Home | History | Annotate | Line # | Download | only in kern
kern_proc.c revision 1.112
      1  1.112        ad /*	$NetBSD: kern_proc.c,v 1.112 2007/06/15 20:17:08 ad Exp $	*/
      2   1.33   thorpej 
      3   1.33   thorpej /*-
      4  1.100        ad  * Copyright (c) 1999, 2006, 2007 The NetBSD Foundation, Inc.
      5   1.33   thorpej  * All rights reserved.
      6   1.33   thorpej  *
      7   1.33   thorpej  * This code is derived from software contributed to The NetBSD Foundation
      8   1.33   thorpej  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
      9  1.100        ad  * NASA Ames Research Center, and by Andrew Doran.
     10   1.33   thorpej  *
     11   1.33   thorpej  * Redistribution and use in source and binary forms, with or without
     12   1.33   thorpej  * modification, are permitted provided that the following conditions
     13   1.33   thorpej  * are met:
     14   1.33   thorpej  * 1. Redistributions of source code must retain the above copyright
     15   1.33   thorpej  *    notice, this list of conditions and the following disclaimer.
     16   1.33   thorpej  * 2. Redistributions in binary form must reproduce the above copyright
     17   1.33   thorpej  *    notice, this list of conditions and the following disclaimer in the
     18   1.33   thorpej  *    documentation and/or other materials provided with the distribution.
     19   1.33   thorpej  * 3. All advertising materials mentioning features or use of this software
     20   1.33   thorpej  *    must display the following acknowledgement:
     21   1.33   thorpej  *	This product includes software developed by the NetBSD
     22   1.33   thorpej  *	Foundation, Inc. and its contributors.
     23   1.33   thorpej  * 4. Neither the name of The NetBSD Foundation nor the names of its
     24   1.33   thorpej  *    contributors may be used to endorse or promote products derived
     25   1.33   thorpej  *    from this software without specific prior written permission.
     26   1.33   thorpej  *
     27   1.33   thorpej  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
     28   1.33   thorpej  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
     29   1.33   thorpej  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
     30   1.33   thorpej  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
     31   1.33   thorpej  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
     32   1.33   thorpej  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
     33   1.33   thorpej  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
     34   1.33   thorpej  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
     35   1.33   thorpej  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
     36   1.33   thorpej  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
     37   1.33   thorpej  * POSSIBILITY OF SUCH DAMAGE.
     38   1.33   thorpej  */
     39    1.9       cgd 
     40    1.1       cgd /*
     41    1.7       cgd  * Copyright (c) 1982, 1986, 1989, 1991, 1993
     42    1.7       cgd  *	The Regents of the University of California.  All rights reserved.
     43    1.1       cgd  *
     44    1.1       cgd  * Redistribution and use in source and binary forms, with or without
     45    1.1       cgd  * modification, are permitted provided that the following conditions
     46    1.1       cgd  * are met:
     47    1.1       cgd  * 1. Redistributions of source code must retain the above copyright
     48    1.1       cgd  *    notice, this list of conditions and the following disclaimer.
     49    1.1       cgd  * 2. Redistributions in binary form must reproduce the above copyright
     50    1.1       cgd  *    notice, this list of conditions and the following disclaimer in the
     51    1.1       cgd  *    documentation and/or other materials provided with the distribution.
     52   1.65       agc  * 3. Neither the name of the University nor the names of its contributors
     53    1.1       cgd  *    may be used to endorse or promote products derived from this software
     54    1.1       cgd  *    without specific prior written permission.
     55    1.1       cgd  *
     56    1.1       cgd  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
     57    1.1       cgd  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
     58    1.1       cgd  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
     59    1.1       cgd  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
     60    1.1       cgd  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
     61    1.1       cgd  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
     62    1.1       cgd  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
     63    1.1       cgd  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
     64    1.1       cgd  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
     65    1.1       cgd  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
     66    1.1       cgd  * SUCH DAMAGE.
     67    1.1       cgd  *
     68   1.23      fvdl  *	@(#)kern_proc.c	8.7 (Berkeley) 2/14/95
     69    1.1       cgd  */
     70   1.45     lukem 
     71   1.45     lukem #include <sys/cdefs.h>
     72  1.112        ad __KERNEL_RCSID(0, "$NetBSD: kern_proc.c,v 1.112 2007/06/15 20:17:08 ad Exp $");
     73   1.48      yamt 
     74   1.48      yamt #include "opt_kstack.h"
     75   1.88      onoe #include "opt_maxuprc.h"
     76   1.90       rjs #include "opt_multiprocessor.h"
     77   1.90       rjs #include "opt_lockdebug.h"
     78    1.1       cgd 
     79    1.5   mycroft #include <sys/param.h>
     80    1.5   mycroft #include <sys/systm.h>
     81    1.5   mycroft #include <sys/kernel.h>
     82    1.5   mycroft #include <sys/proc.h>
     83   1.28   thorpej #include <sys/resourcevar.h>
     84    1.5   mycroft #include <sys/buf.h>
     85    1.5   mycroft #include <sys/acct.h>
     86    1.5   mycroft #include <sys/wait.h>
     87    1.5   mycroft #include <sys/file.h>
     88    1.8   mycroft #include <ufs/ufs/quota.h>
     89    1.5   mycroft #include <sys/uio.h>
     90    1.5   mycroft #include <sys/malloc.h>
     91   1.24   thorpej #include <sys/pool.h>
     92    1.5   mycroft #include <sys/mbuf.h>
     93    1.5   mycroft #include <sys/ioctl.h>
     94    1.5   mycroft #include <sys/tty.h>
     95   1.11       cgd #include <sys/signalvar.h>
     96   1.51  gmcgarry #include <sys/ras.h>
     97   1.81  junyoung #include <sys/filedesc.h>
     98  1.103       dsl #include "sys/syscall_stats.h"
     99   1.89      elad #include <sys/kauth.h>
    100  1.100        ad #include <sys/sleepq.h>
    101   1.81  junyoung 
    102   1.81  junyoung #include <uvm/uvm.h>
    103   1.79      yamt #include <uvm/uvm_extern.h>
    104    1.5   mycroft 
    105    1.7       cgd /*
    106   1.10   mycroft  * Other process lists
    107    1.7       cgd  */
    108   1.31   thorpej 
    109   1.10   mycroft struct proclist allproc;
    110   1.32   thorpej struct proclist zombproc;	/* resources have been freed */
    111   1.32   thorpej 
    112   1.32   thorpej /*
    113  1.100        ad  * There are two locks on global process state.
    114  1.100        ad  *
    115  1.107        ad  * 1. proclist_lock is an adaptive mutex and is used when modifying
    116  1.107        ad  * or examining process state from a process context.  It protects
    117  1.107        ad  * the internal tables, all of the process lists, and a number of
    118  1.107        ad  * members of struct proc.
    119  1.107        ad  *
    120  1.100        ad  * 2. proclist_mutex is used when allproc must be traversed from an
    121  1.107        ad  * interrupt context, or when changing the state of processes.  The
    122  1.107        ad  * proclist_lock should always be used in preference.  In some cases,
    123  1.107        ad  * both locks need to be held.
    124   1.33   thorpej  *
    125  1.100        ad  *	proclist_lock	proclist_mutex	structure
    126  1.100        ad  *	--------------- --------------- -----------------
    127  1.100        ad  *	x				zombproc
    128  1.100        ad  *	x		x		pid_table
    129  1.100        ad  *	x				proc::p_pptr
    130  1.100        ad  *	x				proc::p_sibling
    131  1.100        ad  *	x				proc::p_children
    132  1.100        ad  *	x		x		allproc
    133  1.100        ad  *	x		x		proc::p_pgrp
    134  1.100        ad  *	x		x		proc::p_pglist
    135  1.100        ad  *	x		x		proc::p_session
    136  1.100        ad  *	x		x		proc::p_list
    137  1.100        ad  *			x		alllwp
    138  1.100        ad  *			x		lwp::l_list
    139   1.33   thorpej  *
    140  1.100        ad  * The lock order for processes and LWPs is approximately as following:
    141   1.33   thorpej  *
    142  1.107        ad  * kernel_lock
    143  1.100        ad  * -> proclist_lock
    144  1.107        ad  *   -> proc::p_mutex
    145  1.107        ad  *      -> proclist_mutex
    146  1.100        ad  *         -> proc::p_smutex
    147  1.107        ad  *           -> proc::p_stmutex
    148  1.107        ad  *
    149  1.107        ad  * XXX p_smutex can be run at IPL_VM once audio drivers on the x86
    150  1.107        ad  * platform are made MP safe.  Currently it blocks interrupts at
    151  1.107        ad  * IPL_SCHED and below.
    152  1.107        ad  *
    153  1.107        ad  * XXX The two process locks (p_smutex + p_mutex), and the two global
    154  1.107        ad  * state locks (proclist_lock + proclist_mutex) should be merged
    155  1.107        ad  * together.  However, to do so requires interrupts that interrupts
    156  1.107        ad  * be run with LWP context.
    157   1.33   thorpej  */
    158  1.107        ad kmutex_t	proclist_lock;
    159  1.100        ad kmutex_t	proclist_mutex;
    160   1.33   thorpej 
    161   1.33   thorpej /*
    162   1.72  junyoung  * pid to proc lookup is done by indexing the pid_table array.
    163   1.61       dsl  * Since pid numbers are only allocated when an empty slot
    164   1.61       dsl  * has been found, there is no need to search any lists ever.
    165   1.61       dsl  * (an orphaned pgrp will lock the slot, a session will lock
    166   1.61       dsl  * the pgrp with the same number.)
    167   1.61       dsl  * If the table is too small it is reallocated with twice the
    168   1.61       dsl  * previous size and the entries 'unzipped' into the two halves.
    169   1.61       dsl  * A linked list of free entries is passed through the pt_proc
    170   1.61       dsl  * field of 'free' items - set odd to be an invalid ptr.
