Home | History | Annotate | Line # | Download | only in kern
kern_proc.c revision 1.121
      1  1.121        ad /*	$NetBSD: kern_proc.c,v 1.121 2007/11/06 00:42:42 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  1.121        ad __KERNEL_RCSID(0, "$NetBSD: kern_proc.c,v 1.121 2007/11/06 00:42:42 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  1.100        ad 	mutex_init(&p->p_smutex, MUTEX_SPIN, IPL_SCHED);
    310  1.112        ad 	mutex_init(&p->p_stmutex, MUTEX_SPIN, IPL_HIGH);
    311  1.120        ad 	mutex_init(&p->p_raslock, MUTEX_DEFAULT, IPL_NONE);
    312  1.100        ad 	mutex_init(&p->p_mutex, MUTEX_DEFAULT, IPL_NONE);
    313  1.113        ad 	mutex_init(&l->l_swaplock, MUTEX_DEFAULT, IPL_NONE);
    314  1.107        ad 
    315  1.100        ad 	cv_init(&p->p_refcv, "drainref");
    316  1.100        ad 	cv_init(&p->p_waitcv, "wait");
    317  1.100        ad 	cv_init(&p->p_lwpcv, "lwpwait");
    318  1.100        ad 
    319   1.81  junyoung 	LIST_INIT(&p->p_lwps);
    320  1.100        ad 	LIST_INIT(&p->p_sigwaiters);
    321   1.81  junyoung 	LIST_INSERT_HEAD(&p->p_lwps, l, l_sibling);
    322  1.100        ad 
    323   1.81  junyoung 	p->p_nlwps = 1;
    324  1.100        ad 	p->p_nrlwps = 1;
    325  1.110      yamt 	p->p_nlwpid = l->l_lid;
    326  1.100        ad 	p->p_refcnt = 1;
    327   1.81  junyoung 
    328   1.81  junyoung 	pid_table[0].pt_proc = p;
    329   1.81  junyoung 	LIST_INSERT_HEAD(&allproc, p, p_list);
    330   1.81  junyoung 	LIST_INSERT_HEAD(&alllwp, l, l_list);
    331   1.81  junyoung 
    332   1.81  junyoung 	p->p_pgrp = pg;
    333   1.81  junyoung 	pid_table[0].pt_pgrp = pg;
    334   1.81  junyoung 	LIST_INIT(&pg->pg_members);
    335   1.81  junyoung 	LIST_INSERT_HEAD(&pg->pg_members, p, p_pglist);
    336   1.81  junyoung 
    337   1.81  junyoung 	pg->pg_session = sess;
    338   1.81  junyoung 	sess->s_count = 1;
    339   1.81  junyoung 	sess->s_sid = 0;
    340   1.81  junyoung 	sess->s_leader = p;
    341   1.81  junyoung 
    342   1.81  junyoung 	/*
    343   1.81  junyoung 	 * Set P_NOCLDWAIT so that kernel threads are reparented to
    344   1.81  junyoung 	 * init(8) when they exit.  init(8) can easily wait them out
    345   1.81  junyoung 	 * for us.
    346   1.81  junyoung 	 */
    347  1.102     pavel 	p->p_flag = PK_SYSTEM | PK_NOCLDWAIT;
    348   1.81  junyoung 	p->p_stat = SACTIVE;
    349   1.81  junyoung 	p->p_nice = NZERO;
    350   1.81  junyoung 	p->p_emul = &emul_netbsd;
    351   1.81  junyoung #ifdef __HAVE_SYSCALL_INTERN
    352   1.81  junyoung 	(*p->p_emul->e_syscall_intern)(p);
    353   1.81  junyoung #endif
    354  1.110      yamt 	strlcpy(p->p_comm, "system", sizeof(p->p_comm));
    355   1.81  junyoung 
    356  1.102     pavel 	l->l_flag = LW_INMEM | LW_SYSTEM;
    357   1.81  junyoung 	l->l_stat = LSONPROC;
    358  1.100        ad 	l->l_ts = &turnstile0;
    359  1.100        ad 	l->l_syncobj = &sched_syncobj;
    360  1.100        ad 	l->l_refcnt = 1;
    361  1.100        ad 	l->l_cpu = curcpu();
    362  1.121        ad 	l->l_priority = PRI_USER + NPRI_USER - 1;
    363  1.121        ad 	l->l_inheritedprio = -1;
    364  1.121        ad 	l->l_class = SCHED_OTHER;
    365  1.105      yamt 	SLIST_INIT(&l->l_pi_lenders);
    366  1.113        ad 	l->l_name = __UNCONST("swapper");
    367   1.81  junyoung 
    368  1.121        ad 	callout_init(&l->l_timeout_ch, CALLOUT_MPSAFE);
    369  1.115        ad 	callout_setfunc(&l->l_timeout_ch, sleepq_timeout, l);
    370  1.100        ad 	cv_init(&l->l_sigcv, "sigwait");
    371   1.81  junyoung 
    372   1.81  junyoung 	/* Create credentials. */
    373   1.89      elad 	cred0 = kauth_cred_alloc();
    374   1.89      elad 	p->p_cred = cred0;
    375  1.100        ad 	kauth_cred_hold(cred0);
    376  1.100        ad 	l->l_cred = cred0;
    377   1.81  junyoung 
    378   1.81  junyoung 	/* Create the CWD info. */
    379   1.81  junyoung 	p->p_cwdi = &cwdi0;
    380   1.81  junyoung 	cwdi0.cwdi_cmask = cmask;
    381   1.81  junyoung 	cwdi0.cwdi_refcnt = 1;
    382  1.113        ad 	rw_init(&cwdi0.cwdi_lock);
    383   1.81  junyoung 
    384   1.81  junyoung 	/* Create the limits structures. */
    385   1.81  junyoung 	p->p_limit = &limit0;
    386  1.116       dsl 	mutex_init(&limit0.pl_lock, MUTEX_DEFAULT, IPL_NONE);
    387   1.81  junyoung 	for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++)
    388   1.81  junyoung 		limit0.pl_rlimit[i].rlim_cur =
    389   1.81  junyoung 		    limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
    390   1.81  junyoung 
    391   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_max = maxfiles;
    392   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur =
    393   1.81  junyoung 	    maxfiles < nofile ? maxfiles : nofile;
    394   1.81  junyoung 
    395   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
    396   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur =
    397   1.81  junyoung 	    maxproc < maxuprc ? maxproc : maxuprc;
    398   1.81  junyoung 
    399   1.81  junyoung 	lim = ptoa(uvmexp.free);
    400   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_RSS].rlim_max = lim;
    401   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_max = lim;
    402   1.81  junyoung 	limit0.pl_rlimit[RLIMIT_MEMLOCK].rlim_cur = lim / 3;
    403   1.81  junyoung 	limit0.pl_corename = defcorename;
    404  1.116       dsl 	limit0.pl_refcnt = 1;
    405  1.117       dsl 	limit0.pl_sv_limit = NULL;
    406   1.81  junyoung 
    407   1.81  junyoung 	/* Configure virtual memory system, set vm rlimits. */
    408   1.81  junyoung 	uvm_init_limits(p);
    409   1.81  junyoung 
    410   1.81  junyoung 	/* Initialize file descriptor table for proc0. */
    411   1.81  junyoung 	p->p_fd = &filedesc0.fd_fd;
    412   1.81  junyoung 	fdinit1(&filedesc0);
    413   1.81  junyoung 
    414   1.81  junyoung 	/*
    415   1.81  junyoung 	 * Initialize proc0's vmspace, which uses the kernel pmap.
    416   1.81  junyoung 	 * All kernel processes (which never have user space mappings)
    417   1.81  junyoung 	 * share proc0's vmspace, and thus, the kernel pmap.
    418   1.81  junyoung 	 */
    419   1.81  junyoung 	uvmspace_init(&vmspace0, pmap_kernel(), round_page(VM_MIN_ADDRESS),
    420   1.81  junyoung 	    trunc_page(VM_MAX_ADDRESS));
    421   1.81  junyoung 	p->p_vmspace = &vmspace0;
    422   1.81  junyoung 
    423   1.81  junyoung 	l->l_addr = proc0paddr;				/* XXX */
    424   1.81  junyoung 
    425   1.81  junyoung 	p->p_stats = &pstat0;
    426   1.81  junyoung 
    427   1.81  junyoung 	/* Initialize signal state for proc0. */
    428   1.81  junyoung 	p->p_sigacts = &sigacts0;
    429  1.100        ad 	mutex_init(&p->p_sigacts->sa_mutex, MUTEX_SPIN, IPL_NONE);
    430   1.81  junyoung 	siginit(p);
    431   1.96  christos 
    432   1.96  christos 	proc_initspecific(p);
    433   1.96  christos 	lwp_initspecific(l);
    434  1.103       dsl 
    435  1.103       dsl 	SYSCALL_TIME_LWP_INIT(l);
    436   1.81  junyoung }
    437   1.81  junyoung 
    438   1.81  junyoung /*
    439   1.74  junyoung  * Check that the specified process group is in the session of the
    440   1.60       dsl  * specified process.
