xnu-517.9.5.tar.gz

[apple/xnu.git] / bsd / kern / kern_fork.c

diff --git a/bsd/kern/kern_fork.c b/bsd/kern/kern_fork.c
index 58aa6ecf4f4a0b15083b924b0e9ebb53571c0947..f8b8cf3d8c283a545e24b9f2d6814a1b4ddb5255 100644 (file)
--- a/bsd/kern/kern_fork.c
+++ b/bsd/kern/kern_fork.c
@@ -1,5 +1,5 @@
 /*
 /*

- * Copyright (c) 2000 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.

  *
  * @APPLE_LICENSE_HEADER_START@
  * 
  *
  * @APPLE_LICENSE_HEADER_START@
  * 


@@ -60,6 +60,7 @@
  *     @(#)kern_fork.c 8.8 (Berkeley) 2/14/95
  */
 
  *     @(#)kern_fork.c 8.8 (Berkeley) 2/14/95
  */
 

+#include <kern/assert.h>

 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/filedesc.h>
 #include <sys/param.h>
 #include <sys/systm.h>
 #include <sys/filedesc.h>


@@ -71,16 +72,23 @@
 #include <sys/vnode.h>
 #include <sys/file.h>
 #include <sys/acct.h>
 #include <sys/vnode.h>
 #include <sys/file.h>
 #include <sys/acct.h>

+#include <sys/wait.h>
+
+#include <bsm/audit_kernel.h>
+
+#if KTRACE

 #include <sys/ktrace.h>
 #include <sys/ktrace.h>

+#include <sys/ubc.h>
+#endif

 
 #include <mach/mach_types.h>
 #include <kern/mach_param.h>
 
 #include <machine/spl.h>
 
 
 #include <mach/mach_types.h>
 #include <kern/mach_param.h>
 
 #include <machine/spl.h>
 

-thread_t cloneproc(struct proc *, int); 
+thread_act_t cloneproc(struct proc *, int); 

 struct proc * forkproc(struct proc *, int);
 struct proc * forkproc(struct proc *, int);

-thread_t procdup();
+thread_act_t procdup();

 
 #define        DOFORK  0x1     /* fork() system call */
 #define        DOVFORK 0x2     /* vfork() system call */
 
 #define        DOFORK  0x1     /* fork() system call */
 #define        DOVFORK 0x2     /* vfork() system call */


@@ -141,6 +149,7 @@ vfork(p, uap, retval)
        ut = (struct uthread *)get_bsdthread_info(cur_act);
        if (ut->uu_flag & P_VFORK) {
                printf("vfork called recursively by %s\n", p->p_comm);
        ut = (struct uthread *)get_bsdthread_info(cur_act);
        if (ut->uu_flag & P_VFORK) {
                printf("vfork called recursively by %s\n", p->p_comm);

+               (void)chgproccnt(uid, -1);

                return (EINVAL);
        }
        p->p_flag  |= P_VFORK;
                return (EINVAL);
        }
        p->p_flag  |= P_VFORK;


@@ -149,6 +158,8 @@ vfork(p, uap, retval)
        /* The newly created process comes with signal lock held */
        newproc = (struct proc *)forkproc(p,1);
 
        /* The newly created process comes with signal lock held */
        newproc = (struct proc *)forkproc(p,1);
 

+       AUDIT_ARG(pid, newproc->p_pid);
+

        LIST_INSERT_AFTER(p, newproc, p_pglist);
        newproc->p_pptr = p;
        newproc->task = p->task;
        LIST_INSERT_AFTER(p, newproc, p_pglist);
        newproc->p_pptr = p;
        newproc->task = p->task;


@@ -164,6 +175,7 @@ vfork(p, uap, retval)
        ut->uu_flag |= P_VFORK;
        ut->uu_proc = newproc;
        ut->uu_userstate = (void *)act_thread_csave();
        ut->uu_flag |= P_VFORK;
        ut->uu_proc = newproc;
        ut->uu_userstate = (void *)act_thread_csave();

+       ut->uu_vforkmask = ut->uu_sigmask;

 
        thread_set_child(cur_act, newproc->p_pid);
 
 
        thread_set_child(cur_act, newproc->p_pid);
 