    171   1.61       dsl  */
    172   1.61       dsl 
    173   1.61       dsl struct pid_table {
    174   1.61       dsl 	struct proc	*pt_proc;
    175   1.61       dsl 	struct pgrp	*pt_pgrp;
    176   1.72  junyoung };
    177   1.61       dsl #if 1	/* strongly typed cast - should be a noop */
    178   1.84     perry static inline uint p2u(struct proc *p) { return (uint)(uintptr_t)p; }
    179   1.61       dsl #else
    180   1.61       dsl #define p2u(p) ((uint)p)
    181   1.72  junyoung #endif
    182   1.61       dsl #define P_VALID(p) (!(p2u(p) & 1))
    183   1.61       dsl #define P_NEXT(p) (p2u(p) >> 1)
    184   1.61       dsl #define P_FREE(pid) ((struct proc *)(uintptr_t)((pid) << 1 | 1))
    185   1.61       dsl 
    186   1.61       dsl #define INITIAL_PID_TABLE_SIZE	(1 << 5)
    187   1.61       dsl static struct pid_table *pid_table;
    188   1.61       dsl static uint pid_tbl_mask = INITIAL_PID_TABLE_SIZE - 1;
    189   1.61       dsl static uint pid_alloc_lim;	/* max we allocate before growing table */
    190   1.61       dsl static uint pid_alloc_cnt;	/* number of allocated pids */
    191   1.61       dsl 
    192   1.61       dsl /* links through free slots - never empty! */
    193   1.61       dsl static uint next_free_pt, last_free_pt;
    194   1.61       dsl static pid_t pid_max = PID_MAX;		/* largest value we allocate */
    195   1.31   thorpej 
    196   1.81  junyoung /* Components of the first process -- never freed. */
    197   1.81  junyoung struct session session0;
    198   1.81  junyoung struct pgrp pgrp0;
    199   1.81  junyoung struct proc proc0;
    200  1.100        ad struct lwp lwp0 __aligned(MIN_LWP_ALIGNMENT);
    201   1.89      elad kauth_cred_t cred0;
    202   1.81  junyoung struct filedesc0 filedesc0;
    203   1.81  junyoung struct cwdinfo cwdi0;
    204   1.81  junyoung struct plimit limit0;
    205   1.81  junyoung struct pstats pstat0;
    206   1.81  junyoung struct vmspace vmspace0;
    207   1.81  junyoung struct sigacts sigacts0;
    208  1.100        ad struct turnstile turnstile0;
    209   1.81  junyoung 
    210   1.81  junyoung extern struct user *proc0paddr;
    211   1.81  junyoung 
    212   1.81  junyoung extern const struct emul emul_netbsd;	/* defined in kern_exec.c */
    213   1.81  junyoung 
    214   1.81  junyoung int nofile = NOFILE;
    215   1.81  junyoung int maxuprc = MAXUPRC;
    216   1.81  junyoung int cmask = CMASK;
    217   1.81  junyoung 
    218   1.77    simonb POOL_INIT(proc_pool, sizeof(struct proc), 0, 0, 0, "procpl",
    219  1.108        ad     &pool_allocator_nointr, IPL_NONE);
    220   1.77    simonb POOL_INIT(pgrp_pool, sizeof(struct pgrp), 0, 0, 0, "pgrppl",
    221  1.108        ad     &pool_allocator_nointr, IPL_NONE);
    222   1.77    simonb POOL_INIT(plimit_pool, sizeof(struct plimit), 0, 0, 0, "plimitpl",
    223  1.108        ad     &pool_allocator_nointr, IPL_NONE);
    224   1.77    simonb POOL_INIT(pstats_pool, sizeof(struct pstats), 0, 0, 0, "pstatspl",
    225  1.108        ad     &pool_allocator_nointr, IPL_NONE);
    226   1.77    simonb POOL_INIT(session_pool, sizeof(struct session), 0, 0, 0, "sessionpl",
    227  1.108        ad     &pool_allocator_nointr, IPL_NONE);
    228   1.57   thorpej 
    229   1.57   thorpej MALLOC_DEFINE(M_EMULDATA, "emuldata", "Per-process emulation data");
    230   1.57   thorpej MALLOC_DEFINE(M_PROC, "proc", "Proc structures");
    231   1.57   thorpej MALLOC_DEFINE(M_SUBPROC, "subproc", "Proc sub-structures");
    232   1.10   mycroft 
    233   1.31   thorpej /*
    234   1.31   thorpej  * The process list descriptors, used during pid allocation and
    235   1.31   thorpej  * by sysctl.  No locking on this data structure is needed since
    236   1.31   thorpej  * it is completely static.
    237   1.31   thorpej  */
    238   1.31   thorpej const struct proclist_desc proclists[] = {
    239   1.31   thorpej 	{ &allproc	},
    240   1.31   thorpej 	{ &zombproc	},
    241   1.31   thorpej 	{ NULL		},
    242   1.31   thorpej };
    243   1.31   thorpej 
    244   1.72  junyoung static void orphanpg(struct pgrp *);
    245   1.72  junyoung static void pg_delete(pid_t);
    246   1.13  christos 
    247   1.95   thorpej static specificdata_domain_t proc_specificdata_domain;
    248   1.95   thorpej 
    249   1.10   mycroft /*
    250   1.10   mycroft  * Initialize global process hashing structures.
    251   1.10   mycroft  */
    252   1.11       cgd void
    253   1.59       dsl procinit(void)
    254    1.7       cgd {
    255   1.31   thorpej 	const struct proclist_desc *pd;
    256   1.61       dsl 	int i;
    257   1.61       dsl #define	LINK_EMPTY ((PID_MAX + INITIAL_PID_TABLE_SIZE) & ~(INITIAL_PID_TABLE_SIZE - 1))
    258   1.31   thorpej 
    259   1.31   thorpej 	for (pd = proclists; pd->pd_list != NULL; pd++)
    260   1.31   thorpej 		LIST_INIT(pd->pd_list);
    261    1.7       cgd 
    262  1.107        ad 	mutex_init(&proclist_lock, MUTEX_DEFAULT, IPL_NONE);
    263  1.100        ad 	mutex_init(&proclist_mutex, MUTEX_SPIN, IPL_SCHED);
    264   1.33   thorpej 
    265   1.61       dsl 	pid_table = malloc(INITIAL_PID_TABLE_SIZE * sizeof *pid_table,
    266   1.61       dsl 			    M_PROC, M_WAITOK);
    267   1.61       dsl 	/* Set free list running through table...
    268   1.61       dsl 	   Preset 'use count' above PID_MAX so we allocate pid 1 next. */
    269   1.61       dsl 	for (i = 0; i <= pid_tbl_mask; i++) {
    270   1.61       dsl 		pid_table[i].pt_proc = P_FREE(LINK_EMPTY + i + 1);
    271   1.61       dsl 		pid_table[i].pt_pgrp = 0;
    272   1.61       dsl 	}
    273   1.61       dsl 	/* slot 0 is just grabbed */
    274   1.61       dsl 	next_free_pt = 1;
    275   1.61       dsl 	/* Need to fix last entry. */
    276   1.61       dsl 	last_free_pt = pid_tbl_mask;
    277   1.61       dsl 	pid_table[last_free_pt].pt_proc = P_FREE(LINK_EMPTY);
    278   1.61       dsl 	/* point at which we grow table - to avoid reusing pids too often */
    279   1.61       dsl 	pid_alloc_lim = pid_tbl_mask - 1;
    280   1.61       dsl #undef LINK_EMPTY
    281   1.61       dsl 
    282   1.55   thorpej 	LIST_INIT(&alllwp);
    283   1.55   thorpej 
    284   1.43        ad 	uihashtbl =
    285   1.43        ad 	    hashinit(maxproc / 16, HASH_LIST, M_PROC, M_WAITOK, &uihash);
    286   1.95   thorpej 
    287   1.95   thorpej 	proc_specificdata_domain = specificdata_domain_create();
    288   1.95   thorpej 	KASSERT(proc_specificdata_domain != NULL);
    289    1.7       cgd }
    290    1.1       cgd 
    291    1.7       cgd /*
    292   1.81  junyoung  * Initialize process 0.
    293   1.81  junyoung  */
    294   1.81  junyoung void
    295   1.81  junyoung proc0_init(void)
    296   1.81  junyoung {
    297   1.81  junyoung 	struct proc *p;
    298   1.81  junyoung 	struct pgrp *pg;
    299   1.81  junyoung 	struct session *sess;
    300   1.81  junyoung 	struct lwp *l;
    301   1.81  junyoung 	u_int i;
    302   1.81  junyoung 	rlim_t lim;
    303   1.81  junyoung 
    304   1.81  junyoung 	p = &proc0;
    305   1.81  junyoung 	pg = &pgrp0;
    306   1.81  junyoung 	sess = &session0;
    307   1.81  junyoung 	l = &lwp0;
    308   1.81  junyoung 
    309  1.107        ad 	/*
    310  1.107        ad 	 * XXX p_rasmutex is run at IPL_SCHED, because of lock order
    311  1.107        ad 	 * issues (kernel_lock -> p_rasmutex).  Ideally ras_lookup
    312  1.107        ad 	 * should operate "lock free".
    313  1.107        ad 	 */
    314  1.100        ad 	mutex_init(&p->p_smutex, MUTEX_SPIN, IPL_SCHED);
    315  1.112        ad 	mutex_init(&p->p_stmutex, MUTEX_SPIN, IPL_HIGH);
    316  1.107        ad 	mutex_init(&p->p_rasmutex, MUTEX_SPIN, IPL_SCHED);
    317  1.100        ad 	mutex_init(&p->p_mutex, MUTEX_DEFAULT, IPL_NONE);
    318  1.107        ad 
    319  1.100        ad 	cv_init(&p->p_refcv, "drainref");
    320  1.100        ad 	cv_init(&p->p_waitcv, "wait");
    321  1.100        ad 	cv_init(&p->p_lwpcv, "lwpwait");
    322  1.100        ad 
    323   1.81  junyoung 	LIST_INIT(&p->p_lwps);
    324  1.100        ad 	LIST_INIT(&p->p_sigwaiters);
    325   1.81  junyoung 	LIST_INSERT_HEAD(&p->p_lwps, l, l_sibling);
    326  1.100        ad 
    327   1.81  junyoung 	p->p_nlwps = 1;
    328  1.100        ad 	p->p_nrlwps = 1;
    329  1.110      yamt 	p->p_nlwpid = l->l_lid;
    330  1.100        ad 	p->p_refcnt = 1;
    331   1.81  junyoung 
    332   1.81  junyoung 	pid_table[0].pt_proc = p;
    333   1.81  junyoung 	LIST_INSERT_HEAD(&allproc, p, p_list);
    334   1.81  junyoung 	LIST_INSERT_HEAD(&alllwp, l, l_list);
    335   1.81  junyoung 
    336   1.81  junyoung 	p->p_pgrp = pg;
    337   1.81  junyoung 	pid_table[0].pt_pgrp = pg;
    338   1.81  junyoung 	LIST_INIT(&pg->pg_members);
    339   1.81  junyoung 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
    340   1.81  junyoung 
    341   1.81  junyoung 	pg->pg_session = sess;
    342   1.81  junyoung 	sess->s_count = 1;
    343   1.81  junyoung 	sess->s_sid = 0;
    344   1.81  junyoung 	sess->s_leader = p;
    345   1.81  junyoung 
    346   1.81  junyoung 	/*
    347   1.81  junyoung 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
    348   1.81  junyoung 	 * init(8) when they exit.  init(8) can easily wait them out
    349   1.81  junyoung 	 * for us.
