr_dir) / sizeof (lxpr_dir[0]))
204
205 /*
206 * Contents of an /lxproc/<pid> directory.
207 */
208 static lxpr_dirent_t piddir[] = {
209 { LXPR_PID_CMDLINE, "cmdline" },
210 { LXPR_PID_CPU, "cpu" },
211 { LXPR_PID_CURDIR, "cwd" },
212 { LXPR_PID_ENV, "environ" },
213 { LXPR_PID_EXE, "exe" },
214 { LXPR_PID_MAPS, "maps" },
215 { LXPR_PID_MEM, "mem" },
216 { LXPR_PID_ROOTDIR, "root" },
217 { LXPR_PID_STAT, "stat" },
218 { LXPR_PID_STATM, "statm" },
219 { LXPR_PID_STATUS, "status" },
220 { LXPR_PID_FDDIR, "fd" }
221 };
222
223 #define PIDDIRFILES (sizeof (piddir) / sizeof (piddir[0]))
224
225 /*
226 * contents of /lxproc/net directory
227 */
228 static lxpr_dirent_t netdir[] = {
229 { LXPR_NET_ARP, "arp" },
230 { LXPR_NET_DEV, "dev" },
231 { LXPR_NET_DEV_MCAST, "dev_mcast" },
232 { LXPR_NET_IGMP, "igmp" },
233 { LXPR_NET_IP_MR_CACHE, "ip_mr_cache" },
234 { LXPR_NET_IP_MR_VIF, "ip_mr_vif" },
235 { LXPR_NET_MCFILTER, "mcfilter" },
236 { LXPR_NET_NETSTAT, "netstat" },
237 { LXPR_NET_RAW, "raw" },
238 { LXPR_NET_ROUTE, "route" },
239 { LXPR_NET_RPC, "rpc" },
240 { LXPR_NET_RT_CACHE, "rt_cache" },
241 { LXPR_NET_SOCKSTAT, "sockstat" },
242 { LXPR_NET_SNMP, "snmp" },
243 { LXPR_NET_STAT, "stat" },
244 { LXPR_NET_TCP, "tcp" },
245 { LXPR_NET_UDP, "udp" },
246 { LXPR_NET_UNIX, "unix" }
247 };
248
249 #define NETDIRFILES (sizeof (netdir) / sizeof (netdir[0]))
250
251 /*
252 * These are the major signal number differences between Linux and native:
253 *
254 * ====================================
255 * | Number | Linux | Native |
256 * | ====== | ========= | ========== |
257 * | 7 | SIGBUS | SIGEMT |
258 * | 10 | SIGUSR1 | SIGBUS |
259 * | 12 | SIGUSR2 | SIGSYS |
260 * | 16 | SIGSTKFLT | SIGUSR1 |
261 * | 17 | SIGCHLD | SIGUSR2 |
262 * | 18 | SIGCONT | SIGCHLD |
263 * | 19 | SIGSTOP | SIGPWR |
264 * | 20 | SIGTSTP | SIGWINCH |
265 * | 21 | SIGTTIN | SIGURG |
266 * | 22 | SIGTTOU | SIGPOLL |
267 * | 23 | SIGURG | SIGSTOP |
268 * | 24 | SIGXCPU | SIGTSTP |
269 * | 25 | SIGXFSZ | SIGCONT |
270 * | 26 | SIGVTALARM | SIGTTIN |
271 * | 27 | SIGPROF | SIGTTOU |
272 * | 28 | SIGWINCH | SIGVTALARM |
273 * | 29 | SIGPOLL | SIGPROF |
274 * | 30 | SIGPWR | SIGXCPU |
275 * | 31 | SIGSYS | SIGXFSZ |
276 * ====================================
277 *
278 * Not every Linux signal maps to a native signal, nor does every native
279 * signal map to a Linux counterpart. However, when signals do map, the
280 * mapping is unique.
281 */
282 static int
283 lxpr_sigmap[NSIG] = {
284 0,
285 LX_SIGHUP,
286 LX_SIGINT,
287 LX_SIGQUIT,
288 LX_SIGILL,
289 LX_SIGTRAP,
290 LX_SIGABRT,
291 LX_SIGSTKFLT,
292 LX_SIGFPE,
293 LX_SIGKILL,
294 LX_SIGBUS,
295 LX_SIGSEGV,
296 LX_SIGSYS,
297 LX_SIGPIPE,
298 LX_SIGALRM,
299 LX_SIGTERM,
300 LX_SIGUSR1,
301 LX_SIGUSR2,
302 LX_SIGCHLD,
303 LX_SIGPWR,
304 LX_SIGWINCH,
305 LX_SIGURG,
306 LX_SIGPOLL,
307 LX_SIGSTOP,
308 LX_SIGTSTP,
309 LX_SIGCONT,
310 LX_SIGTTIN,
311 LX_SIGTTOU,
312 LX_SIGVTALRM,
313 LX_SIGPROF,
314 LX_SIGXCPU,
315 LX_SIGXFSZ,
316 -1, /* 32: illumos SIGWAITING */
317 -1, /* 33: illumos SIGLWP */
318 -1, /* 34: illumos SIGFREEZE */
319 -1, /* 35: illumos SIGTHAW */
320 -1, /* 36: illumos SIGCANCEL */
321 -1, /* 37: illumos SIGLOST */
322 -1, /* 38: illumos SIGXRES */
323 -1, /* 39: illumos SIGJVM1 */
324 -1, /* 40: illumos SIGJVM2 */
325 LX_SIGRTMIN, /* 41: illumos _SIGRTMIN */
326 LX_SIGRTMIN + 1,
327 LX_SIGRTMIN + 2,
328 LX_SIGRTMIN + 3,
329 LX_SIGRTMIN + 4,
330 LX_SIGRTMIN + 5,
331 LX_SIGRTMIN + 6,
332 LX_SIGRTMIN + 7,
333 LX_SIGRTMIN + 8,
334 LX_SIGRTMIN + 9,
335 LX_SIGRTMIN + 10,
336 LX_SIGRTMIN + 11,
337 LX_SIGRTMIN + 12,
338 LX_SIGRTMIN + 13,
339 LX_SIGRTMIN + 14,
340 LX_SIGRTMIN + 15,
341 LX_SIGRTMIN + 16,
342 LX_SIGRTMIN + 17,
343 LX_SIGRTMIN + 18,
344 LX_SIGRTMIN + 19,
345 LX_SIGRTMIN + 20,
346 LX_SIGRTMIN + 21,
347 LX_SIGRTMIN + 22,
348 LX_SIGRTMIN + 23,
349 LX_SIGRTMIN + 24,
350 LX_SIGRTMIN + 25,
351 LX_SIGRTMIN + 26,
352 LX_SIGRTMIN + 27,
353 LX_SIGRTMIN + 28,
354 LX_SIGRTMIN + 29,
355 LX_SIGRTMIN + 30,
356 LX_SIGRTMAX,
357 };
358
359 /*
360 * lxpr_open(): Vnode operation for VOP_OPEN()
361 */
362 static int
363 lxpr_open(vnode_t **vpp, int flag, cred_t *cr, caller_context_t *ct)
364 {
365 vnode_t *vp = *vpp;
366 lxpr_node_t *lxpnp = VTOLXP(vp);
367 lxpr_nodetype_t type = lxpnp->lxpr_type;
368 vnode_t *rvp;
369 int error = 0;
370
371 /*
372 * We only allow reading in this file systrem
373 */
374 if (flag & FWRITE)
375 return (EROFS);
376
377 /*
378 * If we are opening an underlying file only allow regular files
379 * reject the open for anything but a regular file.
380 * Just do it if we are opening the current or root directory.
381 */
382 if (lxpnp->lxpr_realvp != NULL) {
383 rvp = lxpnp->lxpr_realvp;
384
385 if (type == LXPR_PID_FD_FD && rvp->v_type != VREG)
386 error = EACCES;
387 else {
388 /*
389 * Need to hold rvp since VOP_OPEN() may release it.
390 */
391 VN_HOLD(rvp);
392 error = VOP_OPEN(&rvp, flag, cr, ct);
393 if (error) {
394 VN_RELE(rvp);
395 } else {
396 *vpp = rvp;
397 VN_RELE(vp);
398 }
399 }
400 }
401
402 if (type == LXPR_KMSG) {
403 ldi_ident_t li = VTOLXPM(vp)->lxprm_li;
404 struct strioctl str;
405 int rv;
406
407 /*
408 * Open the zone's console device using the layered driver
409 * interface.
410 */
411 if ((error = ldi_open_by_name("/dev/log", FREAD, cr,
412 &lxpnp->lxpr_cons_ldih, li)) != 0)
413 return (error);
414
415 /*
416 * Send an ioctl to the underlying console device, letting it
417 * know we're interested in getting console messages.
418 */
419 str.ic_cmd = I_CONSLOG;
420 str.ic_timout = 0;
421 str.ic_len = 0;
422 str.ic_dp = NULL;
423 if ((error = ldi_ioctl(lxpnp->lxpr_cons_ldih, I_STR,
424 (intptr_t)&str, FKIOCTL, cr, &rv)) != 0)
425 return (error);
426 }
427
428 return (error);
429 }
430
431
432 /*
433 * lxpr_close(): Vnode operation for VOP_CLOSE()
434 */
435 /* ARGSUSED */
436 static int
437 lxpr_close(vnode_t *vp, int flag, int count, offset_t offset, cred_t *cr,
438 caller_context_t *ct)
439 {
440 lxpr_node_t *lxpr = VTOLXP(vp);
441 lxpr_nodetype_t type = lxpr->lxpr_type;
442 int err;
443
444 /*
445 * we should never get here because the close is done on the realvp
446 * for these nodes
447 */
448 ASSERT(type != LXPR_PID_FD_FD &&
449 type != LXPR_PID_CURDIR &&
450 type != LXPR_PID_ROOTDIR &&
451 type != LXPR_PID_EXE);
452
453 if (type == LXPR_KMSG) {
454 if ((err = ldi_close(lxpr->lxpr_cons_ldih, 0, cr)) != 0)
455 return (err);
456 }
457
458 return (0);
459 }
460
461 static void (*lxpr_read_function[LXPR_NFILES])() = {
462 lxpr_read_isdir, /* /proc */
463 lxpr_read_isdir, /* /proc/<pid> */
464 lxpr_read_pid_cmdline, /* /proc/<pid>/cmdline */
465 lxpr_read_empty, /* /proc/<pid>/cpu */
466 lxpr_read_invalid, /* /proc/<pid>/cwd */
467 lxpr_read_empty, /* /proc/<pid>/environ */
468 lxpr_read_invalid, /* /proc/<pid>/exe */
469 lxpr_read_pid_maps, /* /proc/<pid>/maps */
470 lxpr_read_empty, /* /proc/<pid>/mem */
471 lxpr_read_invalid, /* /proc/<pid>/root */
472 lxpr_read_pid_stat, /* /proc/<pid>/stat */
473 lxpr_read_pid_statm, /* /proc/<pid>/statm */
474 lxpr_read_pid_status, /* /proc/<pid>/status */
475 lxpr_read_isdir, /* /proc/<pid>/fd */
476 lxpr_read_fd, /* /proc/<pid>/fd/nn */
477 lxpr_read_empty, /* /proc/cmdline */
478 lxpr_read_cpuinfo, /* /proc/cpuinfo */
479 lxpr_read_empty, /* /proc/devices */
480 lxpr_read_empty, /* /proc/dma */
481 lxpr_read_empty, /* /proc/filesystems */
482 lxpr_read_empty, /* /proc/interrupts */
483 lxpr_read_empty, /* /proc/ioports */
484 lxpr_read_empty, /* /proc/kcore */
485 lxpr_read_kmsg, /* /proc/kmsg */
486 lxpr_read_loadavg, /* /proc/loadavg */
487 lxpr_read_meminfo, /* /proc/meminfo */
488 lxpr_read_mounts, /* /proc/mounts */
489 lxpr_read_isdir, /* /proc/net */
490 lxpr_read_net_arp, /* /proc/net/arp */
491 lxpr_read_net_dev, /* /proc/net/dev */
492 lxpr_read_net_dev_mcast, /* /proc/net/dev_mcast */
493 lxpr_read_net_igmp, /* /proc/net/igmp */
494 lxpr_read_net_ip_mr_cache, /* /proc/net/ip_mr_cache */
495 lxpr_read_net_ip_mr_vif, /* /proc/net/ip_mr_vif */
496 lxpr_read_net_mcfilter, /* /proc/net/mcfilter */
497 lxpr_read_net_netstat, /* /proc/net/netstat */
498 lxpr_read_net_raw, /* /proc/net/raw */
499 lxpr_read_net_route, /* /proc/net/route */
500 lxpr_read_net_rpc, /* /proc/net/rpc */
501 lxpr_read_net_rt_cache, /* /proc/net/rt_cache */
502 lxpr_read_net_sockstat, /* /proc/net/sockstat */
503 lxpr_read_net_snmp, /* /proc/net/snmp */
504 lxpr_read_net_stat, /* /proc/net/stat */
505 lxpr_read_net_tcp, /* /proc/net/tcp */
506 lxpr_read_net_udp, /* /proc/net/udp */
507 lxpr_read_net_unix, /* /proc/net/unix */
508 lxpr_read_partitions, /* /proc/partitions */
509 lxpr_read_invalid, /* /proc/self */
510 lxpr_read_stat, /* /proc/stat */
511 lxpr_read_uptime, /* /proc/uptime */
512 lxpr_read_version, /* /proc/version */
513 };
514
515 /*
516 * Array of lookup functions, indexed by /lxproc file type.
