Commit | Line | Data |
---|---|---|
da7c5cc6 | 1 | /* |
436acd38 KB |
2 | * Copyright (c) 1982, 1986, 1989, 1991, 1993 |
3 | * The Regents of the University of California. All rights reserved. | |
adb35f79 KB |
4 | * (c) UNIX System Laboratories, Inc. |
5 | * All or some portions of this file are derived from material licensed | |
6 | * to the University of California by American Telephone and Telegraph | |
7 | * Co. or Unix System Laboratories, Inc. and are reproduced herein with | |
8 | * the permission of UNIX System Laboratories, Inc. | |
da7c5cc6 | 9 | * |
dbf0c423 | 10 | * %sccs.include.redist.c% |
c4ec2128 | 11 | * |
449a479b | 12 | * @(#)kern_fork.c 8.6 (Berkeley) %G% |
da7c5cc6 | 13 | */ |
50108d5c | 14 | |
31c4861a KM |
15 | #include <sys/param.h> |
16 | #include <sys/systm.h> | |
17 | #include <sys/map.h> | |
18 | #include <sys/filedesc.h> | |
19 | #include <sys/kernel.h> | |
20 | #include <sys/malloc.h> | |
21 | #include <sys/proc.h> | |
22 | #include <sys/resourcevar.h> | |
23 | #include <sys/vnode.h> | |
24 | #include <sys/file.h> | |
25 | #include <sys/acct.h> | |
26 | #include <sys/ktrace.h> | |
50108d5c | 27 | |
9e97623a CT |
28 | struct fork_args { |
29 | int dummy; | |
30 | }; | |
c9714ae3 KM |
31 | /* ARGSUSED */ |
32 | fork(p, uap, retval) | |
33 | struct proc *p; | |
9e97623a | 34 | struct fork_args *uap; |
c9714ae3 | 35 | int retval[]; |
50108d5c SL |
36 | { |
37 | ||
d9c2f47f | 38 | return (fork1(p, 0, retval)); |
50108d5c SL |
39 | } |
40 | ||
c9714ae3 KM |
41 | /* ARGSUSED */ |
42 | vfork(p, uap, retval) | |
43 | struct proc *p; | |
9e97623a | 44 | struct fork_args *uap; |
c9714ae3 | 45 | int retval[]; |
50108d5c SL |
46 | { |
47 | ||
d9c2f47f | 48 | return (fork1(p, 1, retval)); |
50108d5c SL |
49 | } |
50 | ||
ae2e0122 MK |
51 | int nprocs = 1; /* process 0 */ |
52 | ||
c9714ae3 KM |
53 | fork1(p1, isvfork, retval) |
54 | register struct proc *p1; | |
55 | int isvfork, retval[]; | |
50108d5c | 56 | { |
c9714ae3 | 57 | register struct proc *p2; |
9c71e14c | 58 | register uid_t uid; |
4bb226b7 | 59 | struct proc *newproc; |
141bed06 | 60 | struct proc **hash; |
9c71e14c | 61 | int count; |
ae2e0122 | 62 | static int nextpid, pidchecked = 0; |
50108d5c | 63 | |
50108d5c | 64 | /* |
141bed06 KM |
65 | * Although process entries are dynamically created, we still keep |
66 | * a global limit on the maximum number we will create. Don't allow | |
449a479b KM |
67 | * a nonprivileged user to use the last process; don't let root |
68 | * exceed the limit. The variable nprocs is the current number of | |
69 | * processes, maxproc is the limit. | |
50108d5c | 70 | */ |
9c71e14c | 71 | uid = p1->p_cred->p_ruid; |
449a479b | 72 | if ((nprocs >= maxproc - 1 && uid != 0) || nprocs >= maxproc) { |
50108d5c | 73 | tablefull("proc"); |
c9714ae3 | 74 | return (EAGAIN); |
50108d5c | 75 | } |
9c71e14c KM |
76 | /* |
77 | * Increment the count of procs running with this uid. Don't allow | |
78 | * a nonprivileged user to exceed their current limit. | |
79 | */ | |
80 | count = chgproccnt(uid, 1); | |
81 | if (uid != 0 && count > p1->p_rlimit[RLIMIT_NPROC].rlim_cur) { | |
82 | (void)chgproccnt(uid, -1); | |
ae2e0122 | 83 | return (EAGAIN); |
9c71e14c | 84 | } |
50108d5c | 85 | |
141bed06 | 86 | /* Allocate new proc. */ |
4bb226b7 | 87 | MALLOC(newproc, struct proc *, sizeof(struct proc), M_PROC, M_WAITOK); |
141bed06 | 88 | |
50108d5c | 89 | /* |
141bed06 KM |
90 | * Find an unused process ID. We remember a range of unused IDs |
91 | * ready to use (from nextpid+1 through pidchecked-1). | |
50108d5c | 92 | */ |
ae2e0122 | 93 | nextpid++; |
1d348849 | 94 | retry: |
ae2e0122 MK |
95 | /* |
96 | * If the process ID prototype has wrapped around, | |
97 | * restart somewhat above 0, as the low-numbered procs | |