    350   1.81  junyoung 	 */
    351  1.102     pavel 	p->p_flag = PK_SYSTEM | PK_NOCLDWAIT;
    352   1.81  junyoung 	p->p_stat = SACTIVE;
    353   1.81  junyoung 	p->p_nice = NZERO;
    354   1.81  junyoung 	p->p_emul = &emul_netbsd;
    355   1.81  junyoung #ifdef __HAVE_SYSCALL_INTERN
    356   1.81  junyoung 	(*p->p_emul->e_syscall_intern)(p);
    357   1.81  junyoung #endif
    358  1.110      yamt 	strlcpy(p->p_comm, "system", sizeof(p->p_comm));
    359   1.81  junyoung 
    360  1.102     pavel 	l->l_flag = LW_INMEM | LW_SYSTEM;
    361   1.81  junyoung 	l->l_stat = LSONPROC;
    362  1.100        ad 	l->l_ts = &turnstile0;
    363  1.100        ad 	l->l_syncobj = &sched_syncobj;
    364  1.100        ad 	l->l_refcnt = 1;
    365  1.100        ad 	l->l_cpu = curcpu();
    366  1.100        ad 	l->l_priority = PRIBIO;
    367  1.100        ad 	l->l_usrpri = PRIBIO;
    368  1.105      yamt 	l->l_inheritedprio = MAXPRI;
    369  1.105      yamt 	SLIST_INIT(&l->l_pi_lenders);
    370  1.110      yamt 	l->l_name = "swapper";
    371   1.81  junyoung 
    372   1.81  junyoung 	callout_init(&l->l_tsleep_ch);
    373  1.100        ad 	cv_init(&l->l_sigcv, "sigwait");
    374   1.81  junyoung 
    375   1.81  junyoung 	/* Create credentials. */
    376   1.89      elad 	cred0 = kauth_cred_alloc();
    377   1.89      elad 	p->p_cred = cred0;
    378  1.100        ad 	kauth_cred_hold(cred0);
    379  1.100        ad 	l->l_cred = cred0;
    380   1.81  junyoung 
    381   1.81  junyoung 	/* Create the CWD info. */
    382   1.81  junyoung 	p->p_cwdi = &cwdi0;
    383   1.81  junyoung 	cwdi0.cwdi_cmask = cmask;
    384   1.81  junyoung 	cwdi0.cwdi_refcnt = 1;
    385   1.81  junyoung 	simple_lock_init(&cwdi0.cwdi_slock);
    386   1.81  junyoung 
    387   1.81  junyoung 	/* Create the limits structures. */
    388   1.81  junyoung 	p->p_limit = &limit0;
    389   1.81  junyoung 	simple_lock_init(&limit0.p_slock);
    390   1.81  junyoung 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    391   1.81  junyoung 		limit0.pl_rlimit[i].rlim_cur =
    392   1.81  junyoung 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    393   1.81  junyoung 
    394   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
    395   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
    396   1.81  junyoung 	    maxfiles < nofile ? maxfiles : nofile;
    397   1.81  junyoung 
    398   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
    399   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
    400   1.81  junyoung 	    maxproc < maxuprc ? maxproc : maxuprc;
    401   1.81  junyoung 
    402   1.81  junyoung 	lim = ptoa(uvmexp.free);
    403   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
    404   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
    405   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
    406   1.81  junyoung 	limit0.pl_corename = defcorename;
    407   1.81  junyoung 	limit0.p_refcnt = 1;
    408   1.81  junyoung 
    409   1.81  junyoung 	/* Configure virtual memory system, set vm rlimits. */
    410   1.81  junyoung 	uvm_init_limits(p);
    411   1.81  junyoung 
    412   1.81  junyoung 	/* Initialize file descriptor table for proc0. */
    413   1.81  junyoung 	p->p_fd = &filedesc0.fd_fd;
    414   1.81  junyoung 	fdinit1(&filedesc0);
    415   1.81  junyoung 
    416   1.81  junyoung 	/*
    417   1.81  junyoung 	 * Initialize proc0's vmspace, which uses the kernel pmap.
    418   1.81  junyoung 	 * All kernel processes (which never have user space mappings)
    419   1.81  junyoung 	 * share proc0's vmspace, and thus, the kernel pmap.
    420   1.81  junyoung 	 */
    421   1.81  junyoung 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    422   1.81  junyoung 	    trunc_page(VM_MAX_ADDRESS));
    423   1.81  junyoung 	p->p_vmspace = &vmspace0;
    424   1.81  junyoung 
    425   1.81  junyoung 	l->l_addr = proc0paddr;				/* XXX */
    426   1.81  junyoung 
    427   1.81  junyoung 	p->p_stats = &pstat0;
    428   1.81  junyoung 
    429   1.81  junyoung 	/* Initialize signal state for proc0. */
    430   1.81  junyoung 	p->p_sigacts = &sigacts0;
    431  1.100        ad 	mutex_init(&p->p_sigacts->sa_mutex, MUTEX_SPIN, IPL_NONE);
    432   1.81  junyoung 	siginit(p);
    433   1.96  christos 
    434   1.96  christos 	proc_initspecific(p);
    435   1.96  christos 	lwp_initspecific(l);
    436  1.103       dsl 
    437  1.103       dsl 	SYSCALL_TIME_LWP_INIT(l);
    438   1.81  junyoung }
    439   1.81  junyoung 
    440   1.81  junyoung /*
    441   1.74  junyoung  * Check that the specified process group is in the session of the
    442   1.60       dsl  * specified process.
    443   1.60       dsl  * Treats -ve ids as process ids.
    444   1.60       dsl  * Used to validate TIOCSPGRP requests.
    445   1.60       dsl  */
    446   1.60       dsl int
    447   1.60       dsl pgid_in_session(struct proc *p, pid_t pg_id)
    448   1.60       dsl {
    449   1.60       dsl 	struct pgrp *pgrp;
    450  1.101       dsl 	struct session *session;
    451  1.107        ad 	int error;
    452  1.101       dsl 
    453  1.107        ad 	mutex_enter(&proclist_lock);
    454   1.60       dsl 	if (pg_id < 0) {
    455  1.101       dsl 		struct proc *p1 = p_find(-pg_id, PFIND_LOCKED | PFIND_UNLOCK_FAIL);
    456   1.64       dsl 		if (p1 == NULL)
    457   1.64       dsl 			return EINVAL;
    458   1.60       dsl 		pgrp = p1->p_pgrp;
    459   1.60       dsl 	} else {
    460  1.101       dsl 		pgrp = pg_find(pg_id, PFIND_LOCKED | PFIND_UNLOCK_FAIL);
    461   1.60       dsl 		if (pgrp == NULL)
    462   1.64       dsl 			return EINVAL;
    463   1.60       dsl 	}
    464  1.101       dsl 	session = pgrp->pg_session;
    465  1.101       dsl 	if (session != p->p_pgrp->pg_session)
    466  1.107        ad 		error = EPERM;
    467  1.107        ad 	else
    468  1.107        ad 		error = 0;
    469  1.107        ad 	mutex_exit(&proclist_lock);
    470  1.107        ad 
    471  1.107        ad 	return error;
    472    1.7       cgd }
    473    1.4    andrew 
    474    1.1       cgd /*
    475   1.41  sommerfe  * Is p an inferior of q?
    476   1.94        ad  *
    477   1.94        ad  * Call with the proclist_lock held.
    478    1.1       cgd  */
    479   1.11       cgd int
    480   1.59       dsl inferior(struct proc *p, struct proc *q)
    481    1.1       cgd {
    482    1.1       cgd 
    483   1.41  sommerfe 	for (; p != q; p = p->p_pptr)
    484    1.1       cgd 		if (p->p_pid == 0)
    485   1.82  junyoung 			return 0;
    486   1.82  junyoung 	return 1;
    487    1.1       cgd }
    488    1.1       cgd 
    489    1.1       cgd /*
    490    1.1       cgd  * Locate a process by number
    491    1.1       cgd  */
    492    1.1       cgd struct proc *
    493   1.68       dsl p_find(pid_t pid, uint flags)
    494    1.1       cgd {
    495   1.33   thorpej 	struct proc *p;
    496   1.68       dsl 	char stat;
    497    1.1       cgd 
    498   1.68       dsl 	if (!(flags & PFIND_LOCKED))
    499  1.107        ad 		mutex_enter(&proclist_lock);
    500  1.100        ad 
    501   1.61       dsl 	p = pid_table[pid & pid_tbl_mask].pt_proc;
    502  1.100        ad 
    503   1.61       dsl 	/* Only allow live processes to be found by pid. */
    504  1.100        ad 	/* XXXSMP p_stat */
    505  1.100        ad 	if (P_VALID(p) && p->p_pid == pid && ((stat = p->p_stat) == SACTIVE ||
    506  1.100        ad 	    stat == SSTOP || ((flags & PFIND_ZOMBIE) &&
    507  1.100        ad 	    (stat == SZOMB || stat == SDEAD || stat == SDYING)))) {
    508   1.68       dsl 		if (flags & PFIND_UNLOCK_OK)
    509  1.107        ad 			 mutex_exit(&proclist_lock);
    510   1.68       dsl 		return p;
    511   1.68       dsl 	}
    512   1.68       dsl 	if (flags & PFIND_UNLOCK_FAIL)
    513  1.107        ad 		mutex_exit(&proclist_lock);
    514   1.68       dsl 	return NULL;
    515    1.1       cgd }
    516    1.1       cgd 
    517   1.61       dsl 
    518    1.1       cgd /*
    519    1.1       cgd  * Locate a process group by number
    520    1.1       cgd  */
    521    1.1       cgd struct pgrp *
    522   1.68       dsl pg_find(pid_t pgid, uint flags)
    523    1.1       cgd {
    524   1.68       dsl 	struct pgrp *pg;
    525    1.1       cgd 
    526   1.68       dsl 	if (!(flags & PFIND_LOCKED))
    527  1.107        ad 		mutex_enter(&proclist_lock);
    528   1.68       dsl 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
    529   1.61       dsl 	/*
    530   1.61       dsl 	 * Can't look up a pgrp that only exists because the session
    531   1.61       dsl 	 * hasn't died yet (traditional)
    532   1.61       dsl 	 */
    533   1.68       dsl 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
    534   1.68       dsl 		if (flags & PFIND_UNLOCK_FAIL)
    535  1.107        ad 			 mutex_exit(&proclist_lock);
    536   1.68       dsl 		return NULL;
    537   1.68       dsl 	}
    538   1.68       dsl 
    539   1.68       dsl 	if (flags & PFIND_UNLOCK_OK)
    540  1.107        ad 		mutex_exit(&proclist_lock);
    541   1.68       dsl 	return pg;
    542    1.1       cgd }
    543    1.1       cgd 
    544   1.61       dsl static void
    545   1.61       dsl expand_pid_table(void)
    546    1.1       cgd {
    547   1.61       dsl 	uint pt_size = pid_tbl_mask + 1;
    548   1.61       dsl 	struct pid_table *n_pt, *new_pt;
    549   1.61       dsl 	struct proc *proc;
    550   1.61       dsl 	struct pgrp *pgrp;
    551   1.61       dsl 	int i;
    552   1.61       dsl 	pid_t pid;
    553    1.1       cgd 
    554   1.61       dsl 	new_pt = malloc(pt_size * 2 * sizeof *new_pt, M_PROC, M_WAITOK);
    555   1.61       dsl 
    556  1.107        ad 	mutex_enter(&proclist_lock);
    557   1.61       dsl 	if (pt_size != pid_tbl_mask + 1) {
    558   1.61       dsl 		/* Another process beat us to it... */
    559  1.107        ad 		mutex_exit(&proclist_lock);
    560   1.61       dsl 		FREE(new_pt, M_PROC);
    561   1.61       dsl 		return;
    562   1.61       dsl 	}
    563   1.72  junyoung 
    564   1.61       dsl 	/*
    565   1.61       dsl 	 * Copy entries from old table into new one.