517 */
518 static vnode_t *(*lxpr_lookup_function[LXPR_NFILES])() = {
519 lxpr_lookup_procdir, /* /proc */
520 lxpr_lookup_piddir, /* /proc/<pid> */
521 lxpr_lookup_not_a_dir, /* /proc/<pid>/cmdline */
522 lxpr_lookup_not_a_dir, /* /proc/<pid>/cpu */
523 lxpr_lookup_not_a_dir, /* /proc/<pid>/cwd */
524 lxpr_lookup_not_a_dir, /* /proc/<pid>/environ */
525 lxpr_lookup_not_a_dir, /* /proc/<pid>/exe */
526 lxpr_lookup_not_a_dir, /* /proc/<pid>/maps */
527 lxpr_lookup_not_a_dir, /* /proc/<pid>/mem */
528 lxpr_lookup_not_a_dir, /* /proc/<pid>/root */
529 lxpr_lookup_not_a_dir, /* /proc/<pid>/stat */
530 lxpr_lookup_not_a_dir, /* /proc/<pid>/statm */
531 lxpr_lookup_not_a_dir, /* /proc/<pid>/status */
532 lxpr_lookup_fddir, /* /proc/<pid>/fd */
533 lxpr_lookup_not_a_dir, /* /proc/<pid>/fd/nn */
534 lxpr_lookup_not_a_dir, /* /proc/cmdline */
535 lxpr_lookup_not_a_dir, /* /proc/cpuinfo */
536 lxpr_lookup_not_a_dir, /* /proc/devices */
537 lxpr_lookup_not_a_dir, /* /proc/dma */
538 lxpr_lookup_not_a_dir, /* /proc/filesystems */
539 lxpr_lookup_not_a_dir, /* /proc/interrupts */
540 lxpr_lookup_not_a_dir, /* /proc/ioports */
541 lxpr_lookup_not_a_dir, /* /proc/kcore */
542 lxpr_lookup_not_a_dir, /* /proc/kmsg */
543 lxpr_lookup_not_a_dir, /* /proc/loadavg */
544 lxpr_lookup_not_a_dir, /* /proc/meminfo */
545 lxpr_lookup_not_a_dir, /* /proc/mounts */
546 lxpr_lookup_netdir, /* /proc/net */
547 lxpr_lookup_not_a_dir, /* /proc/net/arp */
548 lxpr_lookup_not_a_dir, /* /proc/net/dev */
549 lxpr_lookup_not_a_dir, /* /proc/net/dev_mcast */
550 lxpr_lookup_not_a_dir, /* /proc/net/igmp */
551 lxpr_lookup_not_a_dir, /* /proc/net/ip_mr_cache */
552 lxpr_lookup_not_a_dir, /* /proc/net/ip_mr_vif */
553 lxpr_lookup_not_a_dir, /* /proc/net/mcfilter */
554 lxpr_lookup_not_a_dir, /* /proc/net/netstat */
555 lxpr_lookup_not_a_dir, /* /proc/net/raw */
556 lxpr_lookup_not_a_dir, /* /proc/net/route */
557 lxpr_lookup_not_a_dir, /* /proc/net/rpc */
558 lxpr_lookup_not_a_dir, /* /proc/net/rt_cache */
559 lxpr_lookup_not_a_dir, /* /proc/net/sockstat */
560 lxpr_lookup_not_a_dir, /* /proc/net/snmp */
561 lxpr_lookup_not_a_dir, /* /proc/net/stat */
562 lxpr_lookup_not_a_dir, /* /proc/net/tcp */
563 lxpr_lookup_not_a_dir, /* /proc/net/udp */
564 lxpr_lookup_not_a_dir, /* /proc/net/unix */
565 lxpr_lookup_not_a_dir, /* /proc/partitions */
566 lxpr_lookup_not_a_dir, /* /proc/self */
567 lxpr_lookup_not_a_dir, /* /proc/stat */
568 lxpr_lookup_not_a_dir, /* /proc/uptime */
569 lxpr_lookup_not_a_dir, /* /proc/version */
570 };
571
572 /*
573 * Array of readdir functions, indexed by /proc file type.
574 */
575 static int (*lxpr_readdir_function[LXPR_NFILES])() = {
576 lxpr_readdir_procdir, /* /proc */
577 lxpr_readdir_piddir, /* /proc/<pid> */
578 lxpr_readdir_not_a_dir, /* /proc/<pid>/cmdline */
579 lxpr_readdir_not_a_dir, /* /proc/<pid>/cpu */
580 lxpr_readdir_not_a_dir, /* /proc/<pid>/cwd */
581 lxpr_readdir_not_a_dir, /* /proc/<pid>/environ */
582 lxpr_readdir_not_a_dir, /* /proc/<pid>/exe */
583 lxpr_readdir_not_a_dir, /* /proc/<pid>/maps */
584 lxpr_readdir_not_a_dir, /* /proc/<pid>/mem */
585 lxpr_readdir_not_a_dir, /* /proc/<pid>/root */
586 lxpr_readdir_not_a_dir, /* /proc/<pid>/stat */
587 lxpr_readdir_not_a_dir, /* /proc/<pid>/statm */
588 lxpr_readdir_not_a_dir, /* /proc/<pid>/status */
589 lxpr_readdir_fddir, /* /proc/<pid>/fd */
590 lxpr_readdir_not_a_dir, /* /proc/<pid>/fd/nn */
591 lxpr_readdir_not_a_dir, /* /proc/cmdline */
592 lxpr_readdir_not_a_dir, /* /proc/cpuinfo */
593 lxpr_readdir_not_a_dir, /* /proc/devices */
594 lxpr_readdir_not_a_dir, /* /proc/dma */
595 lxpr_readdir_not_a_dir, /* /proc/filesystems */
596 lxpr_readdir_not_a_dir, /* /proc/interrupts */
597 lxpr_readdir_not_a_dir, /* /proc/ioports */
598 lxpr_readdir_not_a_dir, /* /proc/kcore */
599 lxpr_readdir_not_a_dir, /* /proc/kmsg */
600 lxpr_readdir_not_a_dir, /* /proc/loadavg */
601 lxpr_readdir_not_a_dir, /* /proc/meminfo */
602 lxpr_readdir_not_a_dir, /* /proc/mounts */
603 lxpr_readdir_netdir, /* /proc/net */
604 lxpr_readdir_not_a_dir, /* /proc/net/arp */
605 lxpr_readdir_not_a_dir, /* /proc/net/dev */
606 lxpr_readdir_not_a_dir, /* /proc/net/dev_mcast */
607 lxpr_readdir_not_a_dir, /* /proc/net/igmp */
608 lxpr_readdir_not_a_dir, /* /proc/net/ip_mr_cache */
609 lxpr_readdir_not_a_dir, /* /proc/net/ip_mr_vif */
610 lxpr_readdir_not_a_dir, /* /proc/net/mcfilter */
611 lxpr_readdir_not_a_dir, /* /proc/net/netstat */
612 lxpr_readdir_not_a_dir, /* /proc/net/raw */
613 lxpr_readdir_not_a_dir, /* /proc/net/route */
614 lxpr_readdir_not_a_dir, /* /proc/net/rpc */
615 lxpr_readdir_not_a_dir, /* /proc/net/rt_cache */
616 lxpr_readdir_not_a_dir, /* /proc/net/sockstat */
617 lxpr_readdir_not_a_dir, /* /proc/net/snmp */
618 lxpr_readdir_not_a_dir, /* /proc/net/stat */
619 lxpr_readdir_not_a_dir, /* /proc/net/tcp */
620 lxpr_readdir_not_a_dir, /* /proc/net/udp */
621 lxpr_readdir_not_a_dir, /* /proc/net/unix */
622 lxpr_readdir_not_a_dir, /* /proc/partitions */
623 lxpr_readdir_not_a_dir, /* /proc/self */
624 lxpr_readdir_not_a_dir, /* /proc/stat */
625 lxpr_readdir_not_a_dir, /* /proc/uptime */
626 lxpr_readdir_not_a_dir, /* /proc/version */
627 };
628
629
630 /*
631 * lxpr_read(): Vnode operation for VOP_READ()
632 *
633 * As the format of all the files that can be read in lxproc is human readable
634 * and not binary structures there do not have to be different read variants
635 * depending on whether the reading process model is 32- or 64-bit.
636 */
637 /* ARGSUSED */
638 static int
639 lxpr_read(vnode_t *vp, uio_t *uiop, int ioflag, cred_t *cr,
640 caller_context_t *ct)
641 {
642 lxpr_node_t *lxpnp = VTOLXP(vp);
643 lxpr_nodetype_t type = lxpnp->lxpr_type;
644 lxpr_uiobuf_t *uiobuf = lxpr_uiobuf_new(uiop);
645 int error;
646
647 ASSERT(type >= 0 && type < LXPR_NFILES);
648
649 lxpr_read_function[type](lxpnp, uiobuf);
650
651 error = lxpr_uiobuf_flush(uiobuf);
652 lxpr_uiobuf_free(uiobuf);
653
654 return (error);
655 }
656
657 /*
658 * lxpr_read_invalid(), lxpr_read_isdir(), lxpr_read_empty()
659 *
660 * Various special case reads:
661 * - trying to read a directory
662 * - invalid file (used to mean a file that should be implemented,
663 * but isn't yet)
664 * - empty file
665 * - wait to be able to read a file that will never have anything to read
666 */
667 /* ARGSUSED */
668 static void
669 lxpr_read_isdir(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
670 {
671 lxpr_uiobuf_seterr(uiobuf, EISDIR);
672 }
673
674 /* ARGSUSED */
675 static void
676 lxpr_read_invalid(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
677 {
678 lxpr_uiobuf_seterr(uiobuf, EINVAL);
679 }
680
681 /* ARGSUSED */
682 static void
683 lxpr_read_empty(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
684 {
685 }
686
687 /*
688 * lxpr_read_pid_cmdline():
689 *
690 * This is not precisely compatible with Linux: the Linux cmdline returns argv
691 * with the correct separation using \0 between the arguments, but we cannot do
692 * that without copying the real argv from the correct process context. This
693 * is too difficult to attempt so we pretend that the entire cmdline is just
694 * argv[0]. This is good enough for ps and htop to display correctly, but might
695 * cause some other things not to work correctly.