    441   1.60       dsl  * Treats -ve ids as process ids.
    442   1.60       dsl  * Used to validate TIOCSPGRP requests.
    443   1.60       dsl  */
    444   1.60       dsl int
    445   1.60       dsl pgid_in_session(struct proc *p, pid_t pg_id)
    446   1.60       dsl {
    447   1.60       dsl 	struct pgrp *pgrp;
    448  1.101       dsl 	struct session *session;
    449  1.107        ad 	int error;
    450  1.101       dsl 
    451  1.107        ad 	mutex_enter(&proclist_lock);
    452   1.60       dsl 	if (pg_id < 0) {
    453  1.101       dsl 		struct proc *p1 = p_find(-pg_id, PFIND_LOCKED | PFIND_UNLOCK_FAIL);
    454   1.64       dsl 		if (p1 == NULL)
    455   1.64       dsl 			return EINVAL;
    456   1.60       dsl 		pgrp = p1->p_pgrp;
    457   1.60       dsl 	} else {
    458  1.101       dsl 		pgrp = pg_find(pg_id, PFIND_LOCKED | PFIND_UNLOCK_FAIL);
    459   1.60       dsl 		if (pgrp == NULL)
    460   1.64       dsl 			return EINVAL;
    461   1.60       dsl 	}
    462  1.101       dsl 	session = pgrp->pg_session;
    463  1.101       dsl 	if (session != p->p_pgrp->pg_session)
    464  1.107        ad 		error = EPERM;
    465  1.107        ad 	else
    466  1.107        ad 		error = 0;
    467  1.107        ad 	mutex_exit(&proclist_lock);
    468  1.107        ad 
    469  1.107        ad 	return error;
    470    1.7       cgd }
    471    1.4    andrew 
    472    1.1       cgd /*
    473   1.41  sommerfe  * Is p an inferior of q?
    474   1.94        ad  *
    475   1.94        ad  * Call with the proclist_lock held.
    476    1.1       cgd  */
    477   1.11       cgd int
    478   1.59       dsl inferior(struct proc *p, struct proc *q)
    479    1.1       cgd {
    480    1.1       cgd 
    481   1.41  sommerfe 	for (; p != q; p = p->p_pptr)
    482    1.1       cgd 		if (p->p_pid == 0)
    483   1.82  junyoung 			return 0;
    484   1.82  junyoung 	return 1;
    485    1.1       cgd }
    486    1.1       cgd 
    487    1.1       cgd /*
    488    1.1       cgd  * Locate a process by number
    489    1.1       cgd  */
    490    1.1       cgd struct proc *
    491   1.68       dsl p_find(pid_t pid, uint flags)
    492    1.1       cgd {
    493   1.33   thorpej 	struct proc *p;
    494   1.68       dsl 	char stat;
    495    1.1       cgd 
    496   1.68       dsl 	if (!(flags & PFIND_LOCKED))
    497  1.107        ad 		mutex_enter(&proclist_lock);
    498  1.100        ad 
    499   1.61       dsl 	p = pid_table[pid & pid_tbl_mask].pt_proc;
    500  1.100        ad 
    501   1.61       dsl 	/* Only allow live processes to be found by pid. */
    502  1.100        ad 	/* XXXSMP p_stat */
    503  1.100        ad 	if (P_VALID(p) && p->p_pid == pid && ((stat = p->p_stat) == SACTIVE ||
    504  1.100        ad 	    stat == SSTOP || ((flags & PFIND_ZOMBIE) &&
    505  1.100        ad 	    (stat == SZOMB || stat == SDEAD || stat == SDYING)))) {
    506   1.68       dsl 		if (flags & PFIND_UNLOCK_OK)
    507  1.107        ad 			 mutex_exit(&proclist_lock);
    508   1.68       dsl 		return p;
    509   1.68       dsl 	}
    510   1.68       dsl 	if (flags & PFIND_UNLOCK_FAIL)
    511  1.107        ad 		mutex_exit(&proclist_lock);
    512   1.68       dsl 	return NULL;
    513    1.1       cgd }
    514    1.1       cgd 
    515   1.61       dsl 
    516    1.1       cgd /*
    517    1.1       cgd  * Locate a process group by number
    518    1.1       cgd  */
    519    1.1       cgd struct pgrp *
    520   1.68       dsl pg_find(pid_t pgid, uint flags)
    521    1.1       cgd {
    522   1.68       dsl 	struct pgrp *pg;
    523    1.1       cgd 
    524   1.68       dsl 	if (!(flags & PFIND_LOCKED))
    525  1.107        ad 		mutex_enter(&proclist_lock);
    526   1.68       dsl 	pg = pid_table[pgid & pid_tbl_mask].pt_pgrp;
    527   1.61       dsl 	/*
    528   1.61       dsl 	 * Can't look up a pgrp that only exists because the session
    529   1.61       dsl 	 * hasn't died yet (traditional)
    530   1.61       dsl 	 */
    531   1.68       dsl 	if (pg == NULL || pg->pg_id != pgid || LIST_EMPTY(&pg->pg_members)) {
    532   1.68       dsl 		if (flags & PFIND_UNLOCK_FAIL)
    533  1.107        ad 			 mutex_exit(&proclist_lock);
    534   1.68       dsl 		return NULL;
    535   1.68       dsl 	}
    536   1.68       dsl 
    537   1.68       dsl 	if (flags & PFIND_UNLOCK_OK)
    538  1.107        ad 		mutex_exit(&proclist_lock);
    539   1.68       dsl 	return pg;
    540    1.1       cgd }
    541    1.1       cgd 
    542   1.61       dsl static void
    543   1.61       dsl expand_pid_table(void)
    544    1.1       cgd {
    545   1.61       dsl 	uint pt_size = pid_tbl_mask + 1;
    546   1.61       dsl 	struct pid_table *n_pt, *new_pt;
    547   1.61       dsl 	struct proc *proc;
    548   1.61       dsl 	struct pgrp *pgrp;
    549   1.61       dsl 	int i;
    550   1.61       dsl 	pid_t pid;
    551    1.1       cgd 
    552   1.61       dsl 	new_pt = malloc(pt_size * 2 * sizeof *new_pt, M_PROC, M_WAITOK);
    553   1.61       dsl 
    554  1.107        ad 	mutex_enter(&proclist_lock);
    555   1.61       dsl 	if (pt_size != pid_tbl_mask + 1) {
    556   1.61       dsl 		/* Another process beat us to it... */
    557  1.107        ad 		mutex_exit(&proclist_lock);
    558   1.61       dsl 		FREE(new_pt, M_PROC);
    559   1.61       dsl 		return;
    560   1.61       dsl 	}
    561   1.72  junyoung 
    562   1.61       dsl 	/*
    563   1.61       dsl 	 * Copy entries from old table into new one.
    564   1.61       dsl 	 * If 'pid' is 'odd' we need to place in the upper half,
    565   1.61       dsl 	 * even pid's to the lower half.
    566   1.61       dsl 	 * Free items stay in the low half so we don't have to
    567   1.61       dsl 	 * fixup the reference to them.
    568   1.61       dsl 	 * We stuff free items on the front of the freelist
    569   1.61       dsl 	 * because we can't write to unmodified entries.
    570   1.74  junyoung 	 * Processing the table backwards maintains a semblance
    571   1.61       dsl 	 * of issueing pid numbers that increase with time.