98 | * tend to include daemons that don't exit. | |
99 | */ | |
100 | if (nextpid >= PID_MAX) { | |
101 | nextpid = 100; | |
1d348849 | 102 | pidchecked = 0; |
50108d5c | 103 | } |
ae2e0122 | 104 | if (nextpid >= pidchecked) { |
1d348849 | 105 | int doingzomb = 0; |
e9539440 | 106 | |
27ed98f2 | 107 | pidchecked = PID_MAX; |
1d348849 | 108 | /* |
ae2e0122 | 109 | * Scan the active and zombie procs to check whether this pid |
1d348849 | 110 | * is in use. Remember the lowest pid that's greater |
ae2e0122 | 111 | * than nextpid, so we can avoid checking for a while. |
1d348849 | 112 | */ |
80a8e5e6 | 113 | p2 = (struct proc *)allproc; |
1d348849 | 114 | again: |
cf5ef508 | 115 | for (; p2 != NULL; p2 = p2->p_next) { |
141bed06 | 116 | while (p2->p_pid == nextpid || |
ae2e0122 MK |
117 | p2->p_pgrp->pg_id == nextpid) { |
118 | nextpid++; | |
119 | if (nextpid >= pidchecked) | |
1d348849 MK |
120 | goto retry; |
121 | } | |
ae2e0122 MK |
122 | if (p2->p_pid > nextpid && pidchecked > p2->p_pid) |
123 | pidchecked = p2->p_pid; | |
124 | if (p2->p_pgrp->pg_id > nextpid && | |
125 | pidchecked > p2->p_pgrp->pg_id) | |
126 | pidchecked = p2->p_pgrp->pg_id; | |
1d348849 MK |
127 | } |
128 | if (!doingzomb) { | |
129 | doingzomb = 1; | |
ae2e0122 | 130 | p2 = zombproc; |
1d348849 MK |
131 | goto again; |
132 | } | |
50108d5c | 133 | } |
50108d5c | 134 | |
ae2e0122 | 135 | |
773b35ab KM |
136 | /* |
137 | * Link onto allproc (this should probably be delayed). | |
138 | * Heavy use of volatile here to prevent the compiler from | |
139 | * rearranging code. Yes, it *is* terribly ugly, but at least | |
140 | * it works. | |
141 | */ | |
ae2e0122 | 142 | nprocs++; |
4bb226b7 | 143 | p2 = newproc; |
773b35ab KM |
144 | #define Vp2 ((volatile struct proc *)p2) |
145 | Vp2->p_stat = SIDL; /* protect against others */ | |
146 | Vp2->p_pid = nextpid; | |
147 | /* | |
148 | * This is really: | |
cf5ef508 KB |
149 | * p2->p_next = allproc; |
150 | * allproc->p_prev = &p2->p_next; | |
773b35ab KM |
151 | * p2->p_prev = &allproc; |
152 | * allproc = p2; | |
153 | * The assignment via allproc is legal since it is never NULL. | |
154 | */ | |
cf5ef508 | 155 | *(volatile struct proc **)&Vp2->p_next = allproc; |
773b35ab | 156 | *(volatile struct proc ***)&allproc->p_prev = |
cf5ef508 | 157 | (volatile struct proc **)&Vp2->p_next; |
773b35ab KM |
158 | *(volatile struct proc ***)&Vp2->p_prev = &allproc; |
159 | allproc = Vp2; | |
160 | #undef Vp2 | |
cf5ef508 | 161 | p2->p_forw = p2->p_back = NULL; /* shouldn't be necessary */ |
1d348849 | 162 | |
141bed06 KM |
163 | /* Insert on the hash chain. */ |
164 | hash = &pidhash[PIDHASH(p2->p_pid)]; | |
165 | p2->p_hash = *hash; | |
166 | *hash = p2; | |
167 | ||
50108d5c SL |
168 | /* |
169 | * Make a proc table entry for the new process. | |
ae2e0122 MK |
170 | * Start by zeroing the section of proc that is zero-initialized, |
171 | * then copy the section that is copied directly from the parent. | |
50108d5c | 172 | */ |
ae2e0122 MK |
173 | bzero(&p2->p_startzero, |
174 | (unsigned) ((caddr_t)&p2->p_endzero - (caddr_t)&p2->p_startzero)); | |
175 | bcopy(&p1->p_startcopy, &p2->p_startcopy, | |
176 | (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy)); | |
177 | ||
178 | /* | |
179 | * Duplicate sub-structures as needed. | |
180 | * Increase reference counts on shared objects. | |
6795ddce | 181 | * The p_stats and p_sigacts substructs are set in vm_fork. |
ae2e0122 | 182 | */ |
cf5ef508 KB |
183 | p2->p_flag = P_INMEM; |
184 | if (p1->p_flag & P_PROFIL) | |
44d9bc96 | 185 | startprofclock(p2); |
ae2e0122 MK |
186 | MALLOC(p2->p_cred, struct pcred *, sizeof(struct pcred), |
187 | M_SUBPROC, M_WAITOK); | |
188 | bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred)); | |
