]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/kern/bsd_init.c
xnu-792.10.96.tar.gz
[apple/xnu.git] / bsd / kern / bsd_init.c
index 2864a6cfd2b38512210a20eacb0fcf2225c209c6..05b66d5301077a46819f9b1ee061bc7ae3d7641c 100644 (file)
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2000-2003 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved.
  *
  * @APPLE_LICENSE_HEADER_START@
  * 
 #include <sys/param.h>
 #include <sys/filedesc.h>
 #include <sys/kernel.h>
-#include <sys/mount.h>
-#include <sys/proc.h>
+#include <sys/mount_internal.h>
+#include <sys/proc_internal.h>
+#include <sys/kauth.h>
 #include <sys/systm.h>
-#include <sys/vnode.h>
+#include <sys/vnode_internal.h>
 #include <sys/conf.h>
-#include <sys/buf.h>
+#include <sys/buf_internal.h>
 #include <sys/clist.h>
 #include <sys/user.h>
 #include <sys/time.h>
 #include <sys/systm.h>
+#include <sys/mman.h>
 
 #include <bsm/audit_kernel.h>
 
 #include <sys/malloc.h>
 #include <sys/dkstat.h>
 
-#include <machine/spl.h>
+#include <kern/startup.h>
 #include <kern/thread.h>
 #include <kern/task.h>
 #include <kern/ast.h>
 #include <mach/shared_memory_server.h>
 #include <vm/vm_shared_memory_server.h>
 
+#include <net/init.h>
+
+#include <machine/exec.h>
+
 extern int app_profile;                /* on/off switch for pre-heat cache */
 
 char    copyright[] =
@@ -126,7 +132,6 @@ extern void ux_handler();
 struct proc proc0;
 struct session session0;
 struct pgrp pgrp0;
-struct pcred cred0;
 struct filedesc filedesc0;
 struct plimit limit0;
 struct pstats pstats0;
@@ -138,6 +143,7 @@ long tk_nin;
 long tk_nout;
 long tk_rawcc;
 
+int lock_trace = 0;
 /* Global variables to make pstat happy. We do swapping differently */
 int nswdev, nswap;
 int nswapmap;
@@ -152,12 +158,15 @@ char      hostname[MAXHOSTNAMELEN];
 int            hostnamelen;
 char   domainname[MAXDOMNAMELEN];
 int            domainnamelen;
-char   classichandler[32] = {0};  
-long   classichandler_fsid = -1L;
-long   classichandler_fileid = -1L;
+#if __i386__
+struct exec_archhandler exec_archhandler_ppc = {
+       .path = "/usr/libexec/oah/translate",
+};
+#else /* __i386__ */
+struct exec_archhandler exec_archhandler_ppc;
+#endif /* __i386__ */
 
 char rootdevice[16];   /* hfs device names have at least 9 chars */
-struct timeval boottime;               /* GRODY!  This has to go... */
 
 #ifdef  KMEMSTATS
 struct kmemstats kmemstats[M_LAST];
@@ -173,17 +182,30 @@ vm_map_t  mb_map;
 semaphore_t execve_semaphore;
 
 int    cmask = CMASK;
+extern int customnbuf;
 
 int parse_bsd_args(void);
 extern int bsd_hardclockinit;
 extern task_t bsd_init_task;
 extern char    init_task_failure_data[];
 extern void time_zone_slock_init(void);
+static void process_name(char *, struct proc *);
+
+static void setconf(void);
 
 funnel_t *kernel_flock;
-funnel_t *network_flock;
-int disable_funnel = 0;                /* disables split funnel */
-int enable_funnel = 0;         /* disables split funnel */
+
+extern void sysv_shm_lock_init(void);
+extern void sysv_sem_lock_init(void);
+extern void sysv_msg_lock_init(void);
+extern void pshm_lock_init();
+extern void psem_lock_init();
+extern int maxprocperuid;
+
+/* kmem access not enabled by default; can be changed with boot-args */
+int setup_kmem = 0;
+
+extern void stackshot_lock_init();
 
 /*
  * Initialization code.
@@ -200,8 +222,8 @@ int enable_funnel = 0;              /* disables split funnel */
 /*
  *     Sets the name for the given task.
  */
-void
-proc_name(s, p)
+static void
+process_name(s, p)
        char            *s;
        struct proc *p;
 {
@@ -218,31 +240,48 @@ struct rlimit vm_initial_limit_stack = { DFLSSIZ, MAXSSIZ };
 struct rlimit vm_initial_limit_data = { DFLDSIZ, MAXDSIZ };
 struct rlimit vm_initial_limit_core = { DFLCSIZ, MAXCSIZ };
 