    566   1.61       dsl 	 * If 'pid' is 'odd' we need to place in the upper half,
    567   1.61       dsl 	 * even pid's to the lower half.
    568   1.61       dsl 	 * Free items stay in the low half so we don't have to
    569   1.61       dsl 	 * fixup the reference to them.
    570   1.61       dsl 	 * We stuff free items on the front of the freelist
    571   1.61       dsl 	 * because we can't write to unmodified entries.
    572   1.74  junyoung 	 * Processing the table backwards maintains a semblance
    573   1.61       dsl 	 * of issueing pid numbers that increase with time.
    574   1.61       dsl 	 */
    575   1.61       dsl 	i = pt_size - 1;
    576   1.61       dsl 	n_pt = new_pt + i;
    577   1.61       dsl 	for (; ; i--, n_pt--) {
    578   1.61       dsl 		proc = pid_table[i].pt_proc;
    579   1.61       dsl 		pgrp = pid_table[i].pt_pgrp;
    580   1.61       dsl 		if (!P_VALID(proc)) {
    581   1.61       dsl 			/* Up 'use count' so that link is valid */
    582   1.61       dsl 			pid = (P_NEXT(proc) + pt_size) & ~pt_size;
    583   1.61       dsl 			proc = P_FREE(pid);
    584   1.61       dsl 			if (pgrp)
    585   1.61       dsl 				pid = pgrp->pg_id;
    586   1.61       dsl 		} else
    587   1.61       dsl 			pid = proc->p_pid;
    588   1.72  junyoung 
    589   1.61       dsl 		/* Save entry in appropriate half of table */
    590   1.61       dsl 		n_pt[pid & pt_size].pt_proc = proc;
    591   1.61       dsl 		n_pt[pid & pt_size].pt_pgrp = pgrp;
    592   1.61       dsl 
    593   1.61       dsl 		/* Put other piece on start of free list */
    594   1.61       dsl 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
    595   1.61       dsl 		n_pt[pid & pt_size].pt_proc =
    596   1.61       dsl 				    P_FREE((pid & ~pt_size) | next_free_pt);
    597   1.61       dsl 		n_pt[pid & pt_size].pt_pgrp = 0;
    598   1.61       dsl 		next_free_pt = i | (pid & pt_size);
    599   1.61       dsl 		if (i == 0)
    600   1.61       dsl 			break;
    601   1.61       dsl 	}
    602   1.61       dsl 
    603   1.61       dsl 	/* Switch tables */
    604  1.100        ad 	mutex_enter(&proclist_mutex);
    605   1.61       dsl 	n_pt = pid_table;
    606   1.61       dsl 	pid_table = new_pt;
    607  1.100        ad 	mutex_exit(&proclist_mutex);
    608   1.61       dsl 	pid_tbl_mask = pt_size * 2 - 1;
    609   1.61       dsl 
    610   1.61       dsl 	/*
    611   1.61       dsl 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
    612   1.61       dsl 	 * allocated pids we need it to be larger!
    613   1.61       dsl 	 */
    614   1.61       dsl 	if (pid_tbl_mask > PID_MAX) {
    615   1.61       dsl 		pid_max = pid_tbl_mask * 2 + 1;
    616   1.61       dsl 		pid_alloc_lim |= pid_alloc_lim << 1;
    617   1.61       dsl 	} else
    618   1.61       dsl 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
    619   1.61       dsl 
    620  1.107        ad 	mutex_exit(&proclist_lock);
    621   1.61       dsl 	FREE(n_pt, M_PROC);
    622   1.61       dsl }
    623   1.61       dsl 
    624   1.61       dsl struct proc *
    625   1.61       dsl proc_alloc(void)
    626   1.61       dsl {
    627   1.61       dsl 	struct proc *p;
    628  1.100        ad 	int nxt;
    629   1.61       dsl 	pid_t pid;
    630   1.61       dsl 	struct pid_table *pt;
    631   1.61       dsl 
    632   1.61       dsl 	p = pool_get(&proc_pool, PR_WAITOK);
    633   1.61       dsl 	p->p_stat = SIDL;			/* protect against others */
    634   1.61       dsl 
    635   1.96  christos 	proc_initspecific(p);
    636   1.61       dsl 	/* allocate next free pid */
    637   1.61       dsl 
    638   1.61       dsl 	for (;;expand_pid_table()) {
    639   1.61       dsl 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
    640   1.61       dsl 			/* ensure pids cycle through 2000+ values */
    641   1.61       dsl 			continue;
    642  1.107        ad 		mutex_enter(&proclist_lock);
    643   1.61       dsl 		pt = &pid_table[next_free_pt];
    644    1.1       cgd #ifdef DIAGNOSTIC
    645   1.63  christos 		if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
    646   1.61       dsl 			panic("proc_alloc: slot busy");
    647    1.1       cgd #endif
    648   1.61       dsl 		nxt = P_NEXT(pt->pt_proc);
    649   1.61       dsl 		if (nxt & pid_tbl_mask)
    650   1.61       dsl 			break;
    651   1.61       dsl 		/* Table full - expand (NB last entry not used....) */
    652  1.107        ad 		mutex_exit(&proclist_lock);
    653   1.61       dsl 	}
    654   1.61       dsl 
    655   1.61       dsl 	/* pid is 'saved use count' + 'size' + entry */
    656   1.61       dsl 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
    657   1.61       dsl 	if ((uint)pid > (uint)pid_max)
    658   1.61       dsl 		pid &= pid_tbl_mask;
    659   1.61       dsl 	p->p_pid = pid;
    660   1.61       dsl 	next_free_pt = nxt & pid_tbl_mask;
    661   1.61       dsl 
    662   1.61       dsl 	/* Grab table slot */
    663  1.100        ad 	mutex_enter(&proclist_mutex);
    664   1.61       dsl 	pt->pt_proc = p;
    665  1.100        ad 	mutex_exit(&proclist_mutex);
    666   1.61       dsl 	pid_alloc_cnt++;
    667   1.61       dsl 
    668  1.107        ad 	mutex_exit(&proclist_lock);
    669   1.61       dsl 
    670   1.61       dsl 	return p;
    671   1.61       dsl }
    672   1.61       dsl 
    673   1.61       dsl /*
    674   1.61       dsl  * Free last resources of a process - called from proc_free (in kern_exit.c)
    675  1.100        ad  *
    676  1.107        ad  * Called with the proclist_lock held, and releases upon exit.
    677   1.61       dsl  */
    678   1.61       dsl void
    679   1.61       dsl proc_free_mem(struct proc *p)
    680   1.61       dsl {
    681   1.61       dsl 	pid_t pid = p->p_pid;
    682   1.61       dsl 	struct pid_table *pt;
    683   1.61       dsl 
    684  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
    685   1.61       dsl 
    686   1.61       dsl 	pt = &pid_table[pid & pid_tbl_mask];
    687    1.1       cgd #ifdef DIAGNOSTIC
    688   1.63  christos 	if (__predict_false(pt->pt_proc != p))
    689   1.61       dsl 		panic("proc_free: pid_table mismatch, pid %x, proc %p",
    690   1.61       dsl 			pid, p);
    691    1.1       cgd #endif
    692  1.100        ad 	mutex_enter(&proclist_mutex);
    693   1.61       dsl 	/* save pid use count in slot */
    694   1.61       dsl 	pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
    695   1.61       dsl 
    696   1.61       dsl 	if (pt->pt_pgrp == NULL) {
    697   1.61       dsl 		/* link last freed entry onto ours */
    698   1.61       dsl 		pid &= pid_tbl_mask;
    699   1.61       dsl 		pt = &pid_table[last_free_pt];
    700   1.61       dsl 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
    701   1.61       dsl 		last_free_pt = pid;
    702   1.61       dsl 		pid_alloc_cnt--;
    703   1.61       dsl 	}
    704  1.100        ad 	mutex_exit(&proclist_mutex);
    705   1.61       dsl 
    706   1.61       dsl 	nprocs--;
    707  1.107        ad 	mutex_exit(&proclist_lock);
    708   1.61       dsl 
    709   1.61       dsl 	pool_put(&proc_pool, p);
    710   1.61       dsl }
    711   1.61       dsl 
    712   1.61       dsl /*
    713   1.61       dsl  * Move p to a new or existing process group (and session)
    714   1.61       dsl  *
    715   1.61       dsl  * If we are creating a new pgrp, the pgid should equal
    716   1.72  junyoung  * the calling process' pid.
    717   1.61       dsl  * If is only valid to enter a process group that is in the session
    718   1.61       dsl  * of the process.
    719   1.61       dsl  * Also mksess should only be set if we are creating a process group
    720   1.61       dsl  *
    721   1.72  junyoung  * Only called from sys_setsid, sys_setpgid/sys_setpgrp and the
    722  1.100        ad  * SYSV setpgrp support for hpux.