696 */
697 static void
698 lxpr_read_pid_cmdline(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
699 {
700 proc_t *p;
701 char *buf;
702
703 ASSERT(lxpnp->lxpr_type == LXPR_PID_CMDLINE);
704
705 p = lxpr_lock(lxpnp->lxpr_pid);
706 if (p == NULL) {
707 lxpr_uiobuf_seterr(uiobuf, EINVAL);
708 return;
709 }
710
711 buf = PTOU(p)->u_argv != 0 ? PTOU(p)->u_psargs : PTOU(p)->u_comm;
712
713 lxpr_uiobuf_write(uiobuf, buf, strlen(buf) + 1);
714 lxpr_unlock(p);
715 }
716
717 /*
718 * lxpr_read_pid_maps(): memory map file
719 */
720 static void
721 lxpr_read_pid_maps(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
722 {
723 proc_t *p;
724 struct as *as;
725 struct seg *seg;
726 char *buf;
727 int buflen = MAXPATHLEN;
728 struct print_data {
729 caddr_t saddr;
730 caddr_t eaddr;
731 int type;
732 char prot[5];
733 uint32_t offset;
734 vnode_t *vp;
735 struct print_data *next;
736 } *print_head = NULL;
737 struct print_data **print_tail = &print_head;
738 struct print_data *pbuf;
739
740 ASSERT(lxpnp->lxpr_type == LXPR_PID_MAPS);
741
742 p = lxpr_lock(lxpnp->lxpr_pid);
743 if (p == NULL) {
744 lxpr_uiobuf_seterr(uiobuf, EINVAL);
745 return;
746 }
747
748 as = p->p_as;
749
750 if (as == &kas) {
751 lxpr_unlock(p);
752 return;
753 }
754
755 mutex_exit(&p->p_lock);
756
757 /* Iterate over all segments in the address space */
758 AS_LOCK_ENTER(as, &as->a_lock, RW_READER);
759 for (seg = AS_SEGFIRST(as); seg != NULL; seg = AS_SEGNEXT(as, seg)) {
760 vnode_t *vp;
761 uint_t protbits;
762
763 pbuf = kmem_alloc(sizeof (*pbuf), KM_SLEEP);
764
765 pbuf->saddr = seg->s_base;
766 pbuf->eaddr = seg->s_base+seg->s_size;
767 pbuf->type = SEGOP_GETTYPE(seg, seg->s_base);
768
769 /*
770 * Cheat and only use the protection bits of the first page
771 * in the segment
772 */
773 (void) strncpy(pbuf->prot, "----", sizeof (pbuf->prot));
774 (void) SEGOP_GETPROT(seg, seg->s_base, 0, &protbits);
775
776 if (protbits & PROT_READ) pbuf->prot[0] = 'r';
777 if (protbits & PROT_WRITE) pbuf->prot[1] = 'w';
778 if (protbits & PROT_EXEC) pbuf->prot[2] = 'x';
779 if (pbuf->type & MAP_SHARED) pbuf->prot[3] = 's';
780 else if (pbuf->type & MAP_PRIVATE) pbuf->prot[3] = 'p';
781
782 if (seg->s_ops == &segvn_ops &&
783 SEGOP_GETVP(seg, seg->s_base, &vp) == 0 &&
784 vp != NULL && vp->v_type == VREG) {
785 VN_HOLD(vp);
786 pbuf->vp = vp;
787 } else {
788 pbuf->vp = NULL;
789 }
790
791 pbuf->offset = (uint32_t)SEGOP_GETOFFSET(seg, pbuf->saddr);
792
793 pbuf->next = NULL;
794 *print_tail = pbuf;
795 print_tail = &pbuf->next;
796 }
797 AS_LOCK_EXIT(as, &as->a_lock);
798 mutex_enter(&p->p_lock);
799 lxpr_unlock(p);
800
801 buf = kmem_alloc(buflen, KM_SLEEP);
802
803 /* print the data we've extracted */
804 pbuf = print_head;
805 while (pbuf != NULL) {
806 struct print_data *pbuf_next;
807 vattr_t vattr;
808
809 int maj = 0;
810 int min = 0;
811 u_longlong_t inode = 0;
812
813 *buf = '\0';
814 if (pbuf->vp != NULL) {
815 vattr.va_mask = AT_FSID | AT_NODEID;
816 if (VOP_GETATTR(pbuf->vp, &vattr, 0, CRED(),
817 NULL) == 0) {
818 maj = getmajor(vattr.va_fsid);
819 min = getminor(vattr.va_fsid);
820 inode = vattr.va_nodeid;
821 }
822 (void) vnodetopath(NULL, pbuf->vp, buf, buflen, CRED());
823 VN_RELE(pbuf->vp);
824 }
825
826 if (*buf != '\0') {
827 lxpr_uiobuf_printf(uiobuf,
828 "%08x-%08x %s %08x %02d:%03d %lld %s\n",
829 pbuf->saddr, pbuf->eaddr, pbuf->prot, pbuf->offset,
830 maj, min, inode, buf);
831 } else {
832 lxpr_uiobuf_printf(uiobuf,
833 "%08x-%08x %s %08x %02d:%03d %lld\n",
834 pbuf->saddr, pbuf->eaddr, pbuf->prot, pbuf->offset,
835 maj, min, inode);
836 }
837
838 pbuf_next = pbuf->next;
839 kmem_free(pbuf, sizeof (*pbuf));
840 pbuf = pbuf_next;
841 }
842
843 kmem_free(buf, buflen);
844 }
845
846 /*
847 * lxpr_read_pid_statm(): memory status file
848 */
849 static void
850 lxpr_read_pid_statm(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
851 {
852 proc_t *p;
853 struct as *as;
854 size_t vsize;
855 size_t rss;
856
857 ASSERT(lxpnp->lxpr_type == LXPR_PID_STATM);
858
859 p = lxpr_lock(lxpnp->lxpr_pid);
860 if (p == NULL) {
861 lxpr_uiobuf_seterr(uiobuf, EINVAL);
862 return;
863 }
864
865 as = p->p_as;
866
867 mutex_exit(&p->p_lock);
868
869 AS_LOCK_ENTER(as, &as->a_lock, RW_READER);
870 vsize = btopr(as->a_resvsize);
871 rss = rm_asrss(as);
872 AS_LOCK_EXIT(as, &as->a_lock);
873
874 mutex_enter(&p->p_lock);
875 lxpr_unlock(p);
876
877 lxpr_uiobuf_printf(uiobuf,
878 "%lu %lu %lu %lu %lu %lu %lu\n",
879 vsize, rss, 0l, rss, 0l, 0l, 0l);
880 }
881
882 /*
883 * lxpr_read_pid_status(): status file
884 */
885 static void
886 lxpr_read_pid_status(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
887 {
888 proc_t *p;
889 kthread_t *t;
890 user_t *up;
891 cred_t *cr;
892 const gid_t *groups;
893 int ngroups;
894 struct as *as;
895 char *status;
896 pid_t pid, ppid;
897 size_t vsize;
898 size_t rss;
899 k_sigset_t current, ignore, handle;
900 int i, lx_sig;
901
902 ASSERT(lxpnp->lxpr_type == LXPR_PID_STATUS);
903
904 p = lxpr_lock(lxpnp->lxpr_pid);
905 if (p == NULL) {
906 lxpr_uiobuf_seterr(uiobuf, EINVAL);
907 return;
908 }
909
910 pid = p->p_pid;
911
912 /*
913 * Convert pid to the Linux default of 1 if we're the zone's init
914 * process
915 */
916 if (pid == curproc->p_zone->zone_proc_initpid) {
917 pid = 1;
918 ppid = 0; /* parent pid for init is 0 */
919 } else {
920 /*
921 * Make sure not to reference parent PIDs that reside outside
922 * the zone
923 */
924 ppid = ((p->p_flag & SZONETOP)
925 ? curproc->p_zone->zone_zsched->p_pid : p->p_ppid);
926
927 /*
928 * Convert ppid to the Linux default of 1 if our parent is the
929 * zone's init process
930 */
931 if (ppid == curproc->p_zone->zone_proc_initpid)
932 ppid = 1;
933 }
934
935 t = prchoose(p);
936 if (t != NULL) {
937 switch (t->t_state) {
938 case TS_SLEEP:
939 status = "S (sleeping)";
940 break;
941 case TS_RUN:
942 case TS_ONPROC:
943 status = "R (running)";
944 break;
945 case TS_ZOMB:
946 status = "Z (zombie)";
947 break;
948 case TS_STOPPED:
949 status = "T (stopped)";
950 break;
951 default:
952 status = "! (unknown)";
953 break;
954 }
955 thread_unlock(t);
956 } else {
957 /*
958 * there is a hole in the exit code, where a proc can have
959 * no threads but it is yet to be flagged SZOMB. We will
960 * assume we are about to become a zombie
961 */
962 status = "Z (zombie)";
963 }
964
965 up = PTOU(p);
966 mutex_enter(&p->p_crlock);
967 crhold(cr = p->p_cred);
968 mutex_exit(&p->p_crlock);
969
970 lxpr_uiobuf_printf(uiobuf,
971 "Name:\t%s\n"
972 "State:\t%s\n"
973 "Tgid:\t%d\n"
974 "Pid:\t%d\n"
975 "PPid:\t%d\n"
976 "TracerPid:\t%d\n"
977 "Uid:\t%u\t%u\t%u\t%u\n"
978 "Gid:\t%u\t%u\t%u\t%u\n"
979 "FDSize:\t%d\n"
980 "Groups:\t",
981 up->u_comm,
982 status,
983 pid, /* thread group id - same as pid */
984 pid,
985 ppid,
986 0,
987 crgetruid(cr), crgetuid(cr), crgetsuid(cr), crgetuid(cr),
988 crgetrgid(cr), crgetgid(cr), crgetsgid(cr), crgetgid(cr),
989 p->p_fno_ctl);
990
991 ngroups = crgetngroups(cr);
992 groups = crgetgroups(cr);
993 for (i = 0; i < ngroups; i++) {
994 lxpr_uiobuf_printf(uiobuf,
995 "%u ",
996 groups[i]);
997 }
998 crfree(cr);
999
1000 as = p->p_as;
1001 if ((p->p_stat != SZOMB) && !(p->p_flag & SSYS) && (as != &kas)) {
1002 mutex_exit(&p->p_lock);
1003 AS_LOCK_ENTER(as, &as->a_lock, RW_READER);
1004 vsize = as->a_resvsize;
1005 rss = rm_asrss(as);
1006 AS_LOCK_EXIT(as, &as->a_lock);
1007 mutex_enter(&p->p_lock);
1008
1009 lxpr_uiobuf_printf(uiobuf,
1010 "\n"
1011 "VmSize:\t%8lu kB\n"
1012 "VmLck:\t%8lu kB\n"
1013 "VmRSS:\t%8lu kB\n"
1014 "VmData:\t%8lu kB\n"
1015 "VmStk:\t%8lu kB\n"
1016 "VmExe:\t%8lu kB\n"
1017 "VmLib:\t%8lu kB",
1018 btok(vsize),
1019 0l,
1020 ptok(rss),
1021 0l,
1022 btok(p->p_stksize),
1023 ptok(rss),
1024 0l);
1025 }
1026
1027 sigemptyset(¤t);
1028 sigemptyset(&ignore);
1029 sigemptyset(&handle);
1030
1031 for (i = 1; i < NSIG; i++) {
1032 lx_sig = lxpr_sigmap[i];
1033
1034 if ((lx_sig > 0) && (lx_sig < LX_NSIG)) {
1035 if (sigismember(&p->p_sig, i))
1036 sigaddset(¤t, lx_sig);
1037
1038 if (up->u_signal[i - 1] == SIG_IGN)
1039 sigaddset(&ignore, lx_sig);
1040 else if (up->u_signal[i - 1] != SIG_DFL)
1041 sigaddset(&handle, lx_sig);
1042 }
1043 }
1044
1045 lxpr_uiobuf_printf(uiobuf,
1046 "\n"
1047 "SigPnd:\t%08x%08x\n"
1048 "SigBlk:\t%08x%08x\n"
1049 "SigIgn:\t%08x%08x\n"
1050 "SigCgt:\t%08x%08x\n"
1051 "CapInh:\t%016x\n"
1052 "CapPrm:\t%016x\n"
1053 "CapEff:\t%016x\n",
1054 current.__sigbits[1], current.__sigbits[0],
1055 0, 0, /* signals blocked on per thread basis */
1056 ignore.__sigbits[1], ignore.__sigbits[0],
1057 handle.__sigbits[1], handle.__sigbits[0],
1058 /* Can't do anything with linux capabilities */
1059 0,
1060 0,
1061 0);
1062
1063 lxpr_unlock(p);
1064 }
1065
1066
1067 /*
1068 * lxpr_read_pid_stat(): pid stat file
1069 */
1070 static void
1071 lxpr_read_pid_stat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1072 {
1073 proc_t *p;
1074 kthread_t *t;
1075 struct as *as;
1076 char stat;
1077 pid_t pid, ppid, pgpid, spid;
1078 gid_t psgid;
1079 dev_t psdev;
1080 size_t rss, vsize;
1081 int nice, pri;
1082 caddr_t wchan;
1083 processorid_t cpu;
1084
1085 ASSERT(lxpnp->lxpr_type == LXPR_PID_STAT);
1086
1087 p = lxpr_lock(lxpnp->lxpr_pid);
1088 if (p == NULL) {
1089 lxpr_uiobuf_seterr(uiobuf, EINVAL);
1090 return;
1091 }
1092
1093 pid = p->p_pid;
1094
1095 /*
1096 * Set Linux defaults if we're the zone's init process
1097 */
1098 if (pid == curproc->p_zone->zone_proc_initpid) {
1099 pid = 1; /* PID for init */
1100 ppid = 0; /* parent PID for init is 0 */
1101 pgpid = 0; /* process group for init is 0 */
1102 psgid = (gid_t)-1; /* credential GID for init is -1 */
1103 spid = 0; /* session id for init is 0 */
1104 psdev = 0; /* session device for init is 0 */
1105 } else {
1106 /*
1107 * Make sure not to reference parent PIDs that reside outside
1108 * the zone
1109 */
1110 ppid = ((p->p_flag & SZONETOP) ?
1111 curproc->p_zone->zone_zsched->p_pid : p->p_ppid);
1112
1113 /*
1114 * Convert ppid to the Linux default of 1 if our parent is the
1115 * zone's init process
1116 */
1117 if (ppid == curproc->p_zone->zone_proc_initpid)
1118 ppid = 1;
1119
1120 pgpid = p->p_pgrp;
1121
1122 mutex_enter(&p->p_splock);
1123 mutex_enter(&p->p_sessp->s_lock);
1124 spid = p->p_sessp->s_sid;
1125 psdev = p->p_sessp->s_dev;
1126 if (p->p_sessp->s_cred)
1127 psgid = crgetgid(p->p_sessp->s_cred);
1128 else
1129 psgid = crgetgid(p->p_cred);
1130
1131 mutex_exit(&p->p_sessp->s_lock);
1132 mutex_exit(&p->p_splock);
1133 }
1134
1135 t = prchoose(p);
1136 if (t != NULL) {
1137 switch (t->t_state) {
1138 case TS_SLEEP:
1139 stat = 'S'; break;
1140 case TS_RUN:
1141 case TS_ONPROC:
1142 stat = 'R'; break;
1143 case TS_ZOMB:
1144 stat = 'Z'; break;
1145 case TS_STOPPED:
1146 stat = 'T'; break;
1147 default:
1148 stat = '!'; break;
1149 }
1150
1151 if (CL_DONICE(t, NULL, 0, &nice) != 0)
1152 nice = 0;
1153
1154 pri = t->t_pri;
1155 wchan = t->t_wchan;
1156 cpu = t->t_cpu->cpu_id;
1157 thread_unlock(t);
1158 } else {
1159 /* Only zombies have no threads */
1160 stat = 'Z';
1161 nice = 0;
1162 pri = 0;
1163 wchan = 0;
1164 cpu = 0;
1165 }
1166 as = p->p_as;
1167 mutex_exit(&p->p_lock);
1168 AS_LOCK_ENTER(as, &as->a_lock, RW_READER);
1169 vsize = as->a_resvsize;
1170 rss = rm_asrss(as);
1171 AS_LOCK_EXIT(as, &as->a_lock);
1172 mutex_enter(&p->p_lock);
1173
1174 lxpr_uiobuf_printf(uiobuf,
1175 "%d (%s) %c %d %d %d %d %d "
1176 "%lu %lu %lu %lu %lu "
1177 "%lu %lu %ld %ld "
1178 "%d %d %d "
1179 "%lu "
1180 "%lu "
1181 "%lu %ld %llu "
1182 "%lu %lu %u "
1183 "%lu %lu "
1184 "%lu %lu %lu %lu "
1185 "%lu "
1186 "%lu %lu "
1187 "%d "
1188 "%d"
1189 "\n",
1190 pid, PTOU(p)->u_comm, stat, ppid, pgpid, spid, psdev, psgid,
1191 0l, 0l, 0l, 0l, 0l, /* flags, minflt, cminflt, majflt, cmajflt */
1192 p->p_utime, p->p_stime, p->p_cutime, p->p_cstime,
1193 pri, nice, p->p_lwpcnt,
1194 0l, /* itrealvalue (time before next SIGALRM) */
1195 PTOU(p)->u_ticks,
1196 vsize, rss, p->p_vmem_ctl,
1197 0l, 0l, USRSTACK, /* startcode, endcode, startstack */
1198 0l, 0l, /* kstkesp, kstkeip */
1199 0l, 0l, 0l, 0l, /* signal, blocked, sigignore, sigcatch */
1200 wchan,
1201 0l, 0l, /* nswap, cnswap */
1202 0, /* exit_signal */
1203 cpu);
1204
1205 lxpr_unlock(p);
1206 }
1207
1208 /* ARGSUSED */
1209 static void
1210 lxpr_read_net_arp(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1211 {
1212 }
1213
1214 /* ARGSUSED */
1215 static void
1216 lxpr_read_net_dev(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1217 {
1218 lxpr_uiobuf_printf(uiobuf, "Inter-| Receive "
1219 " | Transmit\n");
1220 lxpr_uiobuf_printf(uiobuf, " face |bytes packets errs drop fifo"
1221 " frame compressed multicast|bytes packets errs drop fifo"
1222 " colls carrier compressed\n");
1223
1224 /*
1225 * Data about each interface should go here, but that shouldn't be added
1226 * unless there is an lxproc reader that actually makes use of it (and
1227 * doesn't need anything else that we refuse to provide)...