    572   1.61       dsl 	 */
    573   1.61       dsl 	i = pt_size - 1;
    574   1.61       dsl 	n_pt = new_pt + i;
    575   1.61       dsl 	for (; ; i--, n_pt--) {
    576   1.61       dsl 		proc = pid_table[i].pt_proc;
    577   1.61       dsl 		pgrp = pid_table[i].pt_pgrp;
    578   1.61       dsl 		if (!P_VALID(proc)) {
    579   1.61       dsl 			/* Up 'use count' so that link is valid */
    580   1.61       dsl 			pid = (P_NEXT(proc) + pt_size) & ~pt_size;
    581   1.61       dsl 			proc = P_FREE(pid);
    582   1.61       dsl 			if (pgrp)
    583   1.61       dsl 				pid = pgrp->pg_id;
    584   1.61       dsl 		} else
    585   1.61       dsl 			pid = proc->p_pid;
    586   1.72  junyoung 
    587   1.61       dsl 		/* Save entry in appropriate half of table */
    588   1.61       dsl 		n_pt[pid & pt_size].pt_proc = proc;
    589   1.61       dsl 		n_pt[pid & pt_size].pt_pgrp = pgrp;
    590   1.61       dsl 
    591   1.61       dsl 		/* Put other piece on start of free list */
    592   1.61       dsl 		pid = (pid ^ pt_size) & ~pid_tbl_mask;
    593   1.61       dsl 		n_pt[pid & pt_size].pt_proc =
    594   1.61       dsl 				    P_FREE((pid & ~pt_size) | next_free_pt);
    595   1.61       dsl 		n_pt[pid & pt_size].pt_pgrp = 0;
    596   1.61       dsl 		next_free_pt = i | (pid & pt_size);
    597   1.61       dsl 		if (i == 0)
    598   1.61       dsl 			break;
    599   1.61       dsl 	}
    600   1.61       dsl 
    601   1.61       dsl 	/* Switch tables */
    602  1.100        ad 	mutex_enter(&proclist_mutex);
    603   1.61       dsl 	n_pt = pid_table;
    604   1.61       dsl 	pid_table = new_pt;
    605  1.100        ad 	mutex_exit(&proclist_mutex);
    606   1.61       dsl 	pid_tbl_mask = pt_size * 2 - 1;
    607   1.61       dsl 
    608   1.61       dsl 	/*
    609   1.61       dsl 	 * pid_max starts as PID_MAX (= 30000), once we have 16384
    610   1.61       dsl 	 * allocated pids we need it to be larger!
    611   1.61       dsl 	 */
    612   1.61       dsl 	if (pid_tbl_mask > PID_MAX) {
    613   1.61       dsl 		pid_max = pid_tbl_mask * 2 + 1;
    614   1.61       dsl 		pid_alloc_lim |= pid_alloc_lim << 1;
    615   1.61       dsl 	} else
    616   1.61       dsl 		pid_alloc_lim <<= 1;	/* doubles number of free slots... */
    617   1.61       dsl 
    618  1.107        ad 	mutex_exit(&proclist_lock);
    619   1.61       dsl 	FREE(n_pt, M_PROC);
    620   1.61       dsl }
    621   1.61       dsl 
    622   1.61       dsl struct proc *
    623   1.61       dsl proc_alloc(void)
    624   1.61       dsl {
    625   1.61       dsl 	struct proc *p;
    626  1.100        ad 	int nxt;
    627   1.61       dsl 	pid_t pid;
    628   1.61       dsl 	struct pid_table *pt;
    629   1.61       dsl 
    630   1.61       dsl 	p = pool_get(&proc_pool, PR_WAITOK);
    631   1.61       dsl 	p->p_stat = SIDL;			/* protect against others */
    632   1.61       dsl 
    633   1.96  christos 	proc_initspecific(p);
    634   1.61       dsl 	/* allocate next free pid */
    635   1.61       dsl 
    636   1.61       dsl 	for (;;expand_pid_table()) {
    637   1.61       dsl 		if (__predict_false(pid_alloc_cnt >= pid_alloc_lim))
    638   1.61       dsl 			/* ensure pids cycle through 2000+ values */
    639   1.61       dsl 			continue;
    640  1.107        ad 		mutex_enter(&proclist_lock);
    641   1.61       dsl 		pt = &pid_table[next_free_pt];
    642    1.1       cgd #ifdef DIAGNOSTIC
    643   1.63  christos 		if (__predict_false(P_VALID(pt->pt_proc) || pt->pt_pgrp))
    644   1.61       dsl 			panic("proc_alloc: slot busy");
    645    1.1       cgd #endif
    646   1.61       dsl 		nxt = P_NEXT(pt->pt_proc);
    647   1.61       dsl 		if (nxt & pid_tbl_mask)
    648   1.61       dsl 			break;
    649   1.61       dsl 		/* Table full - expand (NB last entry not used....) */
    650  1.107        ad 		mutex_exit(&proclist_lock);
    651   1.61       dsl 	}
    652   1.61       dsl 
    653   1.61       dsl 	/* pid is 'saved use count' + 'size' + entry */
    654   1.61       dsl 	pid = (nxt & ~pid_tbl_mask) + pid_tbl_mask + 1 + next_free_pt;
    655   1.61       dsl 	if ((uint)pid > (uint)pid_max)
    656   1.61       dsl 		pid &= pid_tbl_mask;
    657   1.61       dsl 	p->p_pid = pid;
    658   1.61       dsl 	next_free_pt = nxt & pid_tbl_mask;
    659   1.61       dsl 
    660   1.61       dsl 	/* Grab table slot */
    661  1.100        ad 	mutex_enter(&proclist_mutex);
    662   1.61       dsl 	pt->pt_proc = p;
    663  1.100        ad 	mutex_exit(&proclist_mutex);
    664   1.61       dsl 	pid_alloc_cnt++;
    665   1.61       dsl 
    666  1.107        ad 	mutex_exit(&proclist_lock);
    667   1.61       dsl 
    668   1.61       dsl 	return p;
    669   1.61       dsl }
    670   1.61       dsl 
    671   1.61       dsl /*
    672  1.118        ad  * Free a process id - called from proc_free (in kern_exit.c)
    673  1.100        ad  *
    674  1.118        ad  * Called with the proclist_lock held.
    675   1.61       dsl  */
    676   1.61       dsl void
    677  1.118        ad proc_free_pid(struct proc *p)
    678   1.61       dsl {
    679   1.61       dsl 	pid_t pid = p->p_pid;
    680   1.61       dsl 	struct pid_table *pt;
    681   1.61       dsl 
    682  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
    683   1.61       dsl 
    684   1.61       dsl 	pt = &pid_table[pid & pid_tbl_mask];
    685    1.1       cgd #ifdef DIAGNOSTIC
    686   1.63  christos 	if (__predict_false(pt->pt_proc != p))
    687   1.61       dsl 		panic("proc_free: pid_table mismatch, pid %x, proc %p",
    688   1.61       dsl 			pid, p);
    689    1.1       cgd #endif
    690  1.100        ad 	mutex_enter(&proclist_mutex);
    691   1.61       dsl 	/* save pid use count in slot */
    692   1.61       dsl 	pt->pt_proc = P_FREE(pid & ~pid_tbl_mask);
    693   1.61       dsl 
    694   1.61       dsl 	if (pt->pt_pgrp == NULL) {
    695   1.61       dsl 		/* link last freed entry onto ours */
    696   1.61       dsl 		pid &= pid_tbl_mask;
    697   1.61       dsl 		pt = &pid_table[last_free_pt];
    698   1.61       dsl 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pid);
    699   1.61       dsl 		last_free_pt = pid;
    700   1.61       dsl 		pid_alloc_cnt--;
    701   1.61       dsl 	}
    702  1.100        ad 	mutex_exit(&proclist_mutex);
    703   1.61       dsl 
    704   1.61       dsl 	nprocs--;
    705   1.61       dsl }
    706   1.61       dsl 
    707   1.61       dsl /*
    708   1.61       dsl  * Move p to a new or existing process group (and session)
    709   1.61       dsl  *
    710   1.61       dsl  * If we are creating a new pgrp, the pgid should equal
    711   1.72  junyoung  * the calling process' pid.
    712   1.61       dsl  * If is only valid to enter a process group that is in the session
    713   1.61       dsl  * of the process.
    714   1.61       dsl  * Also mksess should only be set if we are creating a process group
    715   1.61       dsl  *
    716   1.72  junyoung  * Only called from sys_setsid, sys_setpgid/sys_setpgrp and the
    717  1.100        ad  * SYSV setpgrp support for hpux.