@@ -198,29 +210,30 @@ vfork_return(th_act, p, p2, retval)
 {
        long flags;
        register uid_t uid;
 {
        long flags;
        register uid_t uid;

-       thread_t newth, self = current_thread();
-       thread_act_t cur_act = (thread_act_t)current_act();

        int s, count;
        task_t t;
        uthread_t ut;
        
        int s, count;
        task_t t;
        uthread_t ut;
        

-       ut = (struct uthread *)get_bsdthread_info(cur_act);
+       ut = (struct uthread *)get_bsdthread_info(th_act);

 
        act_thread_catt(ut->uu_userstate);
 
        /* Make sure only one at this time */
 
        act_thread_catt(ut->uu_userstate);
 
        /* Make sure only one at this time */

-       p->p_vforkcnt--;
-       if (p->p_vforkcnt <0)
-               panic("vfork cnt is -ve");
-       if (p->p_vforkcnt <=0)
-               p->p_flag  &= ~P_VFORK;
+       if (p) {
+               p->p_vforkcnt--;
+               if (p->p_vforkcnt <0)
+                       panic("vfork cnt is -ve");
+               if (p->p_vforkcnt <=0)
+                       p->p_flag  &= ~P_VFORK;
+       }

        ut->uu_userstate = 0;
        ut->uu_flag &= ~P_VFORK;
        ut->uu_proc = 0;
        ut->uu_userstate = 0;
        ut->uu_flag &= ~P_VFORK;
        ut->uu_proc = 0;

+       ut->uu_sigmask = ut->uu_vforkmask;

        p2->p_flag  &= ~P_INVFORK;
        p2->p_vforkact = (void *)0;
 
        p2->p_flag  &= ~P_INVFORK;
        p2->p_vforkact = (void *)0;
 

-       thread_set_parent(cur_act, p2->p_pid);
+       thread_set_parent(th_act, p2->p_pid);

 
        if (retval) {
                retval[0] = p2->p_pid;
 
        if (retval) {
                retval[0] = p2->p_pid;


@@ -230,20 +243,21 @@ vfork_return(th_act, p, p2, retval)
        return;
 }
 
        return;
 }
 

-thread_t
+thread_act_t

 procdup(
        struct proc             *child,
        struct proc             *parent)
 {
 procdup(
        struct proc             *child,
        struct proc             *parent)
 {

-       thread_t                thread;
+       thread_act_t            thread;

        task_t                  task;
        kern_return_t   result;
        task_t                  task;
        kern_return_t   result;

+       pmap_t                  pmap;

        extern task_t kernel_task;
 
        if (parent->task == kernel_task)
        extern task_t kernel_task;
 
        if (parent->task == kernel_task)

-               result = task_create_local(TASK_NULL, FALSE, FALSE, &task);
+               result = task_create_internal(TASK_NULL, FALSE, &task);

        else
        else

-               result = task_create_local(parent->task, TRUE, FALSE, &task);
+               result = task_create_internal(parent->task, TRUE, &task);

        if (result != KERN_SUCCESS)
            printf("fork/procdup: task_create failed. Code: 0x%x\n", result);
        child->task = task;
        if (result != KERN_SUCCESS)
            printf("fork/procdup: task_create failed. Code: 0x%x\n", result);
        child->task = task;


@@ -251,6 +265,7 @@ procdup(
        set_bsdtask_info(task, child);
        if (child->p_nice != 0)
                resetpriority(child);
        set_bsdtask_info(task, child);
        if (child->p_nice != 0)
                resetpriority(child);

+               

        result = thread_create(task, &thread);
        if (result != KERN_SUCCESS)
            printf("fork/procdup: thread_create failed. Code: 0x%x\n", result);
        result = thread_create(task, &thread);
        if (result != KERN_SUCCESS)
            printf("fork/procdup: thread_create failed. Code: 0x%x\n", result);


@@ -267,7 +282,7 @@ fork1(p1, flags, retval)
 {
        register struct proc *p2;
        register uid_t uid;
 {
        register struct proc *p2;
        register uid_t uid;