    723   1.61       dsl  */
    724   1.61       dsl int
    725  1.100        ad enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, int mksess)
    726   1.61       dsl {
    727   1.61       dsl 	struct pgrp *new_pgrp, *pgrp;
    728   1.61       dsl 	struct session *sess;
    729  1.100        ad 	struct proc *p;
    730   1.61       dsl 	int rval;
    731   1.61       dsl 	pid_t pg_id = NO_PGID;
    732   1.61       dsl 
    733   1.61       dsl 	if (mksess)
    734   1.99     pooka 		sess = pool_get(&session_pool, PR_WAITOK);
    735   1.61       dsl 	else
    736   1.61       dsl 		sess = NULL;
    737   1.61       dsl 
    738  1.107        ad 	/* Allocate data areas we might need before doing any validity checks */
    739  1.107        ad 	mutex_enter(&proclist_lock);		/* Because pid_table might change */
    740  1.107        ad 	if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
    741  1.107        ad 		mutex_exit(&proclist_lock);
    742  1.107        ad 		new_pgrp = pool_get(&pgrp_pool, PR_WAITOK);
    743  1.107        ad 		mutex_enter(&proclist_lock);
    744  1.107        ad 	} else
    745  1.107        ad 		new_pgrp = NULL;
    746   1.61       dsl 	rval = EPERM;	/* most common error (to save typing) */
    747   1.61       dsl 
    748   1.61       dsl 	/* Check pgrp exists or can be created */
    749   1.61       dsl 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
    750   1.61       dsl 	if (pgrp != NULL && pgrp->pg_id != pgid)
    751   1.61       dsl 		goto done;
    752   1.61       dsl 
    753   1.61       dsl 	/* Can only set another process under restricted circumstances. */
    754  1.100        ad 	if (pid != curp->p_pid) {
    755   1.61       dsl 		/* must exist and be one of our children... */
    756  1.100        ad 		if ((p = p_find(pid, PFIND_LOCKED)) == NULL ||
    757  1.100        ad 		    !inferior(p, curp)) {
    758   1.61       dsl 			rval = ESRCH;
    759   1.61       dsl 			goto done;
    760   1.61       dsl 		}
    761   1.61       dsl 		/* ... in the same session... */
    762   1.61       dsl 		if (sess != NULL || p->p_session != curp->p_session)
    763   1.61       dsl 			goto done;
    764   1.61       dsl 		/* ... existing pgid must be in same session ... */
    765   1.61       dsl 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
    766   1.61       dsl 			goto done;
    767   1.61       dsl 		/* ... and not done an exec. */
    768  1.102     pavel 		if (p->p_flag & PK_EXEC) {
    769   1.61       dsl 			rval = EACCES;
    770   1.61       dsl 			goto done;
    771   1.49     enami 		}
    772  1.100        ad 	} else {
    773  1.100        ad 		/* ... setsid() cannot re-enter a pgrp */
    774  1.100        ad 		if (mksess && (curp->p_pgid == curp->p_pid ||
    775  1.100        ad 		    pg_find(curp->p_pid, PFIND_LOCKED)))
    776  1.100        ad 			goto done;
    777  1.100        ad 		p = curp;
    778   1.61       dsl 	}
    779    1.1       cgd 
    780   1.61       dsl 	/* Changing the process group/session of a session
    781   1.61       dsl 	   leader is definitely off limits. */
    782   1.61       dsl 	if (SESS_LEADER(p)) {
    783   1.61       dsl 		if (sess == NULL && p->p_pgrp == pgrp)
    784   1.61       dsl 			/* unless it's a definite noop */
    785   1.61       dsl 			rval = 0;
    786   1.61       dsl 		goto done;
    787   1.61       dsl 	}
    788   1.61       dsl 
    789   1.61       dsl 	/* Can only create a process group with id of process */
    790   1.61       dsl 	if (pgrp == NULL && pgid != pid)
    791   1.61       dsl 		goto done;
    792   1.61       dsl 
    793   1.61       dsl 	/* Can only create a session if creating pgrp */
    794   1.61       dsl 	if (sess != NULL && pgrp != NULL)
    795   1.61       dsl 		goto done;
    796   1.61       dsl 
    797   1.61       dsl 	/* Check we allocated memory for a pgrp... */
    798   1.61       dsl 	if (pgrp == NULL && new_pgrp == NULL)
    799   1.61       dsl 		goto done;
    800   1.61       dsl 
    801   1.61       dsl 	/* Don't attach to 'zombie' pgrp */
    802   1.61       dsl 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
    803   1.61       dsl 		goto done;
    804   1.61       dsl 
    805   1.61       dsl 	/* Expect to succeed now */
    806   1.61       dsl 	rval = 0;
    807   1.61       dsl 
    808   1.61       dsl 	if (pgrp == p->p_pgrp)
    809   1.61       dsl 		/* nothing to do */
    810   1.61       dsl 		goto done;
    811   1.61       dsl 
    812   1.61       dsl 	/* Ok all setup, link up required structures */
    813  1.100        ad 
    814   1.61       dsl 	if (pgrp == NULL) {
    815   1.61       dsl 		pgrp = new_pgrp;
    816   1.61       dsl 		new_pgrp = 0;
    817   1.61       dsl 		if (sess != NULL) {
    818   1.21   thorpej 			sess->s_sid = p->p_pid;
    819    1.1       cgd 			sess->s_leader = p;
    820    1.1       cgd 			sess->s_count = 1;
    821    1.1       cgd 			sess->s_ttyvp = NULL;
    822    1.1       cgd 			sess->s_ttyp = NULL;
    823   1.58       dsl 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
    824   1.25     perry 			memcpy(sess->s_login, p->p_session->s_login,
    825    1.1       cgd 			    sizeof(sess->s_login));
    826  1.100        ad 			p->p_lflag &= ~PL_CONTROLT;
    827    1.1       cgd 		} else {
    828   1.61       dsl 			sess = p->p_pgrp->pg_session;
    829   1.61       dsl 			SESSHOLD(sess);
    830    1.1       cgd 		}
    831   1.61       dsl 		pgrp->pg_session = sess;
    832   1.61       dsl 		sess = 0;
    833   1.61       dsl 
    834    1.1       cgd 		pgrp->pg_id = pgid;
    835   1.10   mycroft 		LIST_INIT(&pgrp->pg_members);
    836   1.61       dsl #ifdef DIAGNOSTIC
    837   1.63  christos 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
    838   1.61       dsl 			panic("enterpgrp: pgrp table slot in use");
    839   1.63  christos 		if (__predict_false(mksess && p != curp))
    840   1.63  christos 			panic("enterpgrp: mksession and p != curproc");
    841   1.61       dsl #endif
    842  1.100        ad 		mutex_enter(&proclist_mutex);
    843   1.61       dsl 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
    844    1.1       cgd 		pgrp->pg_jobc = 0;
    845  1.100        ad 	} else
    846  1.100        ad 		mutex_enter(&proclist_mutex);
    847  1.100        ad 
    848  1.100        ad #ifdef notyet
    849  1.100        ad 	/*
    850  1.100        ad 	 * If there's a controlling terminal for the current session, we
    851  1.100        ad 	 * have to interlock with it.  See ttread().
    852  1.100        ad 	 */
    853  1.100        ad 	if (p->p_session->s_ttyvp != NULL) {
    854  1.100        ad 		tp = p->p_session->s_ttyp;
    855  1.100        ad 		mutex_enter(&tp->t_mutex);
    856  1.100        ad 	} else
    857  1.100        ad 		tp = NULL;
    858  1.100        ad #endif
    859    1.1       cgd 
    860    1.1       cgd 	/*
    861    1.1       cgd 	 * Adjust eligibility of affected pgrps to participate in job control.
    862    1.1       cgd 	 * Increment eligibility counts before decrementing, otherwise we
    863    1.1       cgd 	 * could reach 0 spuriously during the first call.
    864    1.1       cgd 	 */
    865    1.1       cgd 	fixjobc(p, pgrp, 1);
    866    1.1       cgd 	fixjobc(p, p->p_pgrp, 0);
    867    1.1       cgd 
    868  1.100        ad 	/* Move process to requested group. */
    869   1.10   mycroft 	LIST_REMOVE(p, p_pglist);
    870   1.52      matt 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
    871   1.61       dsl 		/* defer delete until we've dumped the lock */
    872   1.61       dsl 		pg_id = p->p_pgrp->pg_id;
    873    1.1       cgd 	p->p_pgrp = pgrp;
    874   1.10   mycroft 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
    875  1.100        ad 	mutex_exit(&proclist_mutex);
    876  1.100        ad 
    877  1.100        ad #ifdef notyet
    878  1.100        ad 	/* Done with the swap; we can release the tty mutex. */
    879  1.100        ad 	if (tp != NULL)
    880  1.100        ad 		mutex_exit(&tp->t_mutex);
    881  1.100        ad #endif
    882   1.61       dsl 
    883   1.61       dsl     done:
    884  1.100        ad 	if (pg_id != NO_PGID)
    885  1.100        ad 		pg_delete(pg_id);
    886  1.107        ad 	mutex_exit(&proclist_lock);
    887   1.61       dsl 	if (sess != NULL)
    888   1.77    simonb 		pool_put(&session_pool, sess);
    889   1.61       dsl 	if (new_pgrp != NULL)
    890   1.61       dsl 		pool_put(&pgrp_pool, new_pgrp);
    891   1.63  christos #ifdef DEBUG_PGRP
    892   1.63  christos 	if (__predict_false(rval))
    893   1.61       dsl 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
    894   1.61       dsl 			pid, pgid, mksess, curp->p_pid, rval);
    895   1.61       dsl #endif
    896   1.61       dsl 	return rval;
    897    1.1       cgd }
    898    1.1       cgd 
    899    1.1       cgd /*
    900  1.100        ad  * Remove a process from its process group.  Must be called with the
    901  1.107        ad  * proclist_lock held.
    902    1.1       cgd  */
    903  1.100        ad void
    904   1.59       dsl leavepgrp(struct proc *p)
    905    1.1       cgd {
    906   1.61       dsl 	struct pgrp *pgrp;
    907    1.1       cgd 
    908  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
    909  1.100        ad 
    910  1.100        ad 	/*
    911  1.100        ad 	 * If there's a controlling terminal for the session, we have to
    912  1.100        ad 	 * interlock with it.  See ttread().