    718   1.61       dsl  */
    719   1.61       dsl int
    720  1.100        ad enterpgrp(struct proc *curp, pid_t pid, pid_t pgid, int mksess)
    721   1.61       dsl {
    722   1.61       dsl 	struct pgrp *new_pgrp, *pgrp;
    723   1.61       dsl 	struct session *sess;
    724  1.100        ad 	struct proc *p;
    725   1.61       dsl 	int rval;
    726   1.61       dsl 	pid_t pg_id = NO_PGID;
    727   1.61       dsl 
    728   1.61       dsl 	if (mksess)
    729   1.99     pooka 		sess = pool_get(&session_pool, PR_WAITOK);
    730   1.61       dsl 	else
    731   1.61       dsl 		sess = NULL;
    732   1.61       dsl 
    733  1.107        ad 	/* Allocate data areas we might need before doing any validity checks */
    734  1.107        ad 	mutex_enter(&proclist_lock);		/* Because pid_table might change */
    735  1.107        ad 	if (pid_table[pgid & pid_tbl_mask].pt_pgrp == 0) {
    736  1.107        ad 		mutex_exit(&proclist_lock);
    737  1.107        ad 		new_pgrp = pool_get(&pgrp_pool, PR_WAITOK);
    738  1.107        ad 		mutex_enter(&proclist_lock);
    739  1.107        ad 	} else
    740  1.107        ad 		new_pgrp = NULL;
    741   1.61       dsl 	rval = EPERM;	/* most common error (to save typing) */
    742   1.61       dsl 
    743   1.61       dsl 	/* Check pgrp exists or can be created */
    744   1.61       dsl 	pgrp = pid_table[pgid & pid_tbl_mask].pt_pgrp;
    745   1.61       dsl 	if (pgrp != NULL && pgrp->pg_id != pgid)
    746   1.61       dsl 		goto done;
    747   1.61       dsl 
    748   1.61       dsl 	/* Can only set another process under restricted circumstances. */
    749  1.100        ad 	if (pid != curp->p_pid) {
    750   1.61       dsl 		/* must exist and be one of our children... */
    751  1.100        ad 		if ((p = p_find(pid, PFIND_LOCKED)) == NULL ||
    752  1.100        ad 		    !inferior(p, curp)) {
    753   1.61       dsl 			rval = ESRCH;
    754   1.61       dsl 			goto done;
    755   1.61       dsl 		}
    756   1.61       dsl 		/* ... in the same session... */
    757   1.61       dsl 		if (sess != NULL || p->p_session != curp->p_session)
    758   1.61       dsl 			goto done;
    759   1.61       dsl 		/* ... existing pgid must be in same session ... */
    760   1.61       dsl 		if (pgrp != NULL && pgrp->pg_session != p->p_session)
    761   1.61       dsl 			goto done;
    762   1.61       dsl 		/* ... and not done an exec. */
    763  1.102     pavel 		if (p->p_flag & PK_EXEC) {
    764   1.61       dsl 			rval = EACCES;
    765   1.61       dsl 			goto done;
    766   1.49     enami 		}
    767  1.100        ad 	} else {
    768  1.100        ad 		/* ... setsid() cannot re-enter a pgrp */
    769  1.100        ad 		if (mksess && (curp->p_pgid == curp->p_pid ||
    770  1.100        ad 		    pg_find(curp->p_pid, PFIND_LOCKED)))
    771  1.100        ad 			goto done;
    772  1.100        ad 		p = curp;
    773   1.61       dsl 	}
    774    1.1       cgd 
    775   1.61       dsl 	/* Changing the process group/session of a session
    776   1.61       dsl 	   leader is definitely off limits. */
    777   1.61       dsl 	if (SESS_LEADER(p)) {
    778   1.61       dsl 		if (sess == NULL && p->p_pgrp == pgrp)
    779   1.61       dsl 			/* unless it's a definite noop */
    780   1.61       dsl 			rval = 0;
    781   1.61       dsl 		goto done;
    782   1.61       dsl 	}
    783   1.61       dsl 
    784   1.61       dsl 	/* Can only create a process group with id of process */
    785   1.61       dsl 	if (pgrp == NULL && pgid != pid)
    786   1.61       dsl 		goto done;
    787   1.61       dsl 
    788   1.61       dsl 	/* Can only create a session if creating pgrp */
    789   1.61       dsl 	if (sess != NULL && pgrp != NULL)
    790   1.61       dsl 		goto done;
    791   1.61       dsl 
    792   1.61       dsl 	/* Check we allocated memory for a pgrp... */
    793   1.61       dsl 	if (pgrp == NULL && new_pgrp == NULL)
    794   1.61       dsl 		goto done;
    795   1.61       dsl 
    796   1.61       dsl 	/* Don't attach to 'zombie' pgrp */
    797   1.61       dsl 	if (pgrp != NULL && LIST_EMPTY(&pgrp->pg_members))
    798   1.61       dsl 		goto done;
    799   1.61       dsl 
    800   1.61       dsl 	/* Expect to succeed now */
    801   1.61       dsl 	rval = 0;
    802   1.61       dsl 
    803   1.61       dsl 	if (pgrp == p->p_pgrp)
    804   1.61       dsl 		/* nothing to do */
    805   1.61       dsl 		goto done;
    806   1.61       dsl 
    807   1.61       dsl 	/* Ok all setup, link up required structures */
    808  1.100        ad 
    809   1.61       dsl 	if (pgrp == NULL) {
    810   1.61       dsl 		pgrp = new_pgrp;
    811   1.61       dsl 		new_pgrp = 0;
    812   1.61       dsl 		if (sess != NULL) {
    813   1.21   thorpej 			sess->s_sid = p->p_pid;
    814    1.1       cgd 			sess->s_leader = p;
    815    1.1       cgd 			sess->s_count = 1;
    816    1.1       cgd 			sess->s_ttyvp = NULL;
    817    1.1       cgd 			sess->s_ttyp = NULL;
    818   1.58       dsl 			sess->s_flags = p->p_session->s_flags & ~S_LOGIN_SET;
    819   1.25     perry 			memcpy(sess->s_login, p->p_session->s_login,
    820    1.1       cgd 			    sizeof(sess->s_login));
    821  1.100        ad 			p->p_lflag &= ~PL_CONTROLT;
    822    1.1       cgd 		} else {
    823   1.61       dsl 			sess = p->p_pgrp->pg_session;
    824   1.61       dsl 			SESSHOLD(sess);
    825    1.1       cgd 		}
    826   1.61       dsl 		pgrp->pg_session = sess;
    827   1.61       dsl 		sess = 0;
    828   1.61       dsl 
    829    1.1       cgd 		pgrp->pg_id = pgid;
    830   1.10   mycroft 		LIST_INIT(&pgrp->pg_members);
    831   1.61       dsl #ifdef DIAGNOSTIC
    832   1.63  christos 		if (__predict_false(pid_table[pgid & pid_tbl_mask].pt_pgrp))
    833   1.61       dsl 			panic("enterpgrp: pgrp table slot in use");
    834   1.63  christos 		if (__predict_false(mksess && p != curp))
    835   1.63  christos 			panic("enterpgrp: mksession and p != curproc");
    836   1.61       dsl #endif
    837  1.100        ad 		mutex_enter(&proclist_mutex);
    838   1.61       dsl 		pid_table[pgid & pid_tbl_mask].pt_pgrp = pgrp;
    839    1.1       cgd 		pgrp->pg_jobc = 0;
    840  1.100        ad 	} else
    841  1.100        ad 		mutex_enter(&proclist_mutex);
    842  1.100        ad 
    843  1.100        ad #ifdef notyet
    844  1.100        ad 	/*
    845  1.100        ad 	 * If there's a controlling terminal for the current session, we
    846  1.100        ad 	 * have to interlock with it.  See ttread().
    847  1.100        ad 	 */
    848  1.100        ad 	if (p->p_session->s_ttyvp != NULL) {
    849  1.100        ad 		tp = p->p_session->s_ttyp;
    850  1.100        ad 		mutex_enter(&tp->t_mutex);
    851  1.100        ad 	} else
    852  1.100        ad 		tp = NULL;
    853  1.100        ad #endif
    854    1.1       cgd 
    855    1.1       cgd 	/*
    856    1.1       cgd 	 * Adjust eligibility of affected pgrps to participate in job control.
    857    1.1       cgd 	 * Increment eligibility counts before decrementing, otherwise we
    858    1.1       cgd 	 * could reach 0 spuriously during the first call.
    859    1.1       cgd 	 */
    860    1.1       cgd 	fixjobc(p, pgrp, 1);
    861    1.1       cgd 	fixjobc(p, p->p_pgrp, 0);
    862    1.1       cgd 
    863  1.100        ad 	/* Move process to requested group. */
    864   1.10   mycroft 	LIST_REMOVE(p, p_pglist);
    865   1.52      matt 	if (LIST_EMPTY(&p->p_pgrp->pg_members))
    866   1.61       dsl 		/* defer delete until we've dumped the lock */
    867   1.61       dsl 		pg_id = p->p_pgrp->pg_id;
    868    1.1       cgd 	p->p_pgrp = pgrp;
    869   1.10   mycroft 	LIST_INSERT_HEAD(&pgrp->pg_members, p, p_pglist);
    870  1.100        ad 	mutex_exit(&proclist_mutex);
    871  1.100        ad 
    872  1.100        ad #ifdef notyet
    873  1.100        ad 	/* Done with the swap; we can release the tty mutex. */
    874  1.100        ad 	if (tp != NULL)
    875  1.100        ad 		mutex_exit(&tp->t_mutex);
    876  1.100        ad #endif
    877   1.61       dsl 
    878   1.61       dsl     done:
    879  1.100        ad 	if (pg_id != NO_PGID)
    880  1.100        ad 		pg_delete(pg_id);
    881  1.107        ad 	mutex_exit(&proclist_lock);
    882   1.61       dsl 	if (sess != NULL)
    883   1.77    simonb 		pool_put(&session_pool, sess);
    884   1.61       dsl 	if (new_pgrp != NULL)
    885   1.61       dsl 		pool_put(&pgrp_pool, new_pgrp);
    886   1.63  christos #ifdef DEBUG_PGRP
    887   1.63  christos 	if (__predict_false(rval))
    888   1.61       dsl 		printf("enterpgrp(%d,%d,%d), curproc %d, rval %d\n",
    889   1.61       dsl 			pid, pgid, mksess, curp->p_pid, rval);
    890   1.61       dsl #endif
    891   1.61       dsl 	return rval;
    892    1.1       cgd }
    893    1.1       cgd 
    894    1.1       cgd /*
    895  1.100        ad  * Remove a process from its process group.  Must be called with the
    896  1.107        ad  * proclist_lock held.