1228 */
1229 }
1230
1231 /* ARGSUSED */
1232 static void
1233 lxpr_read_net_dev_mcast(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1234 {
1235 }
1236
1237 /* ARGSUSED */
1238 static void
1239 lxpr_read_net_igmp(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1240 {
1241 }
1242
1243 /* ARGSUSED */
1244 static void
1245 lxpr_read_net_ip_mr_cache(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1246 {
1247 }
1248
1249 /* ARGSUSED */
1250 static void
1251 lxpr_read_net_ip_mr_vif(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1252 {
1253 }
1254
1255 /* ARGSUSED */
1256 static void
1257 lxpr_read_net_mcfilter(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1258 {
1259 }
1260
1261 /* ARGSUSED */
1262 static void
1263 lxpr_read_net_netstat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1264 {
1265 }
1266
1267 /* ARGSUSED */
1268 static void
1269 lxpr_read_net_raw(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1270 {
1271 }
1272
1273 /* ARGSUSED */
1274 static void
1275 lxpr_read_net_route(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1276 {
1277 }
1278
1279 /* ARGSUSED */
1280 static void
1281 lxpr_read_net_rpc(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1282 {
1283 }
1284
1285 /* ARGSUSED */
1286 static void
1287 lxpr_read_net_rt_cache(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1288 {
1289 }
1290
1291 /* ARGSUSED */
1292 static void
1293 lxpr_read_net_sockstat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1294 {
1295 }
1296
1297 /* ARGSUSED */
1298 static void
1299 lxpr_read_net_snmp(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1300 {
1301 }
1302
1303 /* ARGSUSED */
1304 static void
1305 lxpr_read_net_stat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1306 {
1307 }
1308
1309 /* ARGSUSED */
1310 static void
1311 lxpr_read_net_tcp(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1312 {
1313 }
1314
1315 /* ARGSUSED */
1316 static void
1317 lxpr_read_net_udp(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1318 {
1319 }
1320
1321 /* ARGSUSED */
1322 static void
1323 lxpr_read_net_unix(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1324 {
1325 }
1326
1327 /*
1328 * lxpr_read_kmsg(): read the contents of the kernel message queue. We
1329 * translate this into the reception of console messages for this zone; each
1330 * read copies out a single zone console message, or blocks until the next one
1331 * is produced.
1332 */
1333
1334 #define LX_KMSG_PRI "<0>"
1335
1336 static void
1337 lxpr_read_kmsg(lxpr_node_t *lxpnp, struct lxpr_uiobuf *uiobuf)
1338 {
1339 ldi_handle_t lh = lxpnp->lxpr_cons_ldih;
1340 mblk_t *mp;
1341
1342 if (ldi_getmsg(lh, &mp, NULL) == 0) {
1343 /*
1344 * lxproc doesn't like successive reads to the same file
1345 * descriptor unless we do an explicit rewind each time.
1346 */
1347 lxpr_uiobuf_seek(uiobuf, 0);
1348
1349 lxpr_uiobuf_printf(uiobuf, "%s%s", LX_KMSG_PRI,
1350 mp->b_cont->b_rptr);
1351
1352 freemsg(mp);
1353 }
1354 }
1355
1356 /*
1357 * lxpr_read_loadavg(): read the contents of the "loadavg" file. We do just
1358 * enough for uptime and other simple lxproc readers to work
1359 */
1360 extern int nthread;
1361
1362 static void
1363 lxpr_read_loadavg(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1364 {
1365 ulong_t avenrun1;
1366 ulong_t avenrun5;
1367 ulong_t avenrun15;
1368 ulong_t avenrun1_cs;
1369 ulong_t avenrun5_cs;
1370 ulong_t avenrun15_cs;
1371 int loadavg[3];
1372 int *loadbuf;
1373 cpupart_t *cp;
1374 zone_t *zone = LXPTOZ(lxpnp);
1375
1376 uint_t nrunnable = 0;
1377 rctl_qty_t nlwps;
1378
1379 ASSERT(lxpnp->lxpr_type == LXPR_LOADAVG);
1380
1381 mutex_enter(&cpu_lock);
1382
1383 /*
1384 * Need to add up values over all CPU partitions. If pools are active,
1385 * only report the values of the zone's partition, which by definition
1386 * includes the current CPU.
1387 */
1388 if (pool_pset_enabled()) {
1389 psetid_t psetid = zone_pset_get(curproc->p_zone);
1390
1391 ASSERT(curproc->p_zone != &zone0);
1392 cp = CPU->cpu_part;
1393
1394 nrunnable = cp->cp_nrunning + cp->cp_nrunnable;
1395 (void) cpupart_get_loadavg(psetid, &loadavg[0], 3);
1396 loadbuf = &loadavg[0];
1397 } else {
1398 cp = cp_list_head;
1399 do {
1400 nrunnable += cp->cp_nrunning + cp->cp_nrunnable;
1401 } while ((cp = cp->cp_next) != cp_list_head);
1402
1403 loadbuf = zone == global_zone ?
1404 &avenrun[0] : zone->zone_avenrun;
1405 }
1406
1407 /*
1408 * If we're in the non-global zone, we'll report the total number of
1409 * LWPs in the zone for the "nproc" parameter of /proc/loadavg,
1410 * otherwise will just use nthread (which will include kernel threads,
1411 * but should be good enough for lxproc).
1412 */
1413 nlwps = zone == global_zone ? nthread : zone->zone_nlwps;
1414
1415 mutex_exit(&cpu_lock);
1416
1417 avenrun1 = loadbuf[0] >> FSHIFT;
1418 avenrun1_cs = ((loadbuf[0] & (FSCALE-1)) * 100) >> FSHIFT;
1419 avenrun5 = loadbuf[1] >> FSHIFT;
1420 avenrun5_cs = ((loadbuf[1] & (FSCALE-1)) * 100) >> FSHIFT;
1421 avenrun15 = loadbuf[2] >> FSHIFT;
1422 avenrun15_cs = ((loadbuf[2] & (FSCALE-1)) * 100) >> FSHIFT;
1423
1424 lxpr_uiobuf_printf(uiobuf,
1425 "%ld.%02d %ld.%02d %ld.%02d %d/%d %d\n",
1426 avenrun1, avenrun1_cs,
1427 avenrun5, avenrun5_cs,
1428 avenrun15, avenrun15_cs,
1429 nrunnable, nlwps, 0);
1430 }
1431
1432 /*
1433 * lxpr_read_meminfo(): read the contents of the "meminfo" file.
1434 */
1435 static void
1436 lxpr_read_meminfo(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1437 {
1438 zone_t *zone = LXPTOZ(lxpnp);
1439 int global = zone == global_zone;
1440 long total_mem, free_mem, total_swap, used_swap;
1441
1442 ASSERT(lxpnp->lxpr_type == LXPR_MEMINFO);
1443
1444 if (global || zone->zone_phys_mem_ctl == UINT64_MAX) {
1445 total_mem = physmem * PAGESIZE;
1446 free_mem = freemem * PAGESIZE;
1447 } else {
1448 total_mem = zone->zone_phys_mem_ctl;
1449 free_mem = zone->zone_phys_mem_ctl - zone->zone_phys_mem;
1450 }
1451
1452 if (global || zone->zone_max_swap_ctl == UINT64_MAX) {
1453 total_swap = k_anoninfo.ani_max * PAGESIZE;
1454 used_swap = k_anoninfo.ani_phys_resv * PAGESIZE;
1455 } else {
1456 mutex_enter(&zone->zone_mem_lock);
1457 total_swap = zone->zone_max_swap_ctl;
1458 used_swap = zone->zone_max_swap;
1459 mutex_exit(&zone->zone_mem_lock);
1460 }
1461
1462 lxpr_uiobuf_printf(uiobuf,
1463 " total: used: free: shared: buffers: cached:\n"
1464 "Mem: %8lu %8lu %8lu %8u %8u %8u\n"
1465 "Swap: %8lu %8lu %8lu\n"
1466 "MemTotal: %8lu kB\n"
1467 "MemFree: %8lu kB\n"
1468 "MemShared: %8u kB\n"
1469 "Buffers: %8u kB\n"
1470 "Cached: %8u kB\n"
1471 "SwapCached:%8u kB\n"
1472 "Active: %8u kB\n"
1473 "Inactive: %8u kB\n"
1474 "HighTotal: %8u kB\n"
1475 "HighFree: %8u kB\n"
1476 "LowTotal: %8u kB\n"
1477 "LowFree: %8u kB\n"
1478 "SwapTotal: %8lu kB\n"
1479 "SwapFree: %8lu kB\n",
1480 total_mem, total_mem - free_mem, free_mem, 0, 0, 0,
1481 total_swap, used_swap, total_swap - used_swap,
1482 btok(total_mem), /* MemTotal */
1483 btok(free_mem), /* MemFree */
1484 0, /* MemShared */
1485 0, /* Buffers */
1486 0, /* Cached */
1487 0, /* SwapCached */
1488 0, /* Active */
1489 0, /* Inactive */
1490 0, /* HighTotal */
1491 0, /* HighFree */
1492 btok(total_mem), /* LowTotal */
1493 btok(free_mem), /* LowFree */
1494 btok(total_swap), /* SwapTotal */
1495 btok(total_swap - used_swap)); /* SwapFree */
1496 }
1497
1498 /*
1499 * lxpr_read_mounts():
1500 */
1501 /* ARGSUSED */
1502 static void
1503 lxpr_read_mounts(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1504 {
1505 struct vfs *vfsp;
1506 struct vfs *vfslist;
1507 zone_t *zone = LXPTOZ(lxpnp);
1508 struct print_data {
1509 refstr_t *vfs_mntpt;
1510 refstr_t *vfs_resource;
1511 uint_t vfs_flag;
1512 int vfs_fstype;
1513 struct print_data *next;
1514 } *print_head = NULL;
1515 struct print_data **print_tail = &print_head;
1516 struct print_data *printp;
1517
1518 vfs_list_read_lock();
1519
1520 if (zone == global_zone) {
1521 vfsp = vfslist = rootvfs;
1522 } else {
1523 vfsp = vfslist = zone->zone_vfslist;
1524 /*
1525 * If the zone has a root entry, it will be the first in
1526 * the list. If it doesn't, we conjure one up.
1527 */
1528 if (vfslist == NULL || strcmp(refstr_value(vfsp->vfs_mntpt),
1529 zone->zone_rootpath) != 0) {
1530 struct vfs *tvfsp;
1531 /*
1532 * The root of the zone is not a mount point. The vfs
1533 * we want to report is that of the zone's root vnode.
1534 */
1535 tvfsp = zone->zone_rootvp->v_vfsp;
1536
1537 lxpr_uiobuf_printf(uiobuf,
1538 "/ / %s %s 0 0\n",
1539 vfssw[tvfsp->vfs_fstype].vsw_name,
1540 tvfsp->vfs_flag & VFS_RDONLY ? "ro" : "rw");
1541
1542 }
1543 if (vfslist == NULL) {
1544 vfs_list_unlock();
1545 return;
1546 }
1547 }
1548
1549 /*
1550 * Later on we have to do a lookupname, which can end up causing
1551 * another vfs_list_read_lock() to be called. Which can lead to a
1552 * deadlock. To avoid this, we extract the data we need into a local
1553 * list, then we can run this list without holding vfs_list_read_lock()
1554 * We keep the list in the same order as the vfs_list
1555 */
1556 do {
1557 /* Skip mounts we shouldn't show */
1558 if (vfsp->vfs_flag & VFS_NOMNTTAB) {
1559 goto nextfs;
1560 }
1561
1562 printp = kmem_alloc(sizeof (*printp), KM_SLEEP);
1563 refstr_hold(vfsp->vfs_mntpt);
1564 printp->vfs_mntpt = vfsp->vfs_mntpt;
1565 refstr_hold(vfsp->vfs_resource);
1566 printp->vfs_resource = vfsp->vfs_resource;
1567 printp->vfs_flag = vfsp->vfs_flag;
1568 printp->vfs_fstype = vfsp->vfs_fstype;
1569 printp->next = NULL;
1570
1571 *print_tail = printp;
1572 print_tail = &printp->next;
1573
1574 nextfs:
1575 vfsp = (zone == global_zone) ?