    913  1.100        ad 	 */
    914  1.100        ad 	mutex_enter(&proclist_mutex);
    915  1.100        ad #ifdef notyet
    916  1.100        ad 	if (p_>p_session->s_ttyvp != NULL) {
    917  1.100        ad 		tp = p->p_session->s_ttyp;
    918  1.100        ad 		mutex_enter(&tp->t_mutex);
    919  1.100        ad 	} else
    920  1.100        ad 		tp = NULL;
    921  1.100        ad #endif
    922  1.100        ad 
    923   1.61       dsl 	pgrp = p->p_pgrp;
    924   1.10   mycroft 	LIST_REMOVE(p, p_pglist);
    925   1.94        ad 	p->p_pgrp = NULL;
    926   1.61       dsl 
    927  1.100        ad #ifdef notyet
    928  1.100        ad 	if (tp != NULL)
    929  1.100        ad 		mutex_exit(&tp->t_mutex);
    930  1.100        ad #endif
    931  1.100        ad 	mutex_exit(&proclist_mutex);
    932  1.100        ad 
    933  1.100        ad 	if (LIST_EMPTY(&pgrp->pg_members))
    934  1.100        ad 		pg_delete(pgrp->pg_id);
    935   1.61       dsl }
    936   1.61       dsl 
    937  1.100        ad /*
    938  1.107        ad  * Free a process group.  Must be called with the proclist_lock held.
    939  1.100        ad  */
    940   1.61       dsl static void
    941   1.61       dsl pg_free(pid_t pg_id)
    942   1.61       dsl {
    943   1.61       dsl 	struct pgrp *pgrp;
    944   1.61       dsl 	struct pid_table *pt;
    945   1.61       dsl 
    946  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
    947  1.100        ad 
    948   1.61       dsl 	pt = &pid_table[pg_id & pid_tbl_mask];
    949   1.61       dsl 	pgrp = pt->pt_pgrp;
    950   1.61       dsl #ifdef DIAGNOSTIC
    951   1.63  christos 	if (__predict_false(!pgrp || pgrp->pg_id != pg_id
    952   1.63  christos 	    || !LIST_EMPTY(&pgrp->pg_members)))
    953   1.61       dsl 		panic("pg_free: process group absent or has members");
    954   1.61       dsl #endif
    955   1.61       dsl 	pt->pt_pgrp = 0;
    956   1.61       dsl 
    957   1.61       dsl 	if (!P_VALID(pt->pt_proc)) {
    958   1.61       dsl 		/* orphaned pgrp, put slot onto free list */
    959   1.61       dsl #ifdef DIAGNOSTIC
    960   1.63  christos 		if (__predict_false(P_NEXT(pt->pt_proc) & pid_tbl_mask))
    961   1.61       dsl 			panic("pg_free: process slot on free list");
    962   1.61       dsl #endif
    963  1.100        ad 		mutex_enter(&proclist_mutex);
    964   1.61       dsl 		pg_id &= pid_tbl_mask;
    965   1.61       dsl 		pt = &pid_table[last_free_pt];
    966   1.61       dsl 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
    967  1.100        ad 		mutex_exit(&proclist_mutex);
    968   1.61       dsl 		last_free_pt = pg_id;
    969   1.61       dsl 		pid_alloc_cnt--;
    970   1.61       dsl 	}
    971   1.61       dsl 	pool_put(&pgrp_pool, pgrp);
    972    1.1       cgd }
    973    1.1       cgd 
    974    1.1       cgd /*
    975  1.107        ad  * Delete a process group.  Must be called with the proclist_lock held.
    976    1.1       cgd  */
    977   1.61       dsl static void
    978   1.61       dsl pg_delete(pid_t pg_id)
    979   1.61       dsl {
    980   1.61       dsl 	struct pgrp *pgrp;
    981   1.61       dsl 	struct tty *ttyp;
    982   1.61       dsl 	struct session *ss;
    983  1.100        ad 	int is_pgrp_leader;
    984  1.100        ad 
    985  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
    986   1.61       dsl 
    987   1.61       dsl 	pgrp = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
    988   1.61       dsl 	if (pgrp == NULL || pgrp->pg_id != pg_id ||
    989  1.100        ad 	    !LIST_EMPTY(&pgrp->pg_members))
    990   1.61       dsl 		return;
    991   1.61       dsl 
    992   1.71        pk 	ss = pgrp->pg_session;
    993   1.71        pk 
    994   1.61       dsl 	/* Remove reference (if any) from tty to this process group */
    995   1.71        pk 	ttyp = ss->s_ttyp;
    996   1.71        pk 	if (ttyp != NULL && ttyp->t_pgrp == pgrp) {
    997   1.61       dsl 		ttyp->t_pgrp = NULL;
    998   1.71        pk #ifdef DIAGNOSTIC
    999   1.71        pk 		if (ttyp->t_session != ss)
   1000   1.71        pk 			panic("pg_delete: wrong session on terminal");
   1001   1.71        pk #endif
   1002   1.71        pk 	}
   1003   1.61       dsl 
   1004   1.71        pk 	/*
   1005   1.71        pk 	 * The leading process group in a session is freed
   1006   1.71        pk 	 * by sessdelete() if last reference.
   1007   1.71        pk 	 */
   1008   1.71        pk 	is_pgrp_leader = (ss->s_sid == pgrp->pg_id);
   1009   1.71        pk 	SESSRELE(ss);
   1010   1.61       dsl 
   1011   1.71        pk 	if (is_pgrp_leader)
   1012   1.61       dsl 		return;
   1013   1.61       dsl 
   1014   1.61       dsl 	pg_free(pg_id);
   1015   1.61       dsl }
   1016   1.61       dsl 
   1017   1.61       dsl /*
   1018   1.61       dsl  * Delete session - called from SESSRELE when s_count becomes zero.
   1019  1.107        ad  * Must be called with the proclist_lock held.
   1020   1.61       dsl  */
   1021   1.11       cgd void
   1022   1.61       dsl sessdelete(struct session *ss)
   1023    1.1       cgd {
   1024  1.100        ad 
   1025  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
   1026  1.100        ad 
   1027   1.61       dsl 	/*
   1028   1.61       dsl 	 * We keep the pgrp with the same id as the session in
   1029   1.61       dsl 	 * order to stop a process being given the same pid.
   1030   1.61       dsl 	 * Since the pgrp holds a reference to the session, it
   1031   1.61       dsl 	 * must be a 'zombie' pgrp by now.
   1032   1.61       dsl 	 */
   1033   1.61       dsl 	pg_free(ss->s_sid);
   1034   1.77    simonb 	pool_put(&session_pool, ss);
   1035    1.1       cgd }
   1036    1.1       cgd 
   1037    1.1       cgd /*
   1038    1.1       cgd  * Adjust pgrp jobc counters when specified process changes process group.
   1039    1.1       cgd  * We count the number of processes in each process group that "qualify"
   1040    1.1       cgd  * the group for terminal job control (those with a parent in a different
   1041    1.1       cgd  * process group of the same session).  If that count reaches zero, the
   1042    1.1       cgd  * process group becomes orphaned.  Check both the specified process'
   1043    1.1       cgd  * process group and that of its children.
   1044    1.1       cgd  * entering == 0 => p is leaving specified group.
   1045    1.1       cgd  * entering == 1 => p is entering specified group.
   1046   1.68       dsl  *
   1047  1.107        ad  * Call with proclist_lock held.
   1048    1.1       cgd  */
   1049    1.4    andrew void
   1050   1.59       dsl fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
   1051    1.1       cgd {
   1052   1.39  augustss 	struct pgrp *hispgrp;
   1053   1.39  augustss 	struct session *mysession = pgrp->pg_session;
   1054   1.68       dsl 	struct proc *child;
   1055    1.1       cgd 
   1056  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
   1057  1.107        ad 	KASSERT(mutex_owned(&proclist_mutex));
   1058  1.100        ad 
   1059    1.1       cgd 	/*
   1060    1.1       cgd 	 * Check p's parent to see whether p qualifies its own process
   1061    1.1       cgd 	 * group; if so, adjust count for p's process group.
   1062    1.1       cgd 	 */
   1063   1.68       dsl 	hispgrp = p->p_pptr->p_pgrp;
   1064   1.68       dsl 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
   1065  1.100        ad 		if (entering) {
   1066  1.100        ad 			mutex_enter(&p->p_smutex);
   1067  1.100        ad 			p->p_sflag &= ~PS_ORPHANPG;
   1068  1.100        ad 			mutex_exit(&p->p_smutex);
   1069    1.1       cgd 			pgrp->pg_jobc++;
   1070  1.100        ad 		} else if (--pgrp->pg_jobc == 0)
   1071    1.1       cgd 			orphanpg(pgrp);
   1072   1.26   thorpej 	}
   1073    1.1       cgd 
   1074    1.1       cgd 	/*
   1075    1.1       cgd 	 * Check this process' children to see whether they qualify
   1076    1.1       cgd 	 * their process groups; if so, adjust counts for children's
   1077    1.1       cgd 	 * process groups.
   1078    1.1       cgd 	 */
   1079   1.68       dsl 	LIST_FOREACH(child, &p->p_children, p_sibling) {
   1080   1.68       dsl 		hispgrp = child->p_pgrp;
   1081   1.68       dsl 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
   1082   1.68       dsl 		    !P_ZOMBIE(child)) {
   1083  1.100        ad 			if (entering) {
   1084  1.100        ad 				mutex_enter(&child->p_smutex);
   1085  1.100        ad 				child->p_sflag &= ~PS_ORPHANPG;
   1086  1.100        ad 				mutex_exit(&child->p_smutex);
   1087    1.1       cgd 				hispgrp->pg_jobc++;
   1088  1.100        ad 			} else if (--hispgrp->pg_jobc == 0)
   1089    1.1       cgd 				orphanpg(hispgrp);
   1090   1.26   thorpej 		}
   1091   1.26   thorpej 	}
   1092    1.1       cgd }
   1093    1.1       cgd 
   1094   1.72  junyoung /*
   1095    1.1       cgd  * A process group has become orphaned;
   1096    1.1       cgd  * if there are any stopped processes in the group,
   1097    1.1       cgd  * hang-up all process in that group.
   1098   1.68       dsl  *
   1099  1.107        ad  * Call with proclist_lock held.