    897    1.1       cgd  */
    898  1.100        ad void
    899   1.59       dsl leavepgrp(struct proc *p)
    900    1.1       cgd {
    901   1.61       dsl 	struct pgrp *pgrp;
    902    1.1       cgd 
    903  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
    904  1.100        ad 
    905  1.100        ad 	/*
    906  1.100        ad 	 * If there's a controlling terminal for the session, we have to
    907  1.100        ad 	 * interlock with it.  See ttread().
    908  1.100        ad 	 */
    909  1.100        ad 	mutex_enter(&proclist_mutex);
    910  1.100        ad #ifdef notyet
    911  1.100        ad 	if (p_>p_session->s_ttyvp != NULL) {
    912  1.100        ad 		tp = p->p_session->s_ttyp;
    913  1.100        ad 		mutex_enter(&tp->t_mutex);
    914  1.100        ad 	} else
    915  1.100        ad 		tp = NULL;
    916  1.100        ad #endif
    917  1.100        ad 
    918   1.61       dsl 	pgrp = p->p_pgrp;
    919   1.10   mycroft 	LIST_REMOVE(p, p_pglist);
    920   1.94        ad 	p->p_pgrp = NULL;
    921   1.61       dsl 
    922  1.100        ad #ifdef notyet
    923  1.100        ad 	if (tp != NULL)
    924  1.100        ad 		mutex_exit(&tp->t_mutex);
    925  1.100        ad #endif
    926  1.100        ad 	mutex_exit(&proclist_mutex);
    927  1.100        ad 
    928  1.100        ad 	if (LIST_EMPTY(&pgrp->pg_members))
    929  1.100        ad 		pg_delete(pgrp->pg_id);
    930   1.61       dsl }
    931   1.61       dsl 
    932  1.100        ad /*
    933  1.107        ad  * Free a process group.  Must be called with the proclist_lock held.
    934  1.100        ad  */
    935   1.61       dsl static void
    936   1.61       dsl pg_free(pid_t pg_id)
    937   1.61       dsl {
    938   1.61       dsl 	struct pgrp *pgrp;
    939   1.61       dsl 	struct pid_table *pt;
    940   1.61       dsl 
    941  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
    942  1.100        ad 
    943   1.61       dsl 	pt = &pid_table[pg_id & pid_tbl_mask];
    944   1.61       dsl 	pgrp = pt->pt_pgrp;
    945   1.61       dsl #ifdef DIAGNOSTIC
    946   1.63  christos 	if (__predict_false(!pgrp || pgrp->pg_id != pg_id
    947   1.63  christos 	    || !LIST_EMPTY(&pgrp->pg_members)))
    948   1.61       dsl 		panic("pg_free: process group absent or has members");
    949   1.61       dsl #endif
    950   1.61       dsl 	pt->pt_pgrp = 0;
    951   1.61       dsl 
    952   1.61       dsl 	if (!P_VALID(pt->pt_proc)) {
    953   1.61       dsl 		/* orphaned pgrp, put slot onto free list */
    954   1.61       dsl #ifdef DIAGNOSTIC
    955   1.63  christos 		if (__predict_false(P_NEXT(pt->pt_proc) & pid_tbl_mask))
    956   1.61       dsl 			panic("pg_free: process slot on free list");
    957   1.61       dsl #endif
    958  1.100        ad 		mutex_enter(&proclist_mutex);
    959   1.61       dsl 		pg_id &= pid_tbl_mask;
    960   1.61       dsl 		pt = &pid_table[last_free_pt];
    961   1.61       dsl 		pt->pt_proc = P_FREE(P_NEXT(pt->pt_proc) | pg_id);
    962  1.100        ad 		mutex_exit(&proclist_mutex);
    963   1.61       dsl 		last_free_pt = pg_id;
    964   1.61       dsl 		pid_alloc_cnt--;
    965   1.61       dsl 	}
    966   1.61       dsl 	pool_put(&pgrp_pool, pgrp);
    967    1.1       cgd }
    968    1.1       cgd 
    969    1.1       cgd /*
    970  1.107        ad  * Delete a process group.  Must be called with the proclist_lock held.
    971    1.1       cgd  */
    972   1.61       dsl static void
    973   1.61       dsl pg_delete(pid_t pg_id)
    974   1.61       dsl {
    975   1.61       dsl 	struct pgrp *pgrp;
    976   1.61       dsl 	struct tty *ttyp;
    977   1.61       dsl 	struct session *ss;
    978  1.100        ad 	int is_pgrp_leader;
    979  1.100        ad 
    980  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
    981   1.61       dsl 
    982   1.61       dsl 	pgrp = pid_table[pg_id & pid_tbl_mask].pt_pgrp;
    983   1.61       dsl 	if (pgrp == NULL || pgrp->pg_id != pg_id ||
    984  1.100        ad 	    !LIST_EMPTY(&pgrp->pg_members))
    985   1.61       dsl 		return;
    986   1.61       dsl 
    987   1.71        pk 	ss = pgrp->pg_session;
    988   1.71        pk 
    989   1.61       dsl 	/* Remove reference (if any) from tty to this process group */
    990   1.71        pk 	ttyp = ss->s_ttyp;
    991   1.71        pk 	if (ttyp != NULL && ttyp->t_pgrp == pgrp) {
    992   1.61       dsl 		ttyp->t_pgrp = NULL;
    993   1.71        pk #ifdef DIAGNOSTIC
    994   1.71        pk 		if (ttyp->t_session != ss)
    995   1.71        pk 			panic("pg_delete: wrong session on terminal");
    996   1.71        pk #endif
    997   1.71        pk 	}
    998   1.61       dsl 
    999   1.71        pk 	/*
   1000   1.71        pk 	 * The leading process group in a session is freed
   1001   1.71        pk 	 * by sessdelete() if last reference.
   1002   1.71        pk 	 */
   1003   1.71        pk 	is_pgrp_leader = (ss->s_sid == pgrp->pg_id);
   1004   1.71        pk 	SESSRELE(ss);
   1005   1.61       dsl 
   1006   1.71        pk 	if (is_pgrp_leader)
   1007   1.61       dsl 		return;
   1008   1.61       dsl 
   1009   1.61       dsl 	pg_free(pg_id);
   1010   1.61       dsl }
   1011   1.61       dsl 
   1012   1.61       dsl /*
   1013   1.61       dsl  * Delete session - called from SESSRELE when s_count becomes zero.
   1014  1.107        ad  * Must be called with the proclist_lock held.
   1015   1.61       dsl  */
   1016   1.11       cgd void
   1017   1.61       dsl sessdelete(struct session *ss)
   1018    1.1       cgd {
   1019  1.100        ad 
   1020  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
   1021  1.100        ad 
   1022   1.61       dsl 	/*
   1023   1.61       dsl 	 * We keep the pgrp with the same id as the session in
   1024   1.61       dsl 	 * order to stop a process being given the same pid.
   1025   1.61       dsl 	 * Since the pgrp holds a reference to the session, it
   1026   1.61       dsl 	 * must be a 'zombie' pgrp by now.
   1027   1.61       dsl 	 */
   1028   1.61       dsl 	pg_free(ss->s_sid);
   1029   1.77    simonb 	pool_put(&session_pool, ss);
   1030    1.1       cgd }
   1031    1.1       cgd 
   1032    1.1       cgd /*
   1033    1.1       cgd  * Adjust pgrp jobc counters when specified process changes process group.
   1034    1.1       cgd  * We count the number of processes in each process group that "qualify"
   1035    1.1       cgd  * the group for terminal job control (those with a parent in a different
   1036    1.1       cgd  * process group of the same session).  If that count reaches zero, the
   1037    1.1       cgd  * process group becomes orphaned.  Check both the specified process'
   1038    1.1       cgd  * process group and that of its children.
   1039    1.1       cgd  * entering == 0 => p is leaving specified group.
   1040    1.1       cgd  * entering == 1 => p is entering specified group.
   1041   1.68       dsl  *
   1042  1.107        ad  * Call with proclist_lock held.