-extern thread_t first_thread;
-extern thread_act_t    cloneproc(struct proc *, int);
-extern int     (*mountroot) __P((void));
+extern thread_t        cloneproc(struct proc *, int);
+extern int     (*mountroot)(void);
 extern int     netboot_mountroot();    /* netboot.c */
 extern int     netboot_setup(struct proc * p);
 
+lck_grp_t * proc_lck_grp;
+lck_grp_attr_t * proc_lck_grp_attr;
+lck_attr_t * proc_lck_attr;
+
 /* hook called after root is mounted XXX temporary hack */
 void (*mountroot_post_hook)(void);
 
+/*
+ * This function is called very early on in the Mach startup, from the
+ * function start_kernel_threads() in osfmk/kern/startup.c.  It's called
+ * in the context of the current (startup) task using a call to the
+ * function kernel_thread_create() to jump into start_kernel_threads().
+ * Internally, kernel_thread_create() calls thread_create_internal(),
+ * which calls uthread_alloc().  The function of uthread_alloc() is
+ * normally to allocate a uthread structure, and fill out the uu_sigmask,
+ * uu_act, and uu_ucred fields.  It skips filling these out in the case
+ * of the "task" being "kernel_task", because the order of operation is
+ * inverted.  To account for that, we need to manually fill in at least
+ * the uu_cred field so that the uthread structure can be used like any
+ * other.
+ */
 void
 bsd_init()
 {
        register struct proc *p;
-       extern struct ucred *rootcred;
+       struct uthread *ut;
+       extern kauth_cred_t rootcred;
        register int i;
        int s;
+       int error;
        thread_t        th;
+       struct vfs_context context;
        void            lightning_bolt(void );
        kern_return_t   ret;
        boolean_t funnel_state;
-       extern void uthread_zone_init();
-
-
-       /* split funnel is enabled by default */
-       PE_parse_boot_arg("dfnl", &disable_funnel);
+       struct ucred temp_cred;
+       extern void file_lock_init(void);
 
        kernel_flock = funnel_alloc(KERNEL_FUNNEL);
        if (kernel_flock == (funnel_t *)0 ) {
@@ -251,29 +290,19 @@ bsd_init()
         
        funnel_state = thread_funnel_set(kernel_flock, TRUE);
 
-       if (!disable_funnel) {
-               network_flock = funnel_alloc(NETWORK_FUNNEL);
-               if (network_flock == (funnel_t *)0 ) {
-                       panic("bsd_init: Failed to allocate network funnel");
-               }
-       } else {
-               network_flock = kernel_flock;
-       }
-
        printf(copyright);
-
+       
        kmeminit();
        
        parse_bsd_args();
 
-       bsd_bufferinit();
-
        /* Initialize the uthread zone */
-       uthread_zone_init();
+       //uthread_zone_init();  /* XXX redundant: previous uthread_alloc() */
 
-       /*
-        * Initialize process and pgrp structures.
-        */
+       /* Initialize kauth subsystem before instancing the first credential */
+       kauth_init();
+
+       /* Initialize process and pgrp structures. */
        procinit();
 
        kernproc = &proc0;
@@ -285,12 +314,25 @@ bsd_init()
        p->p_pid = 0;
 
        /* give kernproc a name */
-       proc_name("kernel_task", p);
+       process_name("kernel_task", p);
+
+       /* allocate proc lock group attribute and group */
+       proc_lck_grp_attr= lck_grp_attr_alloc_init();
+       
+       proc_lck_grp = lck_grp_alloc_init("proc",  proc_lck_grp_attr);
+
+       /* Allocate proc lock attribute */
+       proc_lck_attr = lck_attr_alloc_init();
+
+       lck_mtx_init(&p->p_mlock, proc_lck_grp, proc_lck_attr);
+       lck_mtx_init(&p->p_fdmlock, proc_lck_grp, proc_lck_attr);
 
        if (current_task() != kernel_task)
                printf("bsd_init: We have a problem, "
                                "current task is not kernel task\n");
        
+       ut = (uthread_t)get_bsdthread_info(current_thread());
+
        /*
         * Create process 0.
         */
@@ -307,7 +349,7 @@ bsd_init()
        p->task = kernel_task;
        
        p->p_stat = SRUN;
-       p->p_flag = P_INMEM|P_SYSTEM;
+       p->p_flag = P_SYSTEM;
        p->p_nice = NZERO;
        p->p_pptr = p;
        lockinit(&p->signal_lock, PVM, "signal", 0, 0);
@@ -316,20 +358,26 @@ bsd_init()
        p->sigwait_thread = THREAD_NULL;
        p->exit_thread = THREAD_NULL;
 