   1100    1.1       cgd  */
   1101    1.4    andrew static void
   1102   1.59       dsl orphanpg(struct pgrp *pg)
   1103    1.1       cgd {
   1104   1.39  augustss 	struct proc *p;
   1105  1.100        ad 	int doit;
   1106  1.100        ad 
   1107  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
   1108  1.107        ad 	KASSERT(mutex_owned(&proclist_mutex));
   1109  1.100        ad 
   1110  1.100        ad 	doit = 0;
   1111    1.1       cgd 
   1112   1.52      matt 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
   1113  1.100        ad 		mutex_enter(&p->p_smutex);
   1114    1.1       cgd 		if (p->p_stat == SSTOP) {
   1115  1.100        ad 			doit = 1;
   1116  1.100        ad 			p->p_sflag |= PS_ORPHANPG;
   1117    1.1       cgd 		}
   1118  1.100        ad 		mutex_exit(&p->p_smutex);
   1119    1.1       cgd 	}
   1120   1.35    bouyer 
   1121  1.100        ad 	if (doit) {
   1122  1.100        ad 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
   1123  1.100        ad 			psignal(p, SIGHUP);
   1124  1.100        ad 			psignal(p, SIGCONT);
   1125   1.35    bouyer 		}
   1126   1.35    bouyer 	}
   1127   1.35    bouyer }
   1128    1.1       cgd 
   1129   1.61       dsl #ifdef DDB
   1130   1.61       dsl #include <ddb/db_output.h>
   1131   1.61       dsl void pidtbl_dump(void);
   1132   1.14  christos void
   1133   1.61       dsl pidtbl_dump(void)
   1134    1.1       cgd {
   1135   1.61       dsl 	struct pid_table *pt;
   1136   1.61       dsl 	struct proc *p;
   1137   1.39  augustss 	struct pgrp *pgrp;
   1138   1.61       dsl 	int id;
   1139    1.1       cgd 
   1140   1.61       dsl 	db_printf("pid table %p size %x, next %x, last %x\n",
   1141   1.61       dsl 		pid_table, pid_tbl_mask+1,
   1142   1.61       dsl 		next_free_pt, last_free_pt);
   1143   1.61       dsl 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
   1144   1.61       dsl 		p = pt->pt_proc;
   1145   1.61       dsl 		if (!P_VALID(p) && !pt->pt_pgrp)
   1146   1.61       dsl 			continue;
   1147   1.61       dsl 		db_printf("  id %x: ", id);
   1148   1.61       dsl 		if (P_VALID(p))
   1149   1.61       dsl 			db_printf("proc %p id %d (0x%x) %s\n",
   1150   1.61       dsl 				p, p->p_pid, p->p_pid, p->p_comm);
   1151   1.61       dsl 		else
   1152   1.61       dsl 			db_printf("next %x use %x\n",
   1153   1.61       dsl 				P_NEXT(p) & pid_tbl_mask,
   1154   1.61       dsl 				P_NEXT(p) & ~pid_tbl_mask);
   1155   1.61       dsl 		if ((pgrp = pt->pt_pgrp)) {
   1156   1.61       dsl 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
   1157   1.61       dsl 			    pgrp->pg_session, pgrp->pg_session->s_sid,
   1158   1.61       dsl 			    pgrp->pg_session->s_count,
   1159   1.61       dsl 			    pgrp->pg_session->s_login);
   1160   1.61       dsl 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
   1161   1.61       dsl 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
   1162   1.61       dsl 			    pgrp->pg_members.lh_first);
   1163   1.61       dsl 			for (p = pgrp->pg_members.lh_first; p != 0;
   1164   1.61       dsl 			    p = p->p_pglist.le_next) {
   1165   1.72  junyoung 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
   1166   1.61       dsl 				    p->p_pid, p, p->p_pgrp, p->p_comm);
   1167   1.10   mycroft 			}
   1168    1.1       cgd 		}
   1169    1.1       cgd 	}
   1170    1.1       cgd }
   1171   1.61       dsl #endif /* DDB */
   1172   1.48      yamt 
   1173   1.48      yamt #ifdef KSTACK_CHECK_MAGIC
   1174   1.48      yamt #include <sys/user.h>
   1175   1.48      yamt 
   1176   1.48      yamt #define	KSTACK_MAGIC	0xdeadbeaf
   1177   1.48      yamt 
   1178   1.48      yamt /* XXX should be per process basis? */
   1179   1.48      yamt int kstackleftmin = KSTACK_SIZE;
   1180   1.50     enami int kstackleftthres = KSTACK_SIZE / 8; /* warn if remaining stack is
   1181   1.50     enami 					  less than this */
   1182   1.48      yamt 
   1183   1.48      yamt void
   1184   1.56      yamt kstack_setup_magic(const struct lwp *l)
   1185   1.48      yamt {
   1186   1.85     perry 	uint32_t *ip;
   1187   1.85     perry 	uint32_t const *end;
   1188   1.48      yamt 
   1189   1.56      yamt 	KASSERT(l != NULL);
   1190   1.56      yamt 	KASSERT(l != &lwp0);
   1191   1.48      yamt 
   1192   1.48      yamt 	/*
   1193   1.48      yamt 	 * fill all the stack with magic number
   1194   1.48      yamt 	 * so that later modification on it can be detected.
   1195   1.48      yamt 	 */
   1196   1.85     perry 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
   1197  1.106  christos 	end = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
   1198   1.48      yamt 	for (; ip < end; ip++) {
   1199   1.48      yamt 		*ip = KSTACK_MAGIC;
   1200   1.48      yamt 	}
   1201   1.48      yamt }
   1202   1.48      yamt 
   1203   1.48      yamt void
   1204   1.56      yamt kstack_check_magic(const struct lwp *l)
   1205   1.48      yamt {
   1206   1.85     perry 	uint32_t const *ip, *end;
   1207   1.48      yamt 	int stackleft;
   1208   1.48      yamt 
   1209   1.56      yamt 	KASSERT(l != NULL);
   1210   1.48      yamt 
   1211   1.48      yamt 	/* don't check proc0 */ /*XXX*/
   1212   1.56      yamt 	if (l == &lwp0)
   1213   1.48      yamt 		return;
   1214   1.48      yamt 
   1215   1.48      yamt #ifdef __MACHINE_STACK_GROWS_UP
   1216   1.48      yamt 	/* stack grows upwards (eg. hppa) */
   1217  1.106  christos 	ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
   1218   1.85     perry 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
   1219   1.48      yamt 	for (ip--; ip >= end; ip--)
   1220   1.48      yamt 		if (*ip != KSTACK_MAGIC)
   1221   1.48      yamt 			break;
   1222   1.72  junyoung 
   1223  1.106  christos 	stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
   1224   1.48      yamt #else /* __MACHINE_STACK_GROWS_UP */
   1225   1.48      yamt 	/* stack grows downwards (eg. i386) */
   1226   1.85     perry 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
   1227  1.106  christos 	end = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
   1228   1.48      yamt 	for (; ip < end; ip++)
   1229   1.48      yamt 		if (*ip != KSTACK_MAGIC)
   1230   1.48      yamt 			break;
   1231   1.48      yamt 
   1232   1.93  christos 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
   1233   1.48      yamt #endif /* __MACHINE_STACK_GROWS_UP */
   1234   1.48      yamt 
   1235   1.48      yamt 	if (kstackleftmin > stackleft) {
   1236   1.48      yamt 		kstackleftmin = stackleft;
   1237   1.48      yamt 		if (stackleft < kstackleftthres)
   1238   1.56      yamt 			printf("warning: kernel stack left %d bytes"
   1239   1.56      yamt 			    "(pid %u:lid %u)\n", stackleft,
   1240   1.56      yamt 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
   1241   1.48      yamt 	}
   1242   1.48      yamt 
   1243   1.48      yamt 	if (stackleft <= 0) {
   1244   1.56      yamt 		panic("magic on the top of kernel stack changed for "
   1245   1.56      yamt 		    "pid %u, lid %u: maybe kernel stack overflow",
   1246   1.56      yamt 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
   1247   1.48      yamt 	}
   1248   1.48      yamt }
   1249   1.50     enami #endif /* KSTACK_CHECK_MAGIC */
   1250   1.79      yamt 
   1251  1.100        ad /*
   1252  1.100        ad  * XXXSMP this is bust, it grabs a read lock and then messes about
   1253  1.100        ad  * with allproc.
   1254  1.100        ad  */
   1255   1.79      yamt int
   1256   1.79      yamt proclist_foreach_call(struct proclist *list,
   1257   1.79      yamt     int (*callback)(struct proc *, void *arg), void *arg)
   1258   1.79      yamt {
   1259   1.79      yamt 	struct proc marker;
   1260   1.79      yamt 	struct proc *p;
   1261   1.79      yamt 	struct lwp * const l = curlwp;
   1262   1.79      yamt 	int ret = 0;
   1263   1.79      yamt 
   1264  1.102     pavel 	marker.p_flag = PK_MARKER;
   1265   1.79      yamt 	PHOLD(l);
   1266  1.107        ad 	mutex_enter(&proclist_lock);
   1267   1.79      yamt 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
   1268  1.102     pavel 		if (p->p_flag & PK_MARKER) {
   1269   1.79      yamt 			p = LIST_NEXT(p, p_list);
   1270   1.79      yamt 			continue;
   1271   1.79      yamt 		}
   1272   1.79      yamt 		LIST_INSERT_AFTER(p, &marker, p_list);
   1273   1.79      yamt 		ret = (*callback)(p, arg);
   1274  1.107        ad 		KASSERT(mutex_owned(&proclist_lock));
   1275   1.79      yamt 		p = LIST_NEXT(&marker, p_list);
   1276   1.79      yamt 		LIST_REMOVE(&marker, p_list);
   1277   1.79      yamt 	}
   1278  1.107        ad 	mutex_exit(&proclist_lock);
   1279   1.79      yamt 	PRELE(l);
   1280   1.79      yamt 
   1281   1.79      yamt 	return ret;
   1282   1.79      yamt }
   1283   1.86      yamt 
   1284   1.86      yamt int
   1285   1.86      yamt proc_vmspace_getref(struct proc *p, struct vmspace **vm)
   1286   1.86      yamt {
   1287   1.86      yamt 
   1288   1.86      yamt 	/* XXXCDC: how should locking work here? */
   1289   1.86      yamt 
   1290   1.87      yamt 	/* curproc exception is for coredump. */
   1291   1.87      yamt 
   1292  1.100        ad 	if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
   1293   1.86      yamt 	    (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
   1294   1.86      yamt 		return EFAULT;
   1295   1.86      yamt 	}
   1296   1.86      yamt 
   1297   1.86      yamt 	uvmspace_addref(p->p_vmspace);
   1298   1.86      yamt 	*vm = p->p_vmspace;
   1299   1.86      yamt 
   1300   1.86      yamt 	return 0;
   1301   1.86      yamt }
   1302   1.94        ad 
   1303   1.94        ad /*
   1304   1.94        ad  * Acquire a write lock on the process credential.