   1043    1.1       cgd  */
   1044    1.4    andrew void
   1045   1.59       dsl fixjobc(struct proc *p, struct pgrp *pgrp, int entering)
   1046    1.1       cgd {
   1047   1.39  augustss 	struct pgrp *hispgrp;
   1048   1.39  augustss 	struct session *mysession = pgrp->pg_session;
   1049   1.68       dsl 	struct proc *child;
   1050    1.1       cgd 
   1051  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
   1052  1.107        ad 	KASSERT(mutex_owned(&proclist_mutex));
   1053  1.100        ad 
   1054    1.1       cgd 	/*
   1055    1.1       cgd 	 * Check p's parent to see whether p qualifies its own process
   1056    1.1       cgd 	 * group; if so, adjust count for p's process group.
   1057    1.1       cgd 	 */
   1058   1.68       dsl 	hispgrp = p->p_pptr->p_pgrp;
   1059   1.68       dsl 	if (hispgrp != pgrp && hispgrp->pg_session == mysession) {
   1060  1.100        ad 		if (entering) {
   1061  1.100        ad 			mutex_enter(&p->p_smutex);
   1062  1.100        ad 			p->p_sflag &= ~PS_ORPHANPG;
   1063  1.100        ad 			mutex_exit(&p->p_smutex);
   1064    1.1       cgd 			pgrp->pg_jobc++;
   1065  1.100        ad 		} else if (--pgrp->pg_jobc == 0)
   1066    1.1       cgd 			orphanpg(pgrp);
   1067   1.26   thorpej 	}
   1068    1.1       cgd 
   1069    1.1       cgd 	/*
   1070    1.1       cgd 	 * Check this process' children to see whether they qualify
   1071    1.1       cgd 	 * their process groups; if so, adjust counts for children's
   1072    1.1       cgd 	 * process groups.
   1073    1.1       cgd 	 */
   1074   1.68       dsl 	LIST_FOREACH(child, &p->p_children, p_sibling) {
   1075   1.68       dsl 		hispgrp = child->p_pgrp;
   1076   1.68       dsl 		if (hispgrp != pgrp && hispgrp->pg_session == mysession &&
   1077   1.68       dsl 		    !P_ZOMBIE(child)) {
   1078  1.100        ad 			if (entering) {
   1079  1.100        ad 				mutex_enter(&child->p_smutex);
   1080  1.100        ad 				child->p_sflag &= ~PS_ORPHANPG;
   1081  1.100        ad 				mutex_exit(&child->p_smutex);
   1082    1.1       cgd 				hispgrp->pg_jobc++;
   1083  1.100        ad 			} else if (--hispgrp->pg_jobc == 0)
   1084    1.1       cgd 				orphanpg(hispgrp);
   1085   1.26   thorpej 		}
   1086   1.26   thorpej 	}
   1087    1.1       cgd }
   1088    1.1       cgd 
   1089   1.72  junyoung /*
   1090    1.1       cgd  * A process group has become orphaned;
   1091    1.1       cgd  * if there are any stopped processes in the group,
   1092    1.1       cgd  * hang-up all process in that group.
   1093   1.68       dsl  *
   1094  1.107        ad  * Call with proclist_lock held.
   1095    1.1       cgd  */
   1096    1.4    andrew static void
   1097   1.59       dsl orphanpg(struct pgrp *pg)
   1098    1.1       cgd {
   1099   1.39  augustss 	struct proc *p;
   1100  1.100        ad 	int doit;
   1101  1.100        ad 
   1102  1.107        ad 	KASSERT(mutex_owned(&proclist_lock));
   1103  1.107        ad 	KASSERT(mutex_owned(&proclist_mutex));
   1104  1.100        ad 
   1105  1.100        ad 	doit = 0;
   1106    1.1       cgd 
   1107   1.52      matt 	LIST_FOREACH(p, &pg->pg_members, p_pglist) {
   1108  1.100        ad 		mutex_enter(&p->p_smutex);
   1109    1.1       cgd 		if (p->p_stat == SSTOP) {
   1110  1.100        ad 			doit = 1;
   1111  1.100        ad 			p->p_sflag |= PS_ORPHANPG;
   1112    1.1       cgd 		}
   1113  1.100        ad 		mutex_exit(&p->p_smutex);
   1114    1.1       cgd 	}
   1115   1.35    bouyer 
   1116  1.100        ad 	if (doit) {
   1117  1.100        ad 		LIST_FOREACH(p, &pg->pg_members, p_pglist) {
   1118  1.100        ad 			psignal(p, SIGHUP);
   1119  1.100        ad 			psignal(p, SIGCONT);
   1120   1.35    bouyer 		}
   1121   1.35    bouyer 	}
   1122   1.35    bouyer }
   1123    1.1       cgd 
   1124   1.61       dsl #ifdef DDB
   1125   1.61       dsl #include <ddb/db_output.h>
   1126   1.61       dsl void pidtbl_dump(void);
   1127   1.14  christos void
   1128   1.61       dsl pidtbl_dump(void)
   1129    1.1       cgd {
   1130   1.61       dsl 	struct pid_table *pt;
   1131   1.61       dsl 	struct proc *p;
   1132   1.39  augustss 	struct pgrp *pgrp;
   1133   1.61       dsl 	int id;
   1134    1.1       cgd 
   1135   1.61       dsl 	db_printf("pid table %p size %x, next %x, last %x\n",
   1136   1.61       dsl 		pid_table, pid_tbl_mask+1,
   1137   1.61       dsl 		next_free_pt, last_free_pt);
   1138   1.61       dsl 	for (pt = pid_table, id = 0; id <= pid_tbl_mask; id++, pt++) {
   1139   1.61       dsl 		p = pt->pt_proc;
   1140   1.61       dsl 		if (!P_VALID(p) && !pt->pt_pgrp)
   1141   1.61       dsl 			continue;
   1142   1.61       dsl 		db_printf("  id %x: ", id);
   1143   1.61       dsl 		if (P_VALID(p))
   1144   1.61       dsl 			db_printf("proc %p id %d (0x%x) %s\n",
   1145   1.61       dsl 				p, p->p_pid, p->p_pid, p->p_comm);
   1146   1.61       dsl 		else
   1147   1.61       dsl 			db_printf("next %x use %x\n",
   1148   1.61       dsl 				P_NEXT(p) & pid_tbl_mask,
   1149   1.61       dsl 				P_NEXT(p) & ~pid_tbl_mask);
   1150   1.61       dsl 		if ((pgrp = pt->pt_pgrp)) {
   1151   1.61       dsl 			db_printf("\tsession %p, sid %d, count %d, login %s\n",
   1152   1.61       dsl 			    pgrp->pg_session, pgrp->pg_session->s_sid,
   1153   1.61       dsl 			    pgrp->pg_session->s_count,
   1154   1.61       dsl 			    pgrp->pg_session->s_login);
   1155   1.61       dsl 			db_printf("\tpgrp %p, pg_id %d, pg_jobc %d, members %p\n",
   1156   1.61       dsl 			    pgrp, pgrp->pg_id, pgrp->pg_jobc,
   1157   1.61       dsl 			    pgrp->pg_members.lh_first);
   1158   1.61       dsl 			for (p = pgrp->pg_members.lh_first; p != 0;
   1159   1.61       dsl 			    p = p->p_pglist.le_next) {
   1160   1.72  junyoung 				db_printf("\t\tpid %d addr %p pgrp %p %s\n",
   1161   1.61       dsl 				    p->p_pid, p, p->p_pgrp, p->p_comm);
   1162   1.10   mycroft 			}
   1163    1.1       cgd 		}
   1164    1.1       cgd 	}
   1165    1.1       cgd }
   1166   1.61       dsl #endif /* DDB */
   1167   1.48      yamt 
   1168   1.48      yamt #ifdef KSTACK_CHECK_MAGIC
   1169   1.48      yamt #include <sys/user.h>
   1170   1.48      yamt 
   1171   1.48      yamt #define	KSTACK_MAGIC	0xdeadbeaf
   1172   1.48      yamt 
   1173   1.48      yamt /* XXX should be per process basis? */
   1174   1.48      yamt int kstackleftmin = KSTACK_SIZE;
   1175   1.50     enami int kstackleftthres = KSTACK_SIZE / 8; /* warn if remaining stack is
   1176   1.50     enami 					  less than this */
   1177   1.48      yamt 
   1178   1.48      yamt void
   1179   1.56      yamt kstack_setup_magic(const struct lwp *l)
   1180   1.48      yamt {
   1181   1.85     perry 	uint32_t *ip;
   1182   1.85     perry 	uint32_t const *end;
   1183   1.48      yamt 
   1184   1.56      yamt 	KASSERT(l != NULL);
   1185   1.56      yamt 	KASSERT(l != &lwp0);
   1186   1.48      yamt 
   1187   1.48      yamt 	/*
   1188   1.48      yamt 	 * fill all the stack with magic number
   1189   1.48      yamt 	 * so that later modification on it can be detected.