1576 vfsp->vfs_next : vfsp->vfs_zone_next;
1577
1578 } while (vfsp != vfslist);
1579
1580 vfs_list_unlock();
1581
1582 /*
1583 * now we can run through what we've extracted without holding
1584 * vfs_list_read_lock()
1585 */
1586 printp = print_head;
1587 while (printp != NULL) {
1588 struct print_data *printp_next;
1589 const char *resource;
1590 char *mntpt;
1591 struct vnode *vp;
1592 int error;
1593
1594 mntpt = (char *)refstr_value(printp->vfs_mntpt);
1595 resource = refstr_value(printp->vfs_resource);
1596
1597 if (mntpt != NULL && mntpt[0] != '\0')
1598 mntpt = ZONE_PATH_TRANSLATE(mntpt, zone);
1599 else
1600 mntpt = "-";
1601
1602 error = lookupname(mntpt, UIO_SYSSPACE, FOLLOW, NULLVPP, &vp);
1603
1604 if (error != 0)
1605 goto nextp;
1606
1607 if (!(vp->v_flag & VROOT)) {
1608 VN_RELE(vp);
1609 goto nextp;
1610 }
1611 VN_RELE(vp);
1612
1613 if (resource != NULL && resource[0] != '\0') {
1614 if (resource[0] == '/') {
1615 resource = ZONE_PATH_VISIBLE(resource, zone) ?
1616 ZONE_PATH_TRANSLATE(resource, zone) :
1617 mntpt;
1618 }
1619 } else {
1620 resource = "-";
1621 }
1622
1623 lxpr_uiobuf_printf(uiobuf,
1624 "%s %s %s %s 0 0\n",
1625 resource, mntpt, vfssw[printp->vfs_fstype].vsw_name,
1626 printp->vfs_flag & VFS_RDONLY ? "ro" : "rw");
1627
1628 nextp:
1629 printp_next = printp->next;
1630 refstr_rele(printp->vfs_mntpt);
1631 refstr_rele(printp->vfs_resource);
1632 kmem_free(printp, sizeof (*printp));
1633 printp = printp_next;
1634
1635 }
1636 }
1637
1638 /*
1639 * lxpr_read_partitions():
1640 *
1641 * We don't support partitions in a local zone because it requires access to
1642 * physical devices. But we need to fake up enough of the file to show that we
1643 * have no partitions.
1644 */
1645 /* ARGSUSED */
1646 static void
1647 lxpr_read_partitions(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1648 {
1649 lxpr_uiobuf_printf(uiobuf,
1650 "major minor #blocks name rio rmerge rsect ruse "
1651 "wio wmerge wsect wuse running use aveq\n\n");
1652 }
1653
1654 /*
1655 * lxpr_read_version(): read the contents of the "version" file. Note that
1656 * we don't lie here -- we don't pretend that we're Linux. If lxproc is to
1657 * be used in a Linux-branded zone, there will need to be a mount option to
1658 * indicate that Linux should be more fully mimicked.
1659 */
1660 /* ARGSUSED */
1661 static void
1662 lxpr_read_version(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1663 {
1664 lxpr_uiobuf_printf(uiobuf,
1665 "%s version %s (%s version %d.%d.%d) "
1666 "#%s SMP %s\n",
1667 utsname.sysname, utsname.release,
1668 #if defined(__GNUC__)
1669 "gcc",
1670 __GNUC__,
1671 __GNUC_MINOR__,
1672 __GNUC_PATCHLEVEL__,
1673 #else
1674 "Sun C",
1675 __SUNPRO_C / 0x100,
1676 (__SUNPRO_C & 0xff) / 0x10,
1677 __SUNPRO_C & 0xf,
1678 #endif
1679 utsname.version,
1680 "00:00:00 00/00/00");
1681 }
1682
1683 /*
1684 * lxpr_read_stat(): read the contents of the "stat" file.
1685 *
1686 */
1687 /* ARGSUSED */
1688 static void
1689 lxpr_read_stat(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1690 {
1691 cpu_t *cp, *cpstart;
1692 int pools_enabled;
1693 ulong_t idle_cum = 0;
1694 ulong_t sys_cum = 0;
1695 ulong_t user_cum = 0;
1696 ulong_t irq_cum = 0;
1697 uint_t cpu_nrunnable_cum = 0;
1698 uint_t w_io_cum = 0;
1699
1700 ulong_t pgpgin_cum = 0;
1701 ulong_t pgpgout_cum = 0;
1702 ulong_t pgswapout_cum = 0;
1703 ulong_t pgswapin_cum = 0;
1704 ulong_t intr_cum = 0;
1705 ulong_t pswitch_cum = 0;
1706 ulong_t forks_cum = 0;
1707 hrtime_t msnsecs[NCMSTATES];
1708
1709 /* temporary variable since scalehrtime modifies data in place */
1710 hrtime_t tmptime;
1711
1712 ASSERT(lxpnp->lxpr_type == LXPR_STAT);
1713
1714 mutex_enter(&cpu_lock);
1715 pools_enabled = pool_pset_enabled();
1716
1717 /* Calculate cumulative stats */
1718 cp = cpstart = CPU->cpu_part->cp_cpulist;
1719 do {
1720 int i;
1721
1722 /*
1723 * Don't count CPUs that aren't even in the system
1724 * or aren't up yet.
1725 */
1726 if ((cp->cpu_flags & CPU_EXISTS) == 0) {
1727 continue;
1728 }
1729
1730 get_cpu_mstate(cp, msnsecs);
1731
1732 idle_cum += NSEC_TO_TICK(msnsecs[CMS_IDLE]);
1733 sys_cum += NSEC_TO_TICK(msnsecs[CMS_SYSTEM]);
1734 user_cum += NSEC_TO_TICK(msnsecs[CMS_USER]);
1735
1736 pgpgin_cum += CPU_STATS(cp, vm.pgpgin);
1737 pgpgout_cum += CPU_STATS(cp, vm.pgpgout);
1738 pgswapin_cum += CPU_STATS(cp, vm.pgswapin);
1739 pgswapout_cum += CPU_STATS(cp, vm.pgswapout);
1740
1741 cpu_nrunnable_cum += cp->cpu_disp->disp_nrunnable;
1742 w_io_cum += CPU_STATS(cp, sys.iowait);
1743 for (i = 0; i < NCMSTATES; i++) {
1744 tmptime = cp->cpu_intracct[i];
1745 scalehrtime(&tmptime);
1746 irq_cum += NSEC_TO_TICK(tmptime);
1747 }
1748
1749 for (i = 0; i < PIL_MAX; i++)
1750 intr_cum += CPU_STATS(cp, sys.intr[i]);
1751
1752 pswitch_cum += CPU_STATS(cp, sys.pswitch);
1753 forks_cum += CPU_STATS(cp, sys.sysfork);
1754 forks_cum += CPU_STATS(cp, sys.sysvfork);
1755
1756 if (pools_enabled)
1757 cp = cp->cpu_next_part;
1758 else
1759 cp = cp->cpu_next;
1760 } while (cp != cpstart);
1761
1762 lxpr_uiobuf_printf(uiobuf, "cpu %ld %ld %ld %ld %ld %ld %ld\n",
1763 user_cum, 0, sys_cum, idle_cum, 0, irq_cum, 0);
1764
1765 /* Do per processor stats */
1766 do {
1767 int i;
1768
1769 ulong_t idle_ticks;
1770 ulong_t sys_ticks;
1771 ulong_t user_ticks;
1772 ulong_t irq_ticks = 0;
1773
1774 /*
1775 * Don't count CPUs that aren't even in the system
1776 * or aren't up yet.
1777 */
1778 if ((cp->cpu_flags & CPU_EXISTS) == 0) {
1779 continue;
1780 }
1781
1782 get_cpu_mstate(cp, msnsecs);
1783
1784 idle_ticks = NSEC_TO_TICK(msnsecs[CMS_IDLE]);
1785 sys_ticks = NSEC_TO_TICK(msnsecs[CMS_SYSTEM]);
1786 user_ticks = NSEC_TO_TICK(msnsecs[CMS_USER]);
1787
1788 for (i = 0; i < NCMSTATES; i++) {
1789 tmptime = cp->cpu_intracct[i];
1790 scalehrtime(&tmptime);
1791 irq_ticks += NSEC_TO_TICK(tmptime);
1792 }
1793
1794 lxpr_uiobuf_printf(uiobuf,
1795 "cpu%d %ld %ld %ld %ld %ld %ld %ld\n",
1796 cp->cpu_id, user_ticks, 0, sys_ticks, idle_ticks,
1797 0, irq_ticks, 0);
1798
1799 if (pools_enabled)
1800 cp = cp->cpu_next_part;
1801 else
1802 cp = cp->cpu_next;
1803 } while (cp != cpstart);
1804
1805 mutex_exit(&cpu_lock);
1806
1807 lxpr_uiobuf_printf(uiobuf,
1808 "page %lu %lu\n"
1809 "swap %lu %lu\n"
1810 "intr %lu\n"
1811 "ctxt %lu\n"
1812 "btime %lu\n"
1813 "processes %lu\n"
1814 "procs_running %lu\n"
1815 "procs_blocked %lu\n",
1816 pgpgin_cum, pgpgout_cum,
1817 pgswapin_cum, pgswapout_cum,
1818 intr_cum,
1819 pswitch_cum,
1820 boot_time,
1821 forks_cum,
1822 cpu_nrunnable_cum,
1823 w_io_cum);
1824 }
1825
1826 /*
1827 * lxpr_read_uptime(): read the contents of the "uptime" file.
1828 *
1829 * format is: "%.2lf, %.2lf",uptime_secs, idle_secs
1830 * Use fixed point arithmetic to get 2 decimal places
1831 */
1832 /* ARGSUSED */
1833 static void
1834 lxpr_read_uptime(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1835 {
1836 cpu_t *cp, *cpstart;
1837 int pools_enabled;
1838 ulong_t idle_cum = 0;
1839 ulong_t cpu_count = 0;
1840 ulong_t idle_s;
1841 ulong_t idle_cs;
1842 ulong_t up_s;
1843 ulong_t up_cs;
1844 hrtime_t birthtime;
1845 hrtime_t centi_sec = 10000000; /* 10^7 */
1846
1847 ASSERT(lxpnp->lxpr_type == LXPR_UPTIME);
1848
1849 /* Calculate cumulative stats */
1850 mutex_enter(&cpu_lock);
1851 pools_enabled = pool_pset_enabled();
1852
1853 cp = cpstart = CPU;
1854 do {
1855 /*
1856 * Don't count CPUs that aren't even in the system
1857 * or aren't up yet.
1858 */
1859 if ((cp->cpu_flags & CPU_EXISTS) == 0) {
1860 continue;
1861 }
1862
1863 idle_cum += CPU_STATS(cp, sys.cpu_ticks_idle);
1864 idle_cum += CPU_STATS(cp, sys.cpu_ticks_wait);
1865 cpu_count += 1;
1866
1867 if (pools_enabled)
1868 cp = cp->cpu_next_part;
1869 else
1870 cp = cp->cpu_next;
1871 } while (cp != cpstart);
1872 mutex_exit(&cpu_lock);
1873
1874 /* Getting the Zone zsched process startup time */
1875 birthtime = LXPTOZ(lxpnp)->zone_zsched->p_mstart;
1876 up_cs = (gethrtime() - birthtime) / centi_sec;
1877 up_s = up_cs / 100;
1878 up_cs %= 100;
1879
1880 ASSERT(cpu_count > 0);
1881 idle_cum /= cpu_count;
1882 idle_s = idle_cum / hz;
1883 idle_cs = idle_cum % hz;
1884 idle_cs *= 100;
1885 idle_cs /= hz;
1886
1887 lxpr_uiobuf_printf(uiobuf,
1888 "%ld.%02d %ld.%02d\n", up_s, up_cs, idle_s, idle_cs);
1889 }
1890
1891 static const char *amd_x_edx[] = {
1892 NULL, NULL, NULL, NULL,
1893 NULL, NULL, NULL, NULL,
1894 NULL, NULL, NULL, "syscall",
1895 NULL, NULL, NULL, NULL,
1896 NULL, NULL, NULL, "mp",
1897 "nx", NULL, "mmxext", NULL,
1898 NULL, NULL, NULL, NULL,
1899 NULL, "lm", "3dnowext", "3dnow"
1900 };
1901
1902 static const char *amd_x_ecx[] = {
1903 "lahf_lm", NULL, "svm", NULL,
1904 "altmovcr8"
1905 };
1906
1907 static const char *tm_x_edx[] = {
1908 "recovery", "longrun", NULL, "lrti"
1909 };
1910
1911 /*
1912 * Intel calls no-execute "xd" in its docs, but Linux still reports it as "nx."
1913 */
1914 static const char *intc_x_edx[] = {
1915 NULL, NULL, NULL, NULL,
1916 NULL, NULL, NULL, NULL,
1917 NULL, NULL, NULL, "syscall",
1918 NULL, NULL, NULL, NULL,
1919 NULL, NULL, NULL, NULL,
1920 "nx", NULL, NULL, NULL,
1921 NULL, NULL, NULL, NULL,
1922 NULL, "lm", NULL, NULL
1923 };
1924
1925 static const char *intc_edx[] = {
1926 "fpu", "vme", "de", "pse",
1927 "tsc", "msr", "pae", "mce",
1928 "cx8", "apic", NULL, "sep",
1929 "mtrr", "pge", "mca", "cmov",
1930 "pat", "pse36", "pn", "clflush",
1931 NULL, "dts", "acpi", "mmx",
1932 "fxsr", "sse", "sse2", "ss",
1933 "ht", "tm", "ia64", "pbe"
1934 };
1935
1936 /*
1937 * "sse3" on linux is called "pni" (Prescott New Instructions).