   1305   1.94        ad  */
   1306   1.94        ad void
   1307  1.100        ad proc_crmod_enter(void)
   1308   1.94        ad {
   1309  1.100        ad 	struct lwp *l = curlwp;
   1310  1.100        ad 	struct proc *p = l->l_proc;
   1311  1.100        ad 	struct plimit *lim;
   1312  1.100        ad 	kauth_cred_t oc;
   1313  1.100        ad 	char *cn;
   1314   1.94        ad 
   1315  1.100        ad 	mutex_enter(&p->p_mutex);
   1316  1.100        ad 
   1317  1.100        ad 	/* Ensure the LWP cached credentials are up to date. */
   1318  1.100        ad 	if ((oc = l->l_cred) != p->p_cred) {
   1319  1.100        ad 		kauth_cred_hold(p->p_cred);
   1320  1.100        ad 		l->l_cred = p->p_cred;
   1321  1.100        ad 		kauth_cred_free(oc);
   1322  1.100        ad 	}
   1323  1.100        ad 
   1324  1.100        ad 	/* Reset what needs to be reset in plimit. */
   1325  1.100        ad 	lim = p->p_limit;
   1326  1.100        ad 	if (lim->pl_corename != defcorename) {
   1327  1.100        ad 		if (lim->p_refcnt > 1 &&
   1328  1.100        ad 		    (lim->p_lflags & PL_SHAREMOD) == 0) {
   1329  1.100        ad 			p->p_limit = limcopy(p);
   1330  1.100        ad 			limfree(lim);
   1331  1.100        ad 			lim = p->p_limit;
   1332  1.100        ad 		}
   1333  1.100        ad 		simple_lock(&lim->p_slock);
   1334  1.100        ad 		cn = lim->pl_corename;
   1335  1.100        ad 		lim->pl_corename = defcorename;
   1336  1.100        ad 		simple_unlock(&lim->p_slock);
   1337  1.100        ad 		if (cn != defcorename)
   1338  1.100        ad 			free(cn, M_TEMP);
   1339  1.100        ad 	}
   1340   1.94        ad }
   1341   1.94        ad 
   1342   1.94        ad /*
   1343  1.100        ad  * Set in a new process credential, and drop the write lock.  The credential
   1344  1.100        ad  * must have a reference already.  Optionally, free a no-longer required
   1345  1.100        ad  * credential.  The scheduler also needs to inspect p_cred, so we also
   1346  1.100        ad  * briefly acquire the sched state mutex.
   1347   1.94        ad  */
   1348   1.94        ad void
   1349  1.104   thorpej proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
   1350   1.94        ad {
   1351  1.100        ad 	struct lwp *l = curlwp;
   1352  1.100        ad 	struct proc *p = l->l_proc;
   1353  1.100        ad 	kauth_cred_t oc;
   1354  1.100        ad 
   1355  1.100        ad 	/* Is there a new credential to set in? */
   1356  1.100        ad 	if (scred != NULL) {
   1357  1.100        ad 		mutex_enter(&p->p_smutex);
   1358  1.100        ad 		p->p_cred = scred;
   1359  1.100        ad 		mutex_exit(&p->p_smutex);
   1360  1.100        ad 
   1361  1.100        ad 		/* Ensure the LWP cached credentials are up to date. */
   1362  1.100        ad 		if ((oc = l->l_cred) != scred) {
   1363  1.100        ad 			kauth_cred_hold(scred);
   1364  1.100        ad 			l->l_cred = scred;
   1365  1.100        ad 		}
   1366  1.100        ad 	} else
   1367  1.100        ad 		oc = NULL;	/* XXXgcc */
   1368  1.100        ad 
   1369  1.100        ad 	if (sugid) {
   1370  1.100        ad 		/*
   1371  1.100        ad 		 * Mark process as having changed credentials, stops
   1372  1.100        ad 		 * tracing etc.
   1373  1.100        ad 		 */
   1374  1.102     pavel 		p->p_flag |= PK_SUGID;
   1375  1.100        ad 	}
   1376   1.94        ad 
   1377  1.100        ad 	mutex_exit(&p->p_mutex);
   1378  1.100        ad 
   1379  1.100        ad 	/* If there is a credential to be released, free it now. */
   1380  1.100        ad 	if (fcred != NULL) {
   1381  1.100        ad 		KASSERT(scred != NULL);
   1382   1.94        ad 		kauth_cred_free(fcred);
   1383  1.100        ad 		if (oc != scred)
   1384  1.100        ad 			kauth_cred_free(oc);
   1385  1.100        ad 	}
   1386  1.100        ad }
   1387  1.100        ad 
   1388  1.100        ad /*
   1389  1.100        ad  * Acquire a reference on a process, to prevent it from exiting or execing.
   1390  1.100        ad  */
   1391  1.100        ad int
   1392  1.100        ad proc_addref(struct proc *p)
   1393  1.100        ad {
   1394  1.100        ad 
   1395  1.107        ad 	KASSERT(mutex_owned(&p->p_mutex));
   1396  1.100        ad 
   1397  1.100        ad 	if (p->p_refcnt <= 0)
   1398  1.100        ad 		return EAGAIN;
   1399  1.100        ad 	p->p_refcnt++;
   1400  1.100        ad 
   1401  1.100        ad 	return 0;
   1402  1.100        ad }
   1403  1.100        ad 
   1404  1.100        ad /*
   1405  1.100        ad  * Release a reference on a process.
   1406  1.100        ad  */
   1407  1.100        ad void
   1408  1.100        ad proc_delref(struct proc *p)
   1409  1.100        ad {
   1410  1.100        ad 
   1411  1.107        ad 	KASSERT(mutex_owned(&p->p_mutex));
   1412  1.100        ad 
   1413  1.100        ad 	if (p->p_refcnt < 0) {
   1414  1.100        ad 		if (++p->p_refcnt == 0)
   1415  1.100        ad 			cv_broadcast(&p->p_refcv);
   1416  1.100        ad 	} else {
   1417  1.100        ad 		p->p_refcnt--;
   1418  1.100        ad 		KASSERT(p->p_refcnt != 0);
   1419  1.100        ad 	}
   1420  1.100        ad }
   1421  1.100        ad 
   1422  1.100        ad /*
   1423  1.100        ad  * Wait for all references on the process to drain, and prevent new
   1424  1.100        ad  * references from being acquired.
   1425  1.100        ad  */
   1426  1.100        ad void
   1427  1.100        ad proc_drainrefs(struct proc *p)
   1428  1.100        ad {
   1429  1.100        ad 
   1430  1.107        ad 	KASSERT(mutex_owned(&p->p_mutex));
   1431  1.111      yamt 	KASSERT(p->p_refcnt >= 0);
   1432  1.100        ad 
   1433  1.100        ad 	/*
   1434  1.100        ad 	 * The process itself holds the last reference.  Once it's released,
   1435  1.100        ad 	 * no new references will be granted.  If we have already locked out
   1436  1.100        ad 	 * new references (refcnt <= 0), potentially due to a failed exec,
   1437  1.100        ad 	 * there is nothing more to do.
   1438  1.100        ad 	 */
   1439  1.111      yamt 	if (p->p_refcnt == 0)
   1440  1.111      yamt 		return;
   1441  1.100        ad 	p->p_refcnt = 1 - p->p_refcnt;
   1442  1.100        ad 	while (p->p_refcnt != 0)
   1443  1.100        ad 		cv_wait(&p->p_refcv, &p->p_mutex);
   1444   1.94        ad }
   1445   1.95   thorpej 
   1446   1.95   thorpej /*
   1447   1.95   thorpej  * proc_specific_key_create --
   1448   1.95   thorpej  *	Create a key for subsystem proc-specific data.
   1449   1.95   thorpej  */
   1450   1.95   thorpej int
   1451   1.95   thorpej proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
   1452   1.95   thorpej {
   1453   1.95   thorpej 
   1454   1.98   thorpej 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
   1455   1.95   thorpej }
   1456   1.95   thorpej 
   1457   1.95   thorpej /*
   1458   1.95   thorpej  * proc_specific_key_delete --
   1459   1.95   thorpej  *	Delete a key for subsystem proc-specific data.
   1460   1.95   thorpej  */
   1461   1.95   thorpej void
   1462   1.95   thorpej proc_specific_key_delete(specificdata_key_t key)
   1463   1.95   thorpej {
   1464   1.95   thorpej 
   1465   1.95   thorpej 	specificdata_key_delete(proc_specificdata_domain, key);
   1466   1.95   thorpej }
   1467   1.95   thorpej 
   1468   1.98   thorpej /*
   1469   1.98   thorpej  * proc_initspecific --
   1470   1.98   thorpej  *	Initialize a proc's specificdata container.
   1471   1.98   thorpej  */
   1472   1.96  christos void
   1473   1.96  christos proc_initspecific(struct proc *p)
   1474   1.96  christos {
   1475   1.96  christos 	int error;
   1476   1.98   thorpej 
   1477   1.96  christos 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
   1478   1.96  christos 	KASSERT(error == 0);
   1479   1.96  christos }
   1480   1.96  christos 
   1481   1.95   thorpej /*
   1482   1.98   thorpej  * proc_finispecific --
   1483   1.98   thorpej  *	Finalize a proc's specificdata container.
   1484   1.98   thorpej  */
   1485   1.98   thorpej void
   1486   1.98   thorpej proc_finispecific(struct proc *p)
   1487   1.98   thorpej {
   1488   1.98   thorpej 
   1489   1.98   thorpej 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
   1490   1.98   thorpej }
   1491   1.98   thorpej 
   1492   1.98   thorpej /*
   1493   1.95   thorpej  * proc_getspecific --
   1494   1.95   thorpej  *	Return proc-specific data corresponding to the specified key.
   1495   1.95   thorpej  */
   1496   1.95   thorpej void *
   1497   1.95   thorpej proc_getspecific(struct proc *p, specificdata_key_t key)
   1498   1.95   thorpej {
   1499   1.95   thorpej 
   1500   1.95   thorpej 	return (specificdata_getspecific(proc_specificdata_domain,
   1501   1.95   thorpej 					 &p->p_specdataref, key));
   1502   1.95   thorpej }
   1503   1.95   thorpej 
   1504   1.95   thorpej /*
   1505   1.95   thorpej  * proc_setspecific --
   1506   1.95   thorpej  *	Set proc-specific data corresponding to the specified key.
   1507   1.95   thorpej  */
   1508   1.95   thorpej void
   1509   1.95   thorpej proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
   1510   1.95   thorpej {
   1511   1.95   thorpej 
   1512   1.95   thorpej 	specificdata_setspecific(proc_specificdata_domain,
   1513   1.95   thorpej 				 &p->p_specdataref, key, data);
   1514   1.95   thorpej }
   1515