   1190   1.48      yamt 	 */
   1191   1.85     perry 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
   1192  1.114    dyoung 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
   1193   1.48      yamt 	for (; ip < end; ip++) {
   1194   1.48      yamt 		*ip = KSTACK_MAGIC;
   1195   1.48      yamt 	}
   1196   1.48      yamt }
   1197   1.48      yamt 
   1198   1.48      yamt void
   1199   1.56      yamt kstack_check_magic(const struct lwp *l)
   1200   1.48      yamt {
   1201   1.85     perry 	uint32_t const *ip, *end;
   1202   1.48      yamt 	int stackleft;
   1203   1.48      yamt 
   1204   1.56      yamt 	KASSERT(l != NULL);
   1205   1.48      yamt 
   1206   1.48      yamt 	/* don't check proc0 */ /*XXX*/
   1207   1.56      yamt 	if (l == &lwp0)
   1208   1.48      yamt 		return;
   1209   1.48      yamt 
   1210   1.48      yamt #ifdef __MACHINE_STACK_GROWS_UP
   1211   1.48      yamt 	/* stack grows upwards (eg. hppa) */
   1212  1.106  christos 	ip = (uint32_t *)((void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
   1213   1.85     perry 	end = (uint32_t *)KSTACK_LOWEST_ADDR(l);
   1214   1.48      yamt 	for (ip--; ip >= end; ip--)
   1215   1.48      yamt 		if (*ip != KSTACK_MAGIC)
   1216   1.48      yamt 			break;
   1217   1.72  junyoung 
   1218  1.106  christos 	stackleft = (void *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE - (void *)ip;
   1219   1.48      yamt #else /* __MACHINE_STACK_GROWS_UP */
   1220   1.48      yamt 	/* stack grows downwards (eg. i386) */
   1221   1.85     perry 	ip = (uint32_t *)KSTACK_LOWEST_ADDR(l);
   1222  1.114    dyoung 	end = (uint32_t *)((char *)KSTACK_LOWEST_ADDR(l) + KSTACK_SIZE);
   1223   1.48      yamt 	for (; ip < end; ip++)
   1224   1.48      yamt 		if (*ip != KSTACK_MAGIC)
   1225   1.48      yamt 			break;
   1226   1.48      yamt 
   1227   1.93  christos 	stackleft = ((const char *)ip) - (const char *)KSTACK_LOWEST_ADDR(l);
   1228   1.48      yamt #endif /* __MACHINE_STACK_GROWS_UP */
   1229   1.48      yamt 
   1230   1.48      yamt 	if (kstackleftmin > stackleft) {
   1231   1.48      yamt 		kstackleftmin = stackleft;
   1232   1.48      yamt 		if (stackleft < kstackleftthres)
   1233   1.56      yamt 			printf("warning: kernel stack left %d bytes"
   1234   1.56      yamt 			    "(pid %u:lid %u)\n", stackleft,
   1235   1.56      yamt 			    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
   1236   1.48      yamt 	}
   1237   1.48      yamt 
   1238   1.48      yamt 	if (stackleft <= 0) {
   1239   1.56      yamt 		panic("magic on the top of kernel stack changed for "
   1240   1.56      yamt 		    "pid %u, lid %u: maybe kernel stack overflow",
   1241   1.56      yamt 		    (u_int)l->l_proc->p_pid, (u_int)l->l_lid);
   1242   1.48      yamt 	}
   1243   1.48      yamt }
   1244   1.50     enami #endif /* KSTACK_CHECK_MAGIC */
   1245   1.79      yamt 
   1246  1.100        ad /*
   1247  1.100        ad  * XXXSMP this is bust, it grabs a read lock and then messes about
   1248  1.100        ad  * with allproc.
   1249  1.100        ad  */
   1250   1.79      yamt int
   1251   1.79      yamt proclist_foreach_call(struct proclist *list,
   1252   1.79      yamt     int (*callback)(struct proc *, void *arg), void *arg)
   1253   1.79      yamt {
   1254   1.79      yamt 	struct proc marker;
   1255   1.79      yamt 	struct proc *p;
   1256   1.79      yamt 	struct lwp * const l = curlwp;
   1257   1.79      yamt 	int ret = 0;
   1258   1.79      yamt 
   1259  1.102     pavel 	marker.p_flag = PK_MARKER;
   1260  1.113        ad 	uvm_lwp_hold(l);
   1261  1.107        ad 	mutex_enter(&proclist_lock);
   1262   1.79      yamt 	for (p = LIST_FIRST(list); ret == 0 && p != NULL;) {
   1263  1.102     pavel 		if (p->p_flag & PK_MARKER) {
   1264   1.79      yamt 			p = LIST_NEXT(p, p_list);
   1265   1.79      yamt 			continue;
   1266   1.79      yamt 		}
   1267   1.79      yamt 		LIST_INSERT_AFTER(p, &marker, p_list);
   1268   1.79      yamt 		ret = (*callback)(p, arg);
   1269  1.107        ad 		KASSERT(mutex_owned(&proclist_lock));
   1270   1.79      yamt 		p = LIST_NEXT(&marker, p_list);
   1271   1.79      yamt 		LIST_REMOVE(&marker, p_list);
   1272   1.79      yamt 	}
   1273  1.107        ad 	mutex_exit(&proclist_lock);
   1274  1.113        ad 	uvm_lwp_rele(l);
   1275   1.79      yamt 
   1276   1.79      yamt 	return ret;
   1277   1.79      yamt }
   1278   1.86      yamt 
   1279   1.86      yamt int
   1280   1.86      yamt proc_vmspace_getref(struct proc *p, struct vmspace **vm)
   1281   1.86      yamt {
   1282   1.86      yamt 
   1283   1.86      yamt 	/* XXXCDC: how should locking work here? */
   1284   1.86      yamt 
   1285   1.87      yamt 	/* curproc exception is for coredump. */
   1286   1.87      yamt 
   1287  1.100        ad 	if ((p != curproc && (p->p_sflag & PS_WEXIT) != 0) ||
   1288   1.86      yamt 	    (p->p_vmspace->vm_refcnt < 1)) { /* XXX */
   1289   1.86      yamt 		return EFAULT;
   1290   1.86      yamt 	}
   1291   1.86      yamt 
   1292   1.86      yamt 	uvmspace_addref(p->p_vmspace);
   1293   1.86      yamt 	*vm = p->p_vmspace;
   1294   1.86      yamt 
   1295   1.86      yamt 	return 0;
   1296   1.86      yamt }
   1297   1.94        ad 
   1298   1.94        ad /*
   1299   1.94        ad  * Acquire a write lock on the process credential.
   1300   1.94        ad  */
   1301   1.94        ad void
   1302  1.100        ad proc_crmod_enter(void)
   1303   1.94        ad {
   1304  1.100        ad 	struct lwp *l = curlwp;
   1305  1.100        ad 	struct proc *p = l->l_proc;
   1306  1.100        ad 	struct plimit *lim;
   1307  1.100        ad 	kauth_cred_t oc;
   1308  1.100        ad 	char *cn;
   1309   1.94        ad 
   1310  1.117       dsl 	/* Reset what needs to be reset in plimit. */
   1311  1.117       dsl 	if (p->p_limit->pl_corename != defcorename) {
   1312  1.117       dsl 		lim_privatise(p, false);
   1313  1.117       dsl 		lim = p->p_limit;
   1314  1.117       dsl 		mutex_enter(&lim->pl_lock);
   1315  1.117       dsl 		cn = lim->pl_corename;
   1316  1.117       dsl 		lim->pl_corename = defcorename;
   1317  1.117       dsl 		mutex_exit(&lim->pl_lock);
   1318  1.117       dsl 		if (cn != defcorename)
   1319  1.117       dsl 			free(cn, M_TEMP);
   1320  1.117       dsl 	}
   1321  1.117       dsl 
   1322  1.100        ad 	mutex_enter(&p->p_mutex);
   1323  1.100        ad 
   1324  1.100        ad 	/* Ensure the LWP cached credentials are up to date. */
   1325  1.100        ad 	if ((oc = l->l_cred) != p->p_cred) {
   1326  1.100        ad 		kauth_cred_hold(p->p_cred);
   1327  1.100        ad 		l->l_cred = p->p_cred;
   1328  1.100        ad 		kauth_cred_free(oc);
   1329  1.100        ad 	}
   1330  1.100        ad 
   1331   1.94        ad }
   1332   1.94        ad 
   1333   1.94        ad /*
   1334  1.100        ad  * Set in a new process credential, and drop the write lock.  The credential
   1335  1.100        ad  * must have a reference already.  Optionally, free a no-longer required
   1336  1.100        ad  * credential.  The scheduler also needs to inspect p_cred, so we also
   1337  1.100        ad  * briefly acquire the sched state mutex.