1938 */
1939 static const char *intc_ecx[] = {
1940 "pni", NULL, NULL, "monitor",
1941 "ds_cpl", NULL, NULL, "est",
1942 "tm2", NULL, "cid", NULL,
1943 NULL, "cx16", "xtpr"
1944 };
1945
1946 static void
1947 lxpr_read_cpuinfo(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
1948 {
1949 int i;
1950 uint32_t bits;
1951 cpu_t *cp, *cpstart;
1952 int pools_enabled;
1953 const char **fp;
1954 char brandstr[CPU_IDSTRLEN];
1955 struct cpuid_regs cpr;
1956 int maxeax;
1957 int std_ecx, std_edx, ext_ecx, ext_edx;
1958
1959 ASSERT(lxpnp->lxpr_type == LXPR_CPUINFO);
1960
1961 mutex_enter(&cpu_lock);
1962 pools_enabled = pool_pset_enabled();
1963
1964 cp = cpstart = CPU;
1965 do {
1966 /*
1967 * This returns the maximum eax value for standard cpuid
1968 * functions in eax.
1969 */
1970 cpr.cp_eax = 0;
1971 (void) cpuid_insn(cp, &cpr);
1972 maxeax = cpr.cp_eax;
1973
1974 /*
1975 * Get standard x86 feature flags.
1976 */
1977 cpr.cp_eax = 1;
1978 (void) cpuid_insn(cp, &cpr);
1979 std_ecx = cpr.cp_ecx;
1980 std_edx = cpr.cp_edx;
1981
1982 /*
1983 * Now get extended feature flags.
1984 */
1985 cpr.cp_eax = 0x80000001;
1986 (void) cpuid_insn(cp, &cpr);
1987 ext_ecx = cpr.cp_ecx;
1988 ext_edx = cpr.cp_edx;
1989
1990 (void) cpuid_getbrandstr(cp, brandstr, CPU_IDSTRLEN);
1991
1992 lxpr_uiobuf_printf(uiobuf,
1993 "processor\t: %d\n"
1994 "vendor_id\t: %s\n"
1995 "cpu family\t: %d\n"
1996 "model\t\t: %d\n"
1997 "model name\t: %s\n"
1998 "stepping\t: %d\n"
1999 "cpu MHz\t\t: %u.%03u\n",
2000 cp->cpu_id, cpuid_getvendorstr(cp), cpuid_getfamily(cp),
2001 cpuid_getmodel(cp), brandstr, cpuid_getstep(cp),
2002 (uint32_t)(cpu_freq_hz / 1000000),
2003 ((uint32_t)(cpu_freq_hz / 1000)) % 1000);
2004
2005 lxpr_uiobuf_printf(uiobuf, "cache size\t: %u KB\n",
2006 getl2cacheinfo(cp, NULL, NULL, NULL) / 1024);
2007
2008 if (is_x86_feature(x86_featureset, X86FSET_HTT)) {
2009 /*
2010 * 'siblings' is used for HT-style threads
2011 */
2012 lxpr_uiobuf_printf(uiobuf,
2013 "physical id\t: %lu\n"
2014 "siblings\t: %u\n",
2015 pg_plat_hw_instance_id(cp, PGHW_CHIP),
2016 cpuid_get_ncpu_per_chip(cp));
2017 }
2018
2019 /*
2020 * Since we're relatively picky about running on older hardware,
2021 * we can be somewhat cavalier about the answers to these ones.
2022 *
2023 * In fact, given the hardware we support, we just say:
2024 *
2025 * fdiv_bug : no (if we're on a 64-bit kernel)
2026 * hlt_bug : no
2027 * f00f_bug : no
2028 * coma_bug : no
2029 * wp : yes (write protect in supervsr mode)
2030 */
2031 lxpr_uiobuf_printf(uiobuf,
2032 "fdiv_bug\t: %s\n"
2033 "hlt_bug \t: no\n"
2034 "f00f_bug\t: no\n"
2035 "coma_bug\t: no\n"
2036 "fpu\t\t: %s\n"
2037 "fpu_exception\t: %s\n"
2038 "cpuid level\t: %d\n"
2039 "flags\t\t:",
2040 #if defined(__i386)
2041 fpu_pentium_fdivbug ? "yes" : "no",
2042 #else
2043 "no",
2044 #endif /* __i386 */
2045 fpu_exists ? "yes" : "no", fpu_exists ? "yes" : "no",
2046 maxeax);
2047
2048 for (bits = std_edx, fp = intc_edx, i = 0;
2049 i < sizeof (intc_edx) / sizeof (intc_edx[0]); fp++, i++)
2050 if ((bits & (1 << i)) != 0 && *fp)
2051 lxpr_uiobuf_printf(uiobuf, " %s", *fp);
2052
2053 /*
2054 * name additional features where appropriate
2055 */
2056 switch (x86_vendor) {
2057 case X86_VENDOR_Intel:
2058 for (bits = ext_edx, fp = intc_x_edx, i = 0;
2059 i < sizeof (intc_x_edx) / sizeof (intc_x_edx[0]);
2060 fp++, i++)
2061 if ((bits & (1 << i)) != 0 && *fp)
2062 lxpr_uiobuf_printf(uiobuf, " %s", *fp);
2063 break;
2064
2065 case X86_VENDOR_AMD:
2066 for (bits = ext_edx, fp = amd_x_edx, i = 0;
2067 i < sizeof (amd_x_edx) / sizeof (amd_x_edx[0]);
2068 fp++, i++)
2069 if ((bits & (1 << i)) != 0 && *fp)
2070 lxpr_uiobuf_printf(uiobuf, " %s", *fp);
2071
2072 for (bits = ext_ecx, fp = amd_x_ecx, i = 0;
2073 i < sizeof (amd_x_ecx) / sizeof (amd_x_ecx[0]);
2074 fp++, i++)
2075 if ((bits & (1 << i)) != 0 && *fp)
2076 lxpr_uiobuf_printf(uiobuf, " %s", *fp);
2077 break;
2078
2079 case X86_VENDOR_TM:
2080 for (bits = ext_edx, fp = tm_x_edx, i = 0;
2081 i < sizeof (tm_x_edx) / sizeof (tm_x_edx[0]);
2082 fp++, i++)
2083 if ((bits & (1 << i)) != 0 && *fp)
2084 lxpr_uiobuf_printf(uiobuf, " %s", *fp);
2085 break;
2086 default:
2087 break;
2088 }
2089
2090 for (bits = std_ecx, fp = intc_ecx, i = 0;
2091 i < sizeof (intc_ecx) / sizeof (intc_ecx[0]); fp++, i++)
2092 if ((bits & (1 << i)) != 0 && *fp)
2093 lxpr_uiobuf_printf(uiobuf, " %s", *fp);
2094
2095 lxpr_uiobuf_printf(uiobuf, "\n\n");
2096
2097 if (pools_enabled)
2098 cp = cp->cpu_next_part;
2099 else
2100 cp = cp->cpu_next;
2101 } while (cp != cpstart);
2102
2103 mutex_exit(&cpu_lock);
2104 }
2105
2106 /* ARGSUSED */
2107 static void
2108 lxpr_read_fd(lxpr_node_t *lxpnp, lxpr_uiobuf_t *uiobuf)
2109 {
2110 ASSERT(lxpnp->lxpr_type == LXPR_PID_FD_FD);
2111 lxpr_uiobuf_seterr(uiobuf, EFAULT);
2112 }
2113
2114 /*
2115 * lxpr_getattr(): Vnode operation for VOP_GETATTR()
2116 */
2117 static int
2118 lxpr_getattr(vnode_t *vp, vattr_t *vap, int flags, cred_t *cr,
2119 caller_context_t *ct)
2120 {
2121 register lxpr_node_t *lxpnp = VTOLXP(vp);
2122 lxpr_nodetype_t type = lxpnp->lxpr_type;
2123 extern uint_t nproc;
2124 int error;
2125
2126 /*
2127 * Return attributes of underlying vnode if ATTR_REAL
2128 *
2129 * but keep fd files with the symlink permissions
2130 */
2131 if (lxpnp->lxpr_realvp != NULL && (flags & ATTR_REAL)) {
2132 vnode_t *rvp = lxpnp->lxpr_realvp;
2133
2134 /*
2135 * withold attribute information to owner or root
2136 */
2137 if ((error = VOP_ACCESS(rvp, 0, 0, cr, ct)) != 0) {
2138 return (error);
2139 }
2140
2141 /*
2142 * now its attributes
2143 */
2144 if ((error = VOP_GETATTR(rvp, vap, flags, cr, ct)) != 0) {
2145 return (error);
2146 }
2147
2148 /*
2149 * if it's a file in lx /proc/pid/fd/xx then set its
2150 * mode and keep it looking like a symlink
2151 */
2152 if (type == LXPR_PID_FD_FD) {
2153 vap->va_mode = lxpnp->lxpr_mode;
2154 vap->va_type = vp->v_type;
2155 vap->va_size = 0;
2156 vap->va_nlink = 1;
2157 }
2158 return (0);
2159 }
2160
2161 /* Default attributes, that may be overridden below */
2162 bzero(vap, sizeof (*vap));
2163 vap->va_atime = vap->va_mtime = vap->va_ctime = lxpnp->lxpr_time;
2164 vap->va_nlink = 1;
2165 vap->va_type = vp->v_type;
2166 vap->va_mode = lxpnp->lxpr_mode;
2167 vap->va_fsid = vp->v_vfsp->vfs_dev;
2168 vap->va_blksize = DEV_BSIZE;
2169 vap->va_uid = lxpnp->lxpr_uid;
2170 vap->va_gid = lxpnp->lxpr_gid;
2171 vap->va_nodeid = lxpnp->lxpr_ino;
2172
2173 switch (type) {
2174 case LXPR_PROCDIR:
2175 vap->va_nlink = nproc + 2 + PROCDIRFILES;
2176 vap->va_size = (nproc + 2 + PROCDIRFILES) * LXPR_SDSIZE;
2177 break;
2178 case LXPR_PIDDIR:
2179 vap->va_nlink = PIDDIRFILES;
2180 vap->va_size = PIDDIRFILES * LXPR_SDSIZE;
2181 break;
2182 case LXPR_SELF:
2183 vap->va_uid = crgetruid(curproc->p_cred);
2184 vap->va_gid = crgetrgid(curproc->p_cred);
2185 break;
2186 default:
2187 break;
2188 }
2189
2190 vap->va_nblocks = (fsblkcnt64_t)btod(vap->va_size);
2191 return (0);
2192 }
2193
2194 /*
2195 * lxpr_access(): Vnode operation for VOP_ACCESS()
2196 */
2197 static int
2198 lxpr_access(vnode_t *vp, int mode, int flags, cred_t *cr, caller_context_t *ct)
2199 {
2200 lxpr_node_t *lxpnp = VTOLXP(vp);
2201 int shift = 0;
2202 proc_t *tp;
2203
2204 /* lx /proc is a read only file system */
2205 if (mode & VWRITE)
2206 return (EROFS);
2207
2208 /*
2209 * If this is a restricted file, check access permissions.
2210 */
2211 switch (lxpnp->lxpr_type) {
2212 case LXPR_PIDDIR:
2213 return (0);
2214 case LXPR_PID_CURDIR:
2215 case LXPR_PID_ENV:
2216 case LXPR_PID_EXE:
2217 case LXPR_PID_MAPS:
2218 case LXPR_PID_MEM:
2219 case LXPR_PID_ROOTDIR:
2220 case LXPR_PID_FDDIR:
2221 case LXPR_PID_FD_FD:
2222 if ((tp = lxpr_lock(lxpnp->lxpr_pid)) == NULL)
2223 return (ENOENT);
2224 if (tp != curproc && secpolicy_proc_access(cr) != 0 &&
2225 priv_proc_cred_perm(cr, tp, NULL, mode) != 0) {
2226 lxpr_unlock(tp);
2227 return (EACCES);
2228 }
2229 lxpr_unlock(tp);
2230 default:
2231 break;
2232 }
2233
2234 if (lxpnp->lxpr_realvp != NULL) {
2235 /*
2236 * For these we use the underlying vnode's accessibility.
2237 */
2238 return (VOP_ACCESS(lxpnp->lxpr_realvp, mode, flags, cr, ct));
2239 }
2240
2241 /* If user is root allow access regardless of permission bits */
2242 if (secpolicy_proc_access(cr) == 0)
2243 return (0);
2244
2245 /*
2246 * Access check is based on only one of owner, group, public. If not
2247 * owner, then check group. If not a member of the group, then check
2248 * public access.
2249 */
2250 if (crgetuid(cr) != lxpnp->lxpr_uid) {
2251 shift += 3;
2252 if (!groupmember((uid_t)lxpnp->lxpr_gid, cr))
2253 shift += 3;
2254 }
2255
2256 mode &= ~(lxpnp->lxpr_mode << shift);
2257
2258 if (mode == 0)
2259 return (0);
2260
2261 return (EACCES);
2262 }
2263
2264 /* ARGSUSED */
2265 static vnode_t *
2266 lxpr_lookup_not_a_dir(vnode_t *dp, char *comp)
2267 {
2268 return (NULL);
2269 }
2270
2271 /*
2272 * lxpr_lookup(): Vnode operation for VOP_LOOKUP()
2273 */
2274 /* ARGSUSED */
2275 static int
2276 lxpr_lookup(vnode_t *dp, char *comp, vnode_t **vpp, pathname_t *pathp,
2277 int flags, vnode_t *rdir, cred_t *cr, caller_context_t *ct,
2278 int *direntflags, pathname_t *realpnp)
2279 {
2280 lxpr_node_t *lxpnp = VTOLXP(dp);
2281 lxpr_nodetype_t type = lxpnp->lxpr_type;
2282 int error;
2283
2284 ASSERT(dp->v_type == VDIR);
2285 ASSERT(type >= 0 && type < LXPR_NFILES);
2286
2287 /*
2288 * we should never get here because the lookup
2289 * is done on the realvp for these nodes
2290 */
2291 ASSERT(type != LXPR_PID_FD_FD &&
2292 type != LXPR_PID_CURDIR &&
2293 type != LXPR_PID_ROOTDIR);
2294
2295 /*
2296 * restrict lookup permission to owner or root
2297 */
2298 if ((error = lxpr_access(dp, VEXEC, 0, cr, ct)) != 0) {
2299 return (error);
2300 }
2301
2302 /*
2303 * Just return the parent vnode if that's where we are trying to go.