-       thread_t newth, self = current_thread();
+       thread_act_t newth;

        int s, count;
         task_t t;
 
        int s, count;
         task_t t;
 


@@ -297,10 +312,12 @@ fork1(p1, flags, retval)
 
        /* The newly created process comes with signal lock held */
        newth = cloneproc(p1, 1);
 
        /* The newly created process comes with signal lock held */
        newth = cloneproc(p1, 1);

-       thread_dup(current_act(), newth);
+       thread_dup(newth);

        /* p2 = newth->task->proc; */
        p2 = (struct proc *)(get_bsdtask_info(get_threadtask(newth)));
 
        /* p2 = newth->task->proc; */
        p2 = (struct proc *)(get_bsdtask_info(get_threadtask(newth)));
 

+       AUDIT_ARG(pid, p2->p_pid);
+

        thread_set_child(newth, p2->p_pid);
 
        s = splhigh();
        thread_set_child(newth, p2->p_pid);
 
        s = splhigh();


@@ -326,6 +343,8 @@ fork1(p1, flags, retval)
         }
         act_deallocate(newth);
 
         }
         act_deallocate(newth);
 

+       KNOTE(&p1->p_klist, NOTE_FORK | p2->p_pid);
+

        while (p2->p_flag & P_PPWAIT)
                tsleep(p1, PWAIT, "ppwait", 0);
 
        while (p2->p_flag & P_PPWAIT)
                tsleep(p1, PWAIT, "ppwait", 0);
 


@@ -344,15 +363,17 @@ fork1(p1, flags, retval)
  * lock set. fork() code needs to explicity remove this lock 
  * before signals can be delivered
  */
  * lock set. fork() code needs to explicity remove this lock 
  * before signals can be delivered
  */

-thread_t
+thread_act_t

 cloneproc(p1, lock)
        register struct proc *p1;
        register int lock;
 {
        register struct proc *p2;
 cloneproc(p1, lock)
        register struct proc *p1;
        register int lock;
 {
        register struct proc *p2;

-       thread_t th;
+       thread_act_t th;

 
        p2 = (struct proc *)forkproc(p1,lock);
 
        p2 = (struct proc *)forkproc(p1,lock);

+
+

        th = procdup(p2, p1);   /* child, parent */
 
        LIST_INSERT_AFTER(p1, p2, p_pglist);
        th = procdup(p2, p1);   /* child, parent */
 
        LIST_INSERT_AFTER(p1, p2, p_pglist);


@@ -417,7 +438,8 @@ retry:
 again:
                for (; p2 != 0; p2 = p2->p_list.le_next) {
                        while (p2->p_pid == nextpid ||
 again:
                for (; p2 != 0; p2 = p2->p_list.le_next) {
                        while (p2->p_pid == nextpid ||

-                           p2->p_pgrp->pg_id == nextpid) {
+                           p2->p_pgrp->pg_id == nextpid ||
+                               p2->p_session->s_sid == nextpid) {

                                nextpid++;
                                if (nextpid >= pidchecked)
                                        goto retry;
                                nextpid++;
                                if (nextpid >= pidchecked)
                                        goto retry;


@@ -427,6 +449,9 @@ again:
                        if (p2->p_pgrp && p2->p_pgrp->pg_id > nextpid && 
                            pidchecked > p2->p_pgrp->pg_id)
                                pidchecked = p2->p_pgrp->pg_id;
                        if (p2->p_pgrp && p2->p_pgrp->pg_id > nextpid && 
                            pidchecked > p2->p_pgrp->pg_id)
                                pidchecked = p2->p_pgrp->pg_id;

+                       if (p2->p_session->s_sid > nextpid &&
+                               pidchecked > p2->p_session->s_sid)
+                               pidchecked = p2->p_session->s_sid;

                }
                if (!doingzomb) {
                        doingzomb = 1;
                }
                if (!doingzomb) {
                        doingzomb = 1;


@@ -440,6 +465,7 @@ again:
        p2->p_stat = SIDL;
        p2->p_pid = nextpid;
 
        p2->p_stat = SIDL;
        p2->p_pid = nextpid;
 