-       /* Create credentials. */
-       lockinit(&cred0.pc_lock, PLOCK, "proc0 cred", 0, 0);
-       cred0.p_refcnt = 1;
-       p->p_cred = &cred0;
-       p->p_ucred = crget();
-       p->p_ucred->cr_ngroups = 1;     /* group 0 */
+       /*
+        * Create credential.  This also Initializes the audit information.
+        * XXX It is not clear what the initial values should be for audit ID,
+        * XXX session ID, etc..
+        */
+       bzero(&temp_cred, sizeof(temp_cred));
+       temp_cred.cr_ngroups = 1;
+
+       p->p_ucred = kauth_cred_create(&temp_cred); 
+
+       /* give the (already exisiting) initial thread a reference on it */
+       kauth_cred_ref(p->p_ucred);
+       ut->uu_ucred = p->p_ucred;
        
        TAILQ_INIT(&p->aio_activeq);
        TAILQ_INIT(&p->aio_doneq);
        p->aio_active_count = 0;
        p->aio_done_count = 0;
 
-       /* Set the audit info for this process */
-       audit_proc_init(p);
+       file_lock_init();
 
        /* Create the file descriptor table. */
        filedesc0.fd_refcnt = 1+1;      /* +1 so shutdown will not _FREE_ZONE */
@@ -346,7 +394,7 @@ bsd_init()
                limit0.pl_rlimit[i].rlim_cur = 
                        limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY;
        limit0.pl_rlimit[RLIMIT_NOFILE].rlim_cur = NOFILE;
-       limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = MAXUPRC;
+       limit0.pl_rlimit[RLIMIT_NPROC].rlim_cur = maxprocperuid;
        limit0.pl_rlimit[RLIMIT_NPROC].rlim_max = maxproc;
        limit0.pl_rlimit[RLIMIT_STACK] = vm_initial_limit_stack;
        limit0.pl_rlimit[RLIMIT_DATA] = vm_initial_limit_data;
@@ -357,7 +405,7 @@ bsd_init()
        p->p_sigacts = &sigacts0;
 
        /*
-        * Charge root for one process.
+        * Charge root for two  processes: init and mach_init.
         */
        (void)chgproccnt(0, 1);
 
@@ -372,12 +420,21 @@ bsd_init()
                                &min,
                                (vm_size_t)BSD_PAGABLE_MAP_SIZE,
                                TRUE,
-                               TRUE,
+                               VM_FLAGS_ANYWHERE,
                                &bsd_pageable_map);
                if (ret != KERN_SUCCESS) 
                        panic("bsd_init: Failed to allocate bsd pageable map");
        }
 
+       /*
+        * Initialize buffers and hash links for buffers
+        *
+        * SIDE EFFECT: Starts a thread for bcleanbuf_thread(), so must
+        *              happen after a credential has been associated with
+        *              the kernel task.
+        */
+       bsd_bufferinit();
+
        /* Initialize the execve() semaphore */
        ret = semaphore_create(kernel_task, &execve_semaphore,
                        SYNC_POLICY_FIFO, (BSD_PAGABLE_MAP_SIZE / NCARGS));
@@ -397,9 +454,6 @@ bsd_init()
        /* Initialize mbuf's. */
        mbinit();
 
-       /* Initialize syslog */
-       log_init();
-
        /*
         * Initializes security event auditing.
         * XXX: Should/could this occur later?
@@ -412,22 +466,35 @@ bsd_init()
        /* Initialize for async IO */
        aio_init();
 
+       /* Initialize pipes */
+       pipeinit();
+
+       /* Initialize SysV shm subsystem locks; the subsystem proper is
+        * initialized through a sysctl.
+        */
+       sysv_shm_lock_init();
+       sysv_sem_lock_init();
+       sysv_msg_lock_init();
+       pshm_lock_init();
+       psem_lock_init();
+
        /* POSIX Shm and Sem */
        pshm_cache_init();
        psem_cache_init();
        time_zone_slock_init();
 