2304 */
2305 if (strcmp(comp, "..") == 0) {
2306 VN_HOLD(lxpnp->lxpr_parent);
2307 *vpp = lxpnp->lxpr_parent;
2308 return (0);
2309 }
2310
2311 /*
2312 * Special handling for directory searches. Note: null component name
2313 * denotes that the current directory is being searched.
2314 */
2315 if ((dp->v_type == VDIR) && (*comp == '\0' || strcmp(comp, ".") == 0)) {
2316 VN_HOLD(dp);
2317 *vpp = dp;
2318 return (0);
2319 }
2320
2321 *vpp = (lxpr_lookup_function[type](dp, comp));
2322 return ((*vpp == NULL) ? ENOENT : 0);
2323 }
2324
2325 /*
2326 * Do a sequential search on the given directory table
2327 */
2328 static vnode_t *
2329 lxpr_lookup_common(vnode_t *dp, char *comp, proc_t *p,
2330 lxpr_dirent_t *dirtab, int dirtablen)
2331 {
2332 lxpr_node_t *lxpnp;
2333 int count;
2334
2335 for (count = 0; count < dirtablen; count++) {
2336 if (strcmp(dirtab[count].d_name, comp) == 0) {
2337 lxpnp = lxpr_getnode(dp, dirtab[count].d_type, p, 0);
2338 dp = LXPTOV(lxpnp);
2339 ASSERT(dp != NULL);
2340 return (dp);
2341 }
2342 }
2343 return (NULL);
2344 }
2345
2346 static vnode_t *
2347 lxpr_lookup_piddir(vnode_t *dp, char *comp)
2348 {
2349 proc_t *p;
2350
2351 ASSERT(VTOLXP(dp)->lxpr_type == LXPR_PIDDIR);
2352
2353 p = lxpr_lock(VTOLXP(dp)->lxpr_pid);
2354 if (p == NULL)
2355 return (NULL);
2356
2357 dp = lxpr_lookup_common(dp, comp, p, piddir, PIDDIRFILES);
2358
2359 lxpr_unlock(p);
2360
2361 return (dp);
2362 }
2363
2364 /*
2365 * Lookup one of the process's open files.
2366 */
2367 static vnode_t *
2368 lxpr_lookup_fddir(vnode_t *dp, char *comp)
2369 {
2370 lxpr_node_t *dlxpnp = VTOLXP(dp);
2371 lxpr_node_t *lxpnp;
2372 vnode_t *vp = NULL;
2373 proc_t *p;
2374 file_t *fp;
2375 uint_t fd;
2376 int c;
2377 uf_entry_t *ufp;
2378 uf_info_t *fip;
2379
2380 ASSERT(dlxpnp->lxpr_type == LXPR_PID_FDDIR);
2381
2382 /*
2383 * convert the string rendition of the filename
2384 * to a file descriptor
2385 */
2386 fd = 0;
2387 while ((c = *comp++) != '\0') {
2388 int ofd;
2389 if (c < '0' || c > '9')
2390 return (NULL);
2391
2392 ofd = fd;
2393 fd = 10*fd + c - '0';
2394 /* integer overflow */
2395 if (fd / 10 != ofd)
2396 return (NULL);
2397 }
2398
2399 /*
2400 * get the proc to work with and lock it
2401 */
2402 p = lxpr_lock(dlxpnp->lxpr_pid);
2403 if ((p == NULL))
2404 return (NULL);
2405
2406 /*
2407 * If the process is a zombie or system process
2408 * it can't have any open files.
2409 */
2410 if ((p->p_stat == SZOMB) || (p->p_flag & SSYS) || (p->p_as == &kas)) {
2411 lxpr_unlock(p);
2412 return (NULL);
2413 }
2414
2415 /*
2416 * get us a fresh node/vnode
2417 */
2418 lxpnp = lxpr_getnode(dp, LXPR_PID_FD_FD, p, fd);
2419
2420 /*
2421 * get open file info
2422 */
2423 fip = (&(p)->p_user.u_finfo);
2424 mutex_enter(&fip->fi_lock);
2425
2426 /*
2427 * got the fd data so now done with this proc
2428 */
2429 lxpr_unlock(p);
2430
2431 if (fd < fip->fi_nfiles) {
2432 UF_ENTER(ufp, fip, fd);
2433 /*
2434 * ensure the fd is still kosher.
2435 * it may have gone between the readdir and
2436 * the lookup
2437 */
2438 if (fip->fi_list[fd].uf_file == NULL) {
2439 mutex_exit(&fip->fi_lock);
2440 UF_EXIT(ufp);
2441 lxpr_freenode(lxpnp);
2442 return (NULL);
2443 }
2444
2445 if ((fp = ufp->uf_file) != NULL)
2446 vp = fp->f_vnode;
2447 UF_EXIT(ufp);
2448 }
2449 mutex_exit(&fip->fi_lock);
2450
2451 if (vp == NULL) {
2452 lxpr_freenode(lxpnp);
2453 return (NULL);
2454 } else {
2455 /*
2456 * Fill in the lxpr_node so future references will be able to
2457 * find the underlying vnode. The vnode is held on the realvp.
2458 */
2459 lxpnp->lxpr_realvp = vp;
2460 VN_HOLD(lxpnp->lxpr_realvp);
2461 }
2462
2463 dp = LXPTOV(lxpnp);
2464 ASSERT(dp != NULL);
2465
2466 return (dp);
2467 }
2468
2469 static vnode_t *
2470 lxpr_lookup_netdir(vnode_t *dp, char *comp)
2471 {
2472 ASSERT(VTOLXP(dp)->lxpr_type == LXPR_NETDIR);
2473
2474 dp = lxpr_lookup_common(dp, comp, NULL, netdir, NETDIRFILES);
2475
2476 return (dp);
2477 }
2478
2479 static vnode_t *
2480 lxpr_lookup_procdir(vnode_t *dp, char *comp)
2481 {
2482 ASSERT(VTOLXP(dp)->lxpr_type == LXPR_PROCDIR);
2483
2484 /*
2485 * We know all the names of files & dirs in our file system structure
2486 * except those that are pid names. These change as pids are created/
2487 * deleted etc., so we just look for a number as the first char to see
2488 * if we are we doing pid lookups.
2489 *
2490 * Don't need to check for "self" as it is implemented as a symlink
2491 */
2492 if (*comp >= '0' && *comp <= '9') {
2493 pid_t pid = 0;
2494 lxpr_node_t *lxpnp = NULL;
2495 proc_t *p;
2496 int c;
2497
2498 while ((c = *comp++) != '\0')
2499 pid = 10 * pid + c - '0';
2500
2501 /*
2502 * Can't continue if the process is still loading or it doesn't
2503 * really exist yet (or maybe it just died!)
2504 */
2505 p = lxpr_lock(pid);
2506 if (p == NULL)
2507 return (NULL);
2508
2509 if (secpolicy_basic_procinfo(CRED(), p, curproc) != 0) {
2510 lxpr_unlock(p);
2511 return (NULL);
2512 }
2513
2514 /*
2515 * allocate and fill in a new lxpr node
2516 */
2517 lxpnp = lxpr_getnode(dp, LXPR_PIDDIR, p, 0);
2518
2519 lxpr_unlock(p);
2520
2521 dp = LXPTOV(lxpnp);
2522 ASSERT(dp != NULL);
2523
2524 return (dp);
2525 }
2526
2527 /* Lookup fixed names */
2528 return (lxpr_lookup_common(dp, comp, NULL, lxpr_dir, PROCDIRFILES));
2529 }
2530
2531 /*
2532 * lxpr_readdir(): Vnode operation for VOP_READDIR()
2533 */
2534 /* ARGSUSED */
2535 static int
2536 lxpr_readdir(vnode_t *dp, uio_t *uiop, cred_t *cr, int *eofp,
2537 caller_context_t *ct, int flags)
2538 {
2539 lxpr_node_t *lxpnp = VTOLXP(dp);
2540 lxpr_nodetype_t type = lxpnp->lxpr_type;
2541 ssize_t uresid;
2542 off_t uoffset;
2543 int error;
2544
2545 ASSERT(dp->v_type == VDIR);
2546 ASSERT(type >= 0 && type < LXPR_NFILES);
2547
2548 /*
2549 * we should never get here because the readdir
2550 * is done on the realvp for these nodes
2551 */
2552 ASSERT(type != LXPR_PID_FD_FD &&
2553 type != LXPR_PID_CURDIR &&
2554 type != LXPR_PID_ROOTDIR);
2555
2556 /*
2557 * restrict readdir permission to owner or root
2558 */
2559 if ((error = lxpr_access(dp, VREAD, 0, cr, ct)) != 0)
2560 return (error);
2561
2562 uoffset = uiop->uio_offset;
2563 uresid = uiop->uio_resid;
2564
2565 /* can't do negative reads */
2566 if (uoffset < 0 || uresid <= 0)
2567 return (EINVAL);
2568
2569 /* can't read directory entries that don't exist! */
2570 if (uoffset % LXPR_SDSIZE)
2571 return (ENOENT);
2572
2573 return (lxpr_readdir_function[lxpnp->lxpr_type](lxpnp, uiop, eofp));
2574 }
2575
2576 /* ARGSUSED */
2577 static int
2578 lxpr_readdir_not_a_dir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp)
2579 {
2580 return (ENOTDIR);
2581 }
2582
2583 /*
2584 * This has the common logic for returning directory entries
2585 */
2586 static int
2587 lxpr_readdir_common(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp,
2588 lxpr_dirent_t *dirtab, int dirtablen)
2589 {
2590 /* bp holds one dirent64 structure */
2591 longlong_t bp[DIRENT64_RECLEN(LXPNSIZ) / sizeof (longlong_t)];
2592 dirent64_t *dirent = (dirent64_t *)bp;
2593 ssize_t oresid; /* save a copy for testing later */
2594 ssize_t uresid;
2595
2596 oresid = uiop->uio_resid;
2597
2598 /* clear out the dirent buffer */
2599 bzero(bp, sizeof (bp));
2600
2601 /*
2602 * Satisfy user request
2603 */
2604 while ((uresid = uiop->uio_resid) > 0) {
2605 int dirindex;
2606 off_t uoffset;
2607 int reclen;
2608 int error;
2609
2610 uoffset = uiop->uio_offset;
2611 dirindex = (uoffset / LXPR_SDSIZE) - 2;
2612
2613 if (uoffset == 0) {
2614
2615 dirent->d_ino = lxpnp->lxpr_ino;
2616 dirent->d_name[0] = '.';
2617 dirent->d_name[1] = '\0';
2618 reclen = DIRENT64_RECLEN(1);
2619
2620 } else if (uoffset == LXPR_SDSIZE) {
2621
2622 dirent->d_ino = lxpr_parentinode(lxpnp);
2623 dirent->d_name[0] = '.';
2624 dirent->d_name[1] = '.';
2625 dirent->d_name[2] = '\0';
2626 reclen = DIRENT64_RECLEN(2);
2627
2628 } else if (dirindex < dirtablen) {
2629 int slen = strlen(dirtab[dirindex].d_name);
2630
2631 dirent->d_ino = lxpr_inode(dirtab[dirindex].d_type,
2632 lxpnp->lxpr_pid, 0);
2633
2634 ASSERT(slen < LXPNSIZ);
2635 (void) strcpy(dirent->d_name, dirtab[dirindex].d_name);
2636 reclen = DIRENT64_RECLEN(slen);
2637
2638 } else {
2639 /* Run out of table entries */
2640 if (eofp) {
2641 *eofp = 1;
2642 }
2643 return (0);
2644 }
2645
2646 dirent->d_off = (off64_t)(uoffset + LXPR_SDSIZE);
2647 dirent->d_reclen = (ushort_t)reclen;
2648
2649 /*
2650 * if the size of the data to transfer is greater
2651 * that that requested then we can't do it this transfer.
2652 */
2653 if (reclen > uresid) {
2654 /*
2655 * Error if no entries have been returned yet.
2656 */
2657 if (uresid == oresid) {
2658 return (EINVAL);
2659 }
2660 break;
2661 }
2662
2663 /*
2664 * uiomove() updates both uiop->uio_resid and uiop->uio_offset
2665 * by the same amount. But we want uiop->uio_offset to change
2666 * in increments of LXPR_SDSIZE, which is different from the
2667 * number of bytes being returned to the user. So we set
2668 * uiop->uio_offset separately, ignoring what uiomove() does.
2669 */
2670 if ((error = uiomove((caddr_t)dirent, reclen, UIO_READ,
2671 uiop)) != 0)
2672 return (error);
2673
2674 uiop->uio_offset = uoffset + LXPR_SDSIZE;
2675 }
2676
2677 /* Have run out of space, but could have just done last table entry */
2678 if (eofp) {
2679 *eofp =
2680 (uiop->uio_offset >= ((dirtablen+2) * LXPR_SDSIZE)) ? 1 : 0;
2681 }
2682 return (0);
2683 }
2684
2685
2686 static int
2687 lxpr_readdir_procdir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp)
2688 {
2689 /* bp holds one dirent64 structure */
2690 longlong_t bp[DIRENT64_RECLEN(LXPNSIZ) / sizeof (longlong_t)];
2691 dirent64_t *dirent = (dirent64_t *)bp;
2692 ssize_t oresid; /* save a copy for testing later */
2693 ssize_t uresid;
2694 off_t uoffset;
2695 zoneid_t zoneid;
2696 pid_t pid;
2697 int error;
2698 int ceof;
2699
2700 ASSERT(lxpnp->lxpr_type == LXPR_PROCDIR);
2701
2702 oresid = uiop->uio_resid;
2703 zoneid = LXPTOZ(lxpnp)->zone_id;
2704
2705 /*
2706 * We return directory entries in the order: "." and ".." then the
2707 * unique lxproc files, then the directories corresponding to the
2708 * running processes. We have defined this as the ordering because
2709 * it allows us to more easily keep track of where we are betwen calls
2710 * to getdents(). If the number of processes changes between calls
2711 * then we can't lose track of where we are in the lxproc files.