   1338   1.94        ad  */
   1339   1.94        ad void
   1340  1.104   thorpej proc_crmod_leave(kauth_cred_t scred, kauth_cred_t fcred, bool sugid)
   1341   1.94        ad {
   1342  1.100        ad 	struct lwp *l = curlwp;
   1343  1.100        ad 	struct proc *p = l->l_proc;
   1344  1.100        ad 	kauth_cred_t oc;
   1345  1.100        ad 
   1346  1.100        ad 	/* Is there a new credential to set in? */
   1347  1.100        ad 	if (scred != NULL) {
   1348  1.100        ad 		mutex_enter(&p->p_smutex);
   1349  1.100        ad 		p->p_cred = scred;
   1350  1.100        ad 		mutex_exit(&p->p_smutex);
   1351  1.100        ad 
   1352  1.100        ad 		/* Ensure the LWP cached credentials are up to date. */
   1353  1.100        ad 		if ((oc = l->l_cred) != scred) {
   1354  1.100        ad 			kauth_cred_hold(scred);
   1355  1.100        ad 			l->l_cred = scred;
   1356  1.100        ad 		}
   1357  1.100        ad 	} else
   1358  1.100        ad 		oc = NULL;	/* XXXgcc */
   1359  1.100        ad 
   1360  1.100        ad 	if (sugid) {
   1361  1.100        ad 		/*
   1362  1.100        ad 		 * Mark process as having changed credentials, stops
   1363  1.100        ad 		 * tracing etc.
   1364  1.100        ad 		 */
   1365  1.102     pavel 		p->p_flag |= PK_SUGID;
   1366  1.100        ad 	}
   1367   1.94        ad 
   1368  1.100        ad 	mutex_exit(&p->p_mutex);
   1369  1.100        ad 
   1370  1.100        ad 	/* If there is a credential to be released, free it now. */
   1371  1.100        ad 	if (fcred != NULL) {
   1372  1.100        ad 		KASSERT(scred != NULL);
   1373   1.94        ad 		kauth_cred_free(fcred);
   1374  1.100        ad 		if (oc != scred)
   1375  1.100        ad 			kauth_cred_free(oc);
   1376  1.100        ad 	}
   1377  1.100        ad }
   1378  1.100        ad 
   1379  1.100        ad /*
   1380  1.100        ad  * Acquire a reference on a process, to prevent it from exiting or execing.
   1381  1.100        ad  */
   1382  1.100        ad int
   1383  1.100        ad proc_addref(struct proc *p)
   1384  1.100        ad {
   1385  1.100        ad 
   1386  1.107        ad 	KASSERT(mutex_owned(&p->p_mutex));
   1387  1.100        ad 
   1388  1.100        ad 	if (p->p_refcnt <= 0)
   1389  1.100        ad 		return EAGAIN;
   1390  1.100        ad 	p->p_refcnt++;
   1391  1.100        ad 
   1392  1.100        ad 	return 0;
   1393  1.100        ad }
   1394  1.100        ad 
   1395  1.100        ad /*
   1396  1.100        ad  * Release a reference on a process.
   1397  1.100        ad  */
   1398  1.100        ad void
   1399  1.100        ad proc_delref(struct proc *p)
   1400  1.100        ad {
   1401  1.100        ad 
   1402  1.107        ad 	KASSERT(mutex_owned(&p->p_mutex));
   1403  1.100        ad 
   1404  1.100        ad 	if (p->p_refcnt < 0) {
   1405  1.100        ad 		if (++p->p_refcnt == 0)
   1406  1.100        ad 			cv_broadcast(&p->p_refcv);
   1407  1.100        ad 	} else {
   1408  1.100        ad 		p->p_refcnt--;
   1409  1.100        ad 		KASSERT(p->p_refcnt != 0);
   1410  1.100        ad 	}
   1411  1.100        ad }
   1412  1.100        ad 
   1413  1.100        ad /*
   1414  1.100        ad  * Wait for all references on the process to drain, and prevent new
   1415  1.100        ad  * references from being acquired.
   1416  1.100        ad  */
   1417  1.100        ad void
   1418  1.100        ad proc_drainrefs(struct proc *p)
   1419  1.100        ad {
   1420  1.100        ad 
   1421  1.107        ad 	KASSERT(mutex_owned(&p->p_mutex));
   1422  1.111      yamt 	KASSERT(p->p_refcnt >= 0);
   1423  1.100        ad 
   1424  1.100        ad 	/*
   1425  1.100        ad 	 * The process itself holds the last reference.  Once it's released,
   1426  1.100        ad 	 * no new references will be granted.  If we have already locked out
   1427  1.100        ad 	 * new references (refcnt <= 0), potentially due to a failed exec,
   1428  1.100        ad 	 * there is nothing more to do.
   1429  1.100        ad 	 */
   1430  1.111      yamt 	if (p->p_refcnt == 0)
   1431  1.111      yamt 		return;
   1432  1.100        ad 	p->p_refcnt = 1 - p->p_refcnt;
   1433  1.100        ad 	while (p->p_refcnt != 0)
   1434  1.100        ad 		cv_wait(&p->p_refcv, &p->p_mutex);
   1435   1.94        ad }
   1436   1.95   thorpej 
   1437   1.95   thorpej /*
   1438   1.95   thorpej  * proc_specific_key_create --
   1439   1.95   thorpej  *	Create a key for subsystem proc-specific data.
   1440   1.95   thorpej  */
   1441   1.95   thorpej int
   1442   1.95   thorpej proc_specific_key_create(specificdata_key_t *keyp, specificdata_dtor_t dtor)
   1443   1.95   thorpej {
   1444   1.95   thorpej 
   1445   1.98   thorpej 	return (specificdata_key_create(proc_specificdata_domain, keyp, dtor));
   1446   1.95   thorpej }
   1447   1.95   thorpej 
   1448   1.95   thorpej /*
   1449   1.95   thorpej  * proc_specific_key_delete --
   1450   1.95   thorpej  *	Delete a key for subsystem proc-specific data.
   1451   1.95   thorpej  */
   1452   1.95   thorpej void
   1453   1.95   thorpej proc_specific_key_delete(specificdata_key_t key)
   1454   1.95   thorpej {
   1455   1.95   thorpej 
   1456   1.95   thorpej 	specificdata_key_delete(proc_specificdata_domain, key);
   1457   1.95   thorpej }
   1458   1.95   thorpej 
   1459   1.98   thorpej /*
   1460   1.98   thorpej  * proc_initspecific --
   1461   1.98   thorpej  *	Initialize a proc's specificdata container.
   1462   1.98   thorpej  */
   1463   1.96  christos void
   1464   1.96  christos proc_initspecific(struct proc *p)
   1465   1.96  christos {
   1466   1.96  christos 	int error;
   1467   1.98   thorpej 
   1468   1.96  christos 	error = specificdata_init(proc_specificdata_domain, &p->p_specdataref);
   1469   1.96  christos 	KASSERT(error == 0);
   1470   1.96  christos }
   1471   1.96  christos 
   1472   1.95   thorpej /*
   1473   1.98   thorpej  * proc_finispecific --
   1474   1.98   thorpej  *	Finalize a proc's specificdata container.
   1475   1.98   thorpej  */
   1476   1.98   thorpej void
   1477   1.98   thorpej proc_finispecific(struct proc *p)
   1478   1.98   thorpej {
   1479   1.98   thorpej 
   1480   1.98   thorpej 	specificdata_fini(proc_specificdata_domain, &p->p_specdataref);
   1481   1.98   thorpej }
   1482   1.98   thorpej 
   1483   1.98   thorpej /*
   1484   1.95   thorpej  * proc_getspecific --
   1485   1.95   thorpej  *	Return proc-specific data corresponding to the specified key.
   1486   1.95   thorpej  */
   1487   1.95   thorpej void *
   1488   1.95   thorpej proc_getspecific(struct proc *p, specificdata_key_t key)
   1489   1.95   thorpej {
   1490   1.95   thorpej 
   1491   1.95   thorpej 	return (specificdata_getspecific(proc_specificdata_domain,
   1492   1.95   thorpej 					 &p->p_specdataref, key));
   1493   1.95   thorpej }
   1494   1.95   thorpej 
   1495   1.95   thorpej /*
   1496   1.95   thorpej  * proc_setspecific --
   1497   1.95   thorpej  *	Set proc-specific data corresponding to the specified key.
   1498   1.95   thorpej  */
   1499   1.95   thorpej void
   1500   1.95   thorpej proc_setspecific(struct proc *p, specificdata_key_t key, void *data)
   1501   1.95   thorpej {
   1502   1.95   thorpej 
   1503   1.95   thorpej 	specificdata_setspecific(proc_specificdata_domain,
   1504   1.95   thorpej 				 &p->p_specdataref, key, data);
   1505   1.95   thorpej }
   1506