+       /* Stack snapshot facility lock */
+       stackshot_lock_init();
        /*
         * Initialize protocols.  Block reception of incoming packets
         * until everything is ready.
         */
-       s = splimp();
        sysctl_register_fixed(); 
        sysctl_mib_init();
        dlil_init();
+       proto_kpi_init();
        socketinit();
        domaininit();
-       splx(s);
 
        p->p_fd->fd_cdir = NULL;
        p->p_fd->fd_rdir = NULL;
@@ -456,42 +523,53 @@ bsd_init()
         /* Register the built-in dlil ethernet interface family */
        ether_family_init();
 
+       /* Call any kext code that wants to run just after network init */
+       net_init_run();
+
        vnode_pager_bootstrap();
+#if 0
+       /* XXX Hack for early debug stop */
+       printf("\nabout to sleep for 10 seconds\n");
+       IOSleep( 10 * 1000 );
+       /* Debugger("hello"); */
+#endif
+
+       inittodr(0);
 
        /* Mount the root file system. */
        while( TRUE) {
                int err;
 
                setconf();
-               /*
-                * read the time after clock_initialize_calendar()
-                * and before nfs mount
-                */
-               microtime((struct timeval  *)&time);
-
                bsd_hardclockinit = -1; /* start ticking */
 
                if (0 == (err = vfs_mountroot()))
                        break;
+#if NFSCLIENT
                if (mountroot == netboot_mountroot) {
                        printf("cannot mount network root, errno = %d\n", err);
                        mountroot = NULL;
                        if (0 == (err = vfs_mountroot()))
                                break;
                }
+#endif
                printf("cannot mount root, errno = %d\n", err);
                boothowto |= RB_ASKNAME;
        }
 
-       mountlist.cqh_first->mnt_flag |= MNT_ROOTFS;
+       context.vc_proc = p;
+       context.vc_ucred = p->p_ucred;
+       mountlist.tqh_first->mnt_flag |= MNT_ROOTFS;
 
        /* Get the vnode for '/'.  Set fdp->fd_fd.fd_cdir to reference it. */
-       if (VFS_ROOT(mountlist.cqh_first, &rootvnode))
+       if (VFS_ROOT(mountlist.tqh_first, &rootvnode, &context))
                panic("bsd_init: cannot find root vnode");
-       VREF(rootvnode);
+       rootvnode->v_flag |= VROOT;
+       (void)vnode_ref(rootvnode);
+       (void)vnode_put(rootvnode);
        filedesc0.fd_cdir = rootvnode;
-       VOP_UNLOCK(rootvnode, 0, p);
 
+#if NFSCLIENT
        if (mountroot == netboot_mountroot) {
                int err;
                /* post mount setup */
@@ -499,14 +577,10 @@ bsd_init()
                        panic("bsd_init: NetBoot could not find root, %d", err);
                }
        }
+#endif
        
 
-       /*
-        * Now can look at time, having had a chance to verify the time
-        * from the file system.  Reset p->p_rtime as it may have been
-        * munched in mi_switch() after the time got set.
-        */
-       p->p_stats->p_start = boottime = time;
+       microtime(&p->p_stats->p_start);
        p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0;
 
 #if DEVFS
@@ -522,6 +596,13 @@ bsd_init()
 
        bsd_utaskbootstrap();
 
+#if __i386__
+       // this should be done after the root filesystem is mounted
+       error = set_archhandler(p, CPU_TYPE_POWERPC);
+       if (error)
+               exec_archhandler_ppc.path[0] = 0;
+#endif 
+       
        /* invoke post-root-mount hook */
        if (mountroot_post_hook != NULL)
                mountroot_post_hook();
@@ -536,14 +617,14 @@ bsdinit_task(void)
        struct proc *p = current_proc();
        struct uthread *ut;
        kern_return_t   kr;
-       thread_act_t th_act;
+       thread_t th_act;
        shared_region_mapping_t system_region;
 
-       proc_name("init", p);
+       process_name("init", p);
 
        ux_handler_init();
 
-       th_act = current_act();
+       th_act = current_thread();
        (void) host_set_exception_ports(host_priv_self(),
                                        EXC_MASK_ALL & ~(EXC_MASK_SYSCALL |
                                                         EXC_MASK_MACH_SYSCALL |
@@ -562,22 +643,20 @@ bsdinit_task(void)
 
 
        ut = (uthread_t)get_bsdthread_info(th_act);
-       ut->uu_ar0 = (void *)get_user_regs(th_act);
 