+       p2->p_shutdownstate = 0;

        /*
         * Make a proc table entry for the new process.
         * Start by zeroing the section of proc that is zero-initialized,
        /*
         * Make a proc table entry for the new process.
         * Start by zeroing the section of proc that is zero-initialized,


@@ -451,20 +477,28 @@ again:
            (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
        p2->vm_shm = (void *)NULL; /* Make sure it is zero */
 
            (unsigned) ((caddr_t)&p2->p_endcopy - (caddr_t)&p2->p_startcopy));
        p2->vm_shm = (void *)NULL; /* Make sure it is zero */
 

+       /*
+        * Copy the audit info.
+        */
+       audit_proc_fork(p1, p2);
+

        /*
         * Duplicate sub-structures as needed.
         * Increase reference counts on shared objects.
         * The p_stats and p_sigacts substructs are set in vm_fork.
         */
        p2->p_flag = P_INMEM;
        /*
         * Duplicate sub-structures as needed.
         * Increase reference counts on shared objects.
         * The p_stats and p_sigacts substructs are set in vm_fork.
         */
        p2->p_flag = P_INMEM;

+       p2->p_flag |= (p1->p_flag & P_CLASSIC); // copy from parent
+       p2->p_flag |= (p1->p_flag & P_AFFINITY); // copy from parent

        if (p1->p_flag & P_PROFIL)
                startprofclock(p2);
        bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred));
        p2->p_cred->p_refcnt = 1;
        crhold(p1->p_ucred);
        lockinit(&p2->p_cred->pc_lock, PLOCK, "proc cred", 0, 0);
        if (p1->p_flag & P_PROFIL)
                startprofclock(p2);
        bcopy(p1->p_cred, p2->p_cred, sizeof(*p2->p_cred));
        p2->p_cred->p_refcnt = 1;
        crhold(p1->p_ucred);
        lockinit(&p2->p_cred->pc_lock, PLOCK, "proc cred", 0, 0);

+       klist_init(&p2->p_klist);

 
 

-       /* bump references to the text vnode (for procfs) */
+       /* bump references to the text vnode */

        p2->p_textvp = p1->p_textvp;
        if (p2->p_textvp)
                VREF(p2->p_textvp);
        p2->p_textvp = p1->p_textvp;
        if (p2->p_textvp)
                VREF(p2->p_textvp);


@@ -502,6 +536,8 @@ again:
        if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
                p2->p_flag |= P_CONTROLT;
 
        if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
                p2->p_flag |= P_CONTROLT;
 

+       p2->p_argslen = p1->p_argslen;
+       p2->p_argc = p1->p_argc;

        p2->p_xstat = 0;
        p2->p_ru = NULL;
 
        p2->p_xstat = 0;
        p2->p_ru = NULL;
 


@@ -514,9 +550,13 @@ again:
        p2->sigwait_thread = NULL;
        p2->exit_thread = NULL;
        p2->user_stack = p1->user_stack;
        p2->sigwait_thread = NULL;
        p2->exit_thread = NULL;
        p2->user_stack = p1->user_stack;

-       p2->p_sigpending = 0;

        p2->p_vforkcnt = 0;
        p2->p_vforkact = 0;
        p2->p_vforkcnt = 0;
        p2->p_vforkact = 0;

+       TAILQ_INIT(&p2->p_uthlist);
+       TAILQ_INIT(&p2->aio_activeq);
+       TAILQ_INIT(&p2->aio_doneq);
+       p2->aio_active_count = 0;
+       p2->aio_done_count = 0;

 
 #if KTRACE
        /*
 
 #if KTRACE
        /*


@@ -525,8 +565,11 @@ again:
         */
        if (p1->p_traceflag&KTRFAC_INHERIT) {
                p2->p_traceflag = p1->p_traceflag;
         */
        if (p1->p_traceflag&KTRFAC_INHERIT) {
                p2->p_traceflag = p1->p_traceflag;

-               if ((p2->p_tracep = p1->p_tracep) != NULL)
+               if ((p2->p_tracep = p1->p_tracep) != NULL) {
+                       if (UBCINFOEXISTS(p2->p_tracep))
+                               ubc_hold(p2->p_tracep);