2712 */
2713
2714 /* Do the fixed entries */
2715 error = lxpr_readdir_common(lxpnp, uiop, &ceof, lxpr_dir,
2716 PROCDIRFILES);
2717
2718 /* Finished if we got an error or if we couldn't do all the table */
2719 if (error != 0 || ceof == 0)
2720 return (error);
2721
2722 /* clear out the dirent buffer */
2723 bzero(bp, sizeof (bp));
2724
2725 /* Do the process entries */
2726 while ((uresid = uiop->uio_resid) > 0) {
2727 proc_t *p;
2728 int len;
2729 int reclen;
2730 int i;
2731
2732 uoffset = uiop->uio_offset;
2733
2734 /*
2735 * Stop when entire proc table has been examined.
2736 */
2737 i = (uoffset / LXPR_SDSIZE) - 2 - PROCDIRFILES;
2738 if (i >= v.v_proc) {
2739 /* Run out of table entries */
2740 if (eofp) {
2741 *eofp = 1;
2742 }
2743 return (0);
2744 }
2745 mutex_enter(&pidlock);
2746
2747 /*
2748 * Skip indices for which there is no pid_entry, PIDs for
2749 * which there is no corresponding process, a PID of 0,
2750 * and anything the security policy doesn't allow
2751 * us to look at.
2752 */
2753 if ((p = pid_entry(i)) == NULL || p->p_stat == SIDL ||
2754 p->p_pid == 0 ||
2755 secpolicy_basic_procinfo(CRED(), p, curproc) != 0) {
2756 mutex_exit(&pidlock);
2757 goto next;
2758 }
2759 mutex_exit(&pidlock);
2760
2761 /*
2762 * Convert pid to the Linux default of 1 if we're the zone's
2763 * init process, otherwise use the value from the proc
2764 * structure
2765 */
2766 pid = ((p->p_pid != curproc->p_zone->zone_proc_initpid) ?
2767 p->p_pid : 1);
2768
2769 /*
2770 * If this /proc was mounted in the global zone, view
2771 * all procs; otherwise, only view zone member procs.
2772 */
2773 if (zoneid != GLOBAL_ZONEID && p->p_zone->zone_id != zoneid) {
2774 goto next;
2775 }
2776
2777 ASSERT(p->p_stat != 0);
2778
2779 dirent->d_ino = lxpr_inode(LXPR_PIDDIR, pid, 0);
2780 len = snprintf(dirent->d_name, LXPNSIZ, "%d", pid);
2781 ASSERT(len < LXPNSIZ);
2782 reclen = DIRENT64_RECLEN(len);
2783
2784 dirent->d_off = (off64_t)(uoffset + LXPR_SDSIZE);
2785 dirent->d_reclen = (ushort_t)reclen;
2786
2787 /*
2788 * if the size of the data to transfer is greater
2789 * that that requested then we can't do it this transfer.
2790 */
2791 if (reclen > uresid) {
2792 /*
2793 * Error if no entries have been returned yet.
2794 */
2795 if (uresid == oresid)
2796 return (EINVAL);
2797 break;
2798 }
2799
2800 /*
2801 * uiomove() updates both uiop->uio_resid and uiop->uio_offset
2802 * by the same amount. But we want uiop->uio_offset to change
2803 * in increments of LXPR_SDSIZE, which is different from the
2804 * number of bytes being returned to the user. So we set
2805 * uiop->uio_offset separately, in the increment of this for
2806 * the loop, ignoring what uiomove() does.
2807 */
2808 if ((error = uiomove((caddr_t)dirent, reclen, UIO_READ,
2809 uiop)) != 0)
2810 return (error);
2811 next:
2812 uiop->uio_offset = uoffset + LXPR_SDSIZE;
2813 }
2814
2815 if (eofp != NULL) {
2816 *eofp = (uiop->uio_offset >=
2817 ((v.v_proc + PROCDIRFILES + 2) * LXPR_SDSIZE)) ? 1 : 0;
2818 }
2819
2820 return (0);
2821 }
2822
2823 static int
2824 lxpr_readdir_piddir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp)
2825 {
2826 proc_t *p;
2827
2828 ASSERT(lxpnp->lxpr_type == LXPR_PIDDIR);
2829
2830 /* can't read its contents if it died */
2831 mutex_enter(&pidlock);
2832
2833 p = prfind((lxpnp->lxpr_pid == 1) ?
2834 curproc->p_zone->zone_proc_initpid : lxpnp->lxpr_pid);
2835
2836 if (p == NULL || p->p_stat == SIDL) {
2837 mutex_exit(&pidlock);
2838 return (ENOENT);
2839 }
2840 mutex_exit(&pidlock);
2841
2842 return (lxpr_readdir_common(lxpnp, uiop, eofp, piddir, PIDDIRFILES));
2843 }
2844
2845 static int
2846 lxpr_readdir_netdir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp)
2847 {
2848 ASSERT(lxpnp->lxpr_type == LXPR_NETDIR);
2849 return (lxpr_readdir_common(lxpnp, uiop, eofp, netdir, NETDIRFILES));
2850 }
2851
2852 static int
2853 lxpr_readdir_fddir(lxpr_node_t *lxpnp, uio_t *uiop, int *eofp)
2854 {
2855 /* bp holds one dirent64 structure */
2856 longlong_t bp[DIRENT64_RECLEN(LXPNSIZ) / sizeof (longlong_t)];
2857 dirent64_t *dirent = (dirent64_t *)bp;
2858 ssize_t oresid; /* save a copy for testing later */
2859 ssize_t uresid;
2860 off_t uoffset;
2861 int error;
2862 int ceof;
2863 proc_t *p;
2864 int fddirsize;
2865 uf_info_t *fip;
2866
2867 ASSERT(lxpnp->lxpr_type == LXPR_PID_FDDIR);
2868
2869 oresid = uiop->uio_resid;
2870
2871 /* can't read its contents if it died */
2872 p = lxpr_lock(lxpnp->lxpr_pid);
2873 if (p == NULL)
2874 return (ENOENT);
2875
2876 /* Get open file info */
2877 fip = (&(p)->p_user.u_finfo);
2878
2879 if ((p->p_stat == SZOMB) || (p->p_flag & SSYS) || (p->p_as == &kas)) {
2880 fddirsize = 0;
2881 } else {
2882 fddirsize = fip->fi_nfiles;
2883 }
2884
2885 mutex_enter(&fip->fi_lock);
2886 lxpr_unlock(p);
2887
2888 /* Do the fixed entries (in this case just "." & "..") */
2889 error = lxpr_readdir_common(lxpnp, uiop, &ceof, 0, 0);
2890
2891 /* Finished if we got an error or if we couldn't do all the table */
2892 if (error != 0 || ceof == 0)
2893 return (error);
2894
2895 /* clear out the dirent buffer */
2896 bzero(bp, sizeof (bp));
2897
2898 /*
2899 * Loop until user's request is satisfied or until
2900 * all file descriptors have been examined.
2901 */
2902 for (; (uresid = uiop->uio_resid) > 0;
2903 uiop->uio_offset = uoffset + LXPR_SDSIZE) {
2904 int reclen;
2905 int fd;
2906 int len;
2907
2908 uoffset = uiop->uio_offset;
2909
2910 /*
2911 * Stop at the end of the fd list
2912 */
2913 fd = (uoffset / LXPR_SDSIZE) - 2;
2914 if (fd >= fddirsize) {
2915 if (eofp) {
2916 *eofp = 1;
2917 }
2918 goto out;
2919 }
2920
2921 if (fip->fi_list[fd].uf_file == NULL)
2922 continue;
2923
2924 dirent->d_ino = lxpr_inode(LXPR_PID_FD_FD, lxpnp->lxpr_pid, fd);
2925 len = snprintf(dirent->d_name, LXPNSIZ, "%d", fd);
2926 ASSERT(len < LXPNSIZ);
2927 reclen = DIRENT64_RECLEN(len);
2928
2929 dirent->d_off = (off64_t)(uoffset + LXPR_SDSIZE);
2930 dirent->d_reclen = (ushort_t)reclen;
2931
2932 if (reclen > uresid) {
2933 /*
2934 * Error if no entries have been returned yet.
2935 */
2936 if (uresid == oresid)
2937 error = EINVAL;
2938 goto out;
2939 }
2940
2941 if ((error = uiomove((caddr_t)dirent, reclen, UIO_READ,
2942 uiop)) != 0)
2943 goto out;
2944 }
2945
2946 if (eofp != NULL) {
2947 *eofp =
2948 (uiop->uio_offset >= ((fddirsize+2) * LXPR_SDSIZE)) ? 1 : 0;
2949 }
2950
2951 out:
2952 mutex_exit(&fip->fi_lock);
2953 return (error);
2954 }
2955
2956
2957 /*
2958 * lxpr_readlink(): Vnode operation for VOP_READLINK()
2959 */
2960 /* ARGSUSED */
2961 static int
2962 lxpr_readlink(vnode_t *vp, uio_t *uiop, cred_t *cr, caller_context_t *ct)
2963 {
2964 char bp[MAXPATHLEN + 1];
2965 size_t buflen = sizeof (bp);
2966 lxpr_node_t *lxpnp = VTOLXP(vp);
2967 vnode_t *rvp = lxpnp->lxpr_realvp;
2968 pid_t pid;
2969 int error = 0;
2970
2971 /* must be a symbolic link file */
2972 if (vp->v_type != VLNK)
2973 return (EINVAL);
2974
2975 /* Try to produce a symlink name for anything that has a realvp */
2976 if (rvp != NULL) {
2977 if ((error = lxpr_access(vp, VREAD, 0, CRED(), ct)) != 0)
2978 return (error);
2979 if ((error = vnodetopath(NULL, rvp, bp, buflen, CRED())) != 0)
2980 return (error);
2981 } else {
2982 switch (lxpnp->lxpr_type) {
2983 case LXPR_SELF:
2984 /*
2985 * Convert pid to the Linux default of 1 if we're the
2986 * zone's init process
2987 */
2988 pid = ((curproc->p_pid !=
2989 curproc->p_zone->zone_proc_initpid)
2990 ? curproc->p_pid : 1);
2991
2992 /*
2993 * Don't need to check result as every possible int
2994 * will fit within MAXPATHLEN bytes.
2995 */
2996 (void) snprintf(bp, buflen, "%d", pid);
2997 break;
2998 case LXPR_PID_CURDIR:
2999 case LXPR_PID_ROOTDIR:
3000 case LXPR_PID_EXE:
3001 return (EACCES);
3002 default:
3003 /*
3004 * Need to return error so that nothing thinks
3005 * that the symlink is empty and hence "."
3006 */
3007 return (EINVAL);
3008 }
3009 }
3010
3011 /* copy the link data to user space */
3012 return (uiomove(bp, strlen(bp), UIO_READ, uiop));
3013 }
3014
3015 /*
3016 * lxpr_inactive(): Vnode operation for VOP_INACTIVE()
3017 * Vnode is no longer referenced, deallocate the file
3018 * and all its resources.
3019 */
3020 /* ARGSUSED */
3021 static void
3022 lxpr_inactive(vnode_t *vp, cred_t *cr, caller_context_t *ct)
3023 {
3024 lxpr_freenode(VTOLXP(vp));
3025 }
3026
3027 /*
3028 * lxpr_sync(): Vnode operation for VOP_SYNC()
3029 */
3030 static int
3031 lxpr_sync()
3032 {
3033 /*
3034 * Nothing to sync but this function must never fail
3035 */
3036 return (0);
3037 }
3038
3039 /*
3040 * lxpr_cmp(): Vnode operation for VOP_CMP()
3041 */
3042 static int
3043 lxpr_cmp(vnode_t *vp1, vnode_t *vp2, caller_context_t *ct)
3044 {
3045 vnode_t *rvp;
3046
3047 while (vn_matchops(vp1, lxpr_vnodeops) &&
3048 (rvp = VTOLXP(vp1)->lxpr_realvp) != NULL) {
3049 vp1 = rvp;
3050 }
3051
3052 while (vn_matchops(vp2, lxpr_vnodeops) &&
3053 (rvp = VTOLXP(vp2)->lxpr_realvp) != NULL) {
3054 vp2 = rvp;
3055 }
3056
3057 if (vn_matchops(vp1, lxpr_vnodeops) || vn_matchops(vp2, lxpr_vnodeops))
3058 return (vp1 == vp2);
3059
3060 return (VOP_CMP(vp1, vp2, ct));
3061 }
3062
3063 /*
3064 * lxpr_realvp(): Vnode operation for VOP_REALVP()
3065 */
3066 static int
3067 lxpr_realvp(vnode_t *vp, vnode_t **vpp, caller_context_t *ct)
3068 {
3069 vnode_t *rvp;
3070
3071 if ((rvp = VTOLXP(vp)->lxpr_realvp) != NULL) {
3072 vp = rvp;
3073 if (VOP_REALVP(vp, &rvp, ct) == 0)
3074 vp = rvp;
3075 }
3076
3077 *vpp = vp;
3078 return (0);
3079 }