        bsd_hardclockinit = 1;  /* Start bsd hardclock */
        bsd_init_task = get_threadtask(th_act);
        init_task_failure_data[0] = 0;
-       system_region = lookup_default_shared_region(ENV_DEFAULT_ROOT,
-                               machine_slot[cpu_number()].cpu_type);
+       system_region = lookup_default_shared_region(ENV_DEFAULT_ROOT, cpu_type());
         if (system_region == NULL) {
-               shared_file_boot_time_init(ENV_DEFAULT_ROOT,
-                               machine_slot[cpu_number()].cpu_type);
+               shared_file_boot_time_init(ENV_DEFAULT_ROOT, cpu_type());
        } else {
                vm_set_shared_region(get_threadtask(th_act), system_region);
        }
        load_init_program(p);
        /* turn on app-profiling i.e. pre-heating */
        app_profile = 1;
+       lock_trace = 1;
 }
 
 void
@@ -617,7 +696,8 @@ bsd_autoconf()
 
 #include <sys/disklabel.h>  /* for MAXPARTITIONS */
 
-setconf()
+static void
+setconf(void)
 {      
        extern kern_return_t IOFindBSDRoot( char * rootName,
                                dev_t * root, u_int32_t * flags );
@@ -640,25 +720,29 @@ setconf()
                flags = 0;
        }
 
+#if NFSCLIENT
        if( flags & 1 ) {
                /* network device */
                mountroot = netboot_mountroot;
        } else {
+#endif
                /* otherwise have vfs determine root filesystem */
                mountroot = NULL;
+#if NFSCLIENT
        }
+#endif
 
 }
 
 bsd_utaskbootstrap()
 {
-       thread_act_t th_act;
+       thread_t th_act;
        struct uthread *ut;
 
        th_act = cloneproc(kernproc, 0);
        initproc = pfind(1);                            
        /* Set the launch time for init */
-       initproc->p_stats->p_start = time;
+       microtime(&initproc->p_stats->p_start);
 
        ut = (struct uthread *)get_bsdthread_info(th_act);
        ut->uu_sigmask = 0;
@@ -728,61 +812,17 @@ parse_bsd_args()
 
        PE_parse_boot_arg("srv", &srv);
        PE_parse_boot_arg("ncl", &ncl);
-       PE_parse_boot_arg("nbuf", &nbuf);
+       if (PE_parse_boot_arg("nbuf", &max_nbuf_headers)) 
+               customnbuf = 1;
+       PE_parse_boot_arg("kmem", &setup_kmem);
 
        return 0;
 }
 
-boolean_t
-thread_funnel_switch(
-        int    oldfnl,
-       int     newfnl)
+#if !NFSCLIENT
+int 
+netboot_root(void)
 {
-       boolean_t       funnel_state_prev;
-       int curfnl;
-       funnel_t * curflock;
-       funnel_t * oldflock;
-       funnel_t * newflock;
-       funnel_t * exist_funnel;
-       extern int disable_funnel;
-       
-        
-               if (disable_funnel)
-                       return(TRUE);
-
-        if(oldfnl == newfnl) {
-            panic("thread_funnel_switch: can't switch to same funnel");
-        }
-        
-        if ((oldfnl != NETWORK_FUNNEL) && (oldfnl != KERNEL_FUNNEL)) {
-            panic("thread_funnel_switch: invalid oldfunnel");
-        }
-        if ((newfnl != NETWORK_FUNNEL) && (newfnl != KERNEL_FUNNEL)) {
-            panic("thread_funnel_switch: invalid newfunnel");
-        }
-        
-       if((curflock = thread_funnel_get()) == THR_FUNNEL_NULL) {
-            panic("thread_funnel_switch: no funnel held");
-       }
-        
-        if ((oldfnl == NETWORK_FUNNEL) && (curflock != network_flock))
-            panic("thread_funnel_switch: network funnel not held");
-            
-        if ((oldfnl == KERNEL_FUNNEL) && (curflock != kernel_flock))
-            panic("thread_funnel_switch: kernel funnel not held");
-
-        if(oldfnl == NETWORK_FUNNEL) {
-            oldflock = network_flock;
-            newflock = kernel_flock;
-        } else {
-            oldflock = kernel_flock;
-            newflock = network_flock;
-        }
-               KERNEL_DEBUG(0x603242c | DBG_FUNC_NONE, oldflock, 1, 0, 0, 0);
-        thread_funnel_set(oldflock, FALSE);
-               KERNEL_DEBUG(0x6032428 | DBG_FUNC_NONE, newflock, 1, 0, 0, 0);
-        thread_funnel_set(newflock, TRUE);
-               KERNEL_DEBUG(0x6032434 | DBG_FUNC_NONE, newflock, 1, 0, 0, 0);
-
-        return(TRUE);        
+       return(0);
 }
+#endif