53e7449e | 189 | p2->p_cred->p_refcnt = 1; |
ae2e0122 MK |
190 | crhold(p1->p_ucred); |
191 | ||
859437ef JSP |
192 | /* bump references to the text vnode (for procfs) */ |
193 | p2->p_textvp = p1->p_textvp; | |
194 | if (p2->p_textvp) | |
195 | VREF(p2->p_textvp); | |
196 | ||
ae2e0122 | 197 | p2->p_fd = fdcopy(p1); |
ae2e0122 MK |
198 | /* |
199 | * If p_limit is still copy-on-write, bump refcnt, | |
200 | * otherwise get a copy that won't be modified. | |
201 | * (If PL_SHAREMOD is clear, the structure is shared | |
202 | * copy-on-write.) | |
203 | */ | |
204 | if (p1->p_limit->p_lflags & PL_SHAREMOD) | |
205 | p2->p_limit = limcopy(p1->p_limit); | |
206 | else { | |
207 | p2->p_limit = p1->p_limit; | |
208 | p2->p_limit->p_refcnt++; | |
98111078 | 209 | } |
ae2e0122 | 210 | |
cf5ef508 KB |
211 | if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT) |
212 | p2->p_flag |= P_CONTROLT; | |
ae2e0122 | 213 | if (isvfork) |
cf5ef508 | 214 | p2->p_flag |= P_PPWAIT; |
ae2e0122 MK |
215 | p2->p_pgrpnxt = p1->p_pgrpnxt; |
216 | p1->p_pgrpnxt = p2; | |
217 | p2->p_pptr = p1; | |
218 | p2->p_osptr = p1->p_cptr; | |
219 | if (p1->p_cptr) | |
220 | p1->p_cptr->p_ysptr = p2; | |
221 | p1->p_cptr = p2; | |
222 | #ifdef KTRACE | |
50108d5c | 223 | /* |
ae2e0122 MK |
224 | * Copy traceflag and tracefile if enabled. |
225 | * If not inherited, these were zeroed above. | |
50108d5c | 226 | */ |
ae2e0122 MK |
227 | if (p1->p_traceflag&KTRFAC_INHERIT) { |
228 | p2->p_traceflag = p1->p_traceflag; | |
229 | if ((p2->p_tracep = p1->p_tracep) != NULL) | |
230 | VREF(p2->p_tracep); | |
231 | } | |
232 | #endif | |
233 | ||
50108d5c | 234 | /* |
50108d5c SL |
235 | * This begins the section where we must prevent the parent |
236 | * from being swapped. | |
237 | */ | |
cf5ef508 | 238 | p1->p_flag |= P_NOSWAP; |
6795ddce MK |
239 | /* |
240 | * Set return values for child before vm_fork, | |
241 | * so they can be copied to child stack. | |
242 | * We return parent pid, and mark as child in retval[1]. | |
f6fca708 MK |
243 | * NOTE: the kernel stack may be at a different location in the child |
244 | * process, and thus addresses of automatic variables (including retval) | |
245 | * may be invalid after vm_fork returns in the child process. | |
6795ddce MK |
246 | */ |
247 | retval[0] = p1->p_pid; | |
248 | retval[1] = 1; | |
ae2e0122 MK |
249 | if (vm_fork(p1, p2, isvfork)) { |
250 | /* | |
6795ddce | 251 | * Child process. Set start time and get to work. |
ae2e0122 | 252 | */ |
d848aba6 | 253 | (void) splclock(); |
ae2e0122 | 254 | p2->p_stats->p_start = time; |
d848aba6 | 255 | (void) spl0(); |
ae2e0122 MK |
256 | p2->p_acflag = AFORK; |
257 | return (0); | |
d848aba6 | 258 | } |
50108d5c SL |
259 | |
260 | /* | |
261 | * Make child runnable and add to run queue. | |
262 | */ | |
ae2e0122 MK |
263 | (void) splhigh(); |
264 | p2->p_stat = SRUN; | |
cb84e0ab | 265 | setrunqueue(p2); |
50108d5c SL |
266 | (void) spl0(); |
267 | ||
50108d5c SL |
268 | /* |
269 | * Now can be swapped. | |
270 | */ | |
cf5ef508 | 271 | p1->p_flag &= ~P_NOSWAP; |
50108d5c SL |
272 | |
273 | /* | |
cf5ef508 KB |
274 | * Preserve synchronization semantics of vfork. If waiting for |
275 | * child to exec or exit, set P_PPWAIT on child, and sleep on our | |
276 | * proc (in case of exit). | |
50108d5c | 277 | */ |
ae2e0122 | 278 | if (isvfork) |
cf5ef508 KB |
279 | while (p2->p_flag & P_PPWAIT) |
280 | tsleep(p1, PWAIT, "ppwait", 0); | |
50108d5c SL |
281 | |
282 | /* | |
6795ddce MK |
283 | * Return child pid to parent process, |
284 | * marking us as parent via retval[1]. | |
50108d5c | 285 | */ |
ae2e0122 | 286 | retval[0] = p2->p_pid; |
6795ddce | 287 | retval[1] = 0; |
50108d5c SL |
288 | return (0); |
289 | } |