                        VREF(p2->p_tracep);
                        VREF(p2->p_tracep);

+               }

        }
 #endif
        return(p2);
        }
 #endif
        return(p2);


@@ -551,25 +594,63 @@ uthread_zone_init()
 }
 
 void *
 }
 
 void *

-uthread_alloc(void)
+uthread_alloc(task_t task, thread_act_t thr_act )

 {
 {

+       struct proc *p;
+       struct uthread *uth, *uth_parent;

        void *ut;
        void *ut;

+       extern task_t kernel_task;
+       boolean_t funnel_state;

 
        if (!uthread_zone_inited)
                uthread_zone_init();
 
        ut = (void *)zalloc(uthread_zone);
        bzero(ut, sizeof(struct uthread));
 
        if (!uthread_zone_inited)
                uthread_zone_init();
 
        ut = (void *)zalloc(uthread_zone);
        bzero(ut, sizeof(struct uthread));

+
+       if (task != kernel_task) {
+               uth = (struct uthread *)ut;
+               p = (struct proc *) get_bsdtask_info(task);
+
+               funnel_state = thread_funnel_set(kernel_flock, TRUE);
+               uth_parent = (struct uthread *)get_bsdthread_info(current_act());
+               if (uth_parent) {
+                       if (uth_parent->uu_flag & USAS_OLDMASK)
+                               uth->uu_sigmask = uth_parent->uu_oldmask;
+                       else
+                               uth->uu_sigmask = uth_parent->uu_sigmask;
+               }
+               uth->uu_act = thr_act;
+               //signal_lock(p);
+               if (p)
+                       TAILQ_INSERT_TAIL(&p->p_uthlist, uth, uu_list);
+               //signal_unlock(p);
+               (void)thread_funnel_set(kernel_flock, funnel_state);
+       }
+

        return (ut);
 }
 
 
 void
        return (ut);
 }
 
 
 void

-uthread_free(void *uthread)
+uthread_free(task_t task, thread_t act, void *uthread, void * bsd_info)

 {
        struct _select *sel;
        struct uthread *uth = (struct uthread *)uthread;
 {
        struct _select *sel;
        struct uthread *uth = (struct uthread *)uthread;

+       struct proc * p = (struct proc *)bsd_info;
+       extern task_t kernel_task;

        int size;
        int size;

+       boolean_t funnel_state;
+       struct nlminfo *nlmp;
+       struct proc * vproc;
+
+       /*
+        * Per-thread audit state should never last beyond system
+        * call return.  Since we don't audit the thread creation/
+        * removal, the thread state pointer should never be
+        * non-NULL when we get here.
+        */
+       assert(uth->uu_ar == NULL);

 
        sel = &uth->uu_state.ss_select;
        /* cleanup the select bit space */
 
        sel = &uth->uu_state.ss_select;
        /* cleanup the select bit space */


@@ -586,6 +667,25 @@ uthread_free(void *uthread)
                sel->wql = 0;
        }
 
                sel->wql = 0;
        }
 

+       if ((nlmp = uth->uu_nlminfo)) {
+               uth->uu_nlminfo = 0;
+               FREE(nlmp, M_LOCKF);
+       }
+
+       if ((task != kernel_task) ) {
+               int vfork_exit(struct proc *, int);
+
+               funnel_state = thread_funnel_set(kernel_flock, TRUE);
+               if (p)
+                       TAILQ_REMOVE(&p->p_uthlist, uth, uu_list);
+               if ((uth->uu_flag & P_VFORK) && (vproc = uth->uu_proc) 
+                                       && (vproc->p_flag & P_INVFORK)) {
+                       if (!vfork_exit(vproc, W_EXITCODE(0, SIGKILL))) 
+                               vfork_return(act, p, vproc, NULL);
+
+               }
+               (void)thread_funnel_set(kernel_flock, funnel_state);
+       }

        /* and free the uthread itself */
        zfree(uthread_zone, (vm_offset_t)uthread);
 }
        /* and free the uthread itself */
        zfree(uthread_zone, (vm_offset_t)uthread);
 }