X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/39236c6e673c41db228275375ab7fdb0f837b292..9d749ea394c01276fa19e397e70f46858e849c76:/bsd/kern/bsd_init.c diff --git a/bsd/kern/bsd_init.c b/bsd/kern/bsd_init.c index ace8c58ba..36b5db056 100644 --- a/bsd/kern/bsd_init.c +++ b/bsd/kern/bsd_init.c @@ -1,5 +1,5 @@ /* - * Copyright (c) 2000-2013 Apple Inc. All rights reserved. + * Copyright (c) 2000-2015 Apple Inc. All rights reserved. * * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * @@ -94,6 +94,7 @@ #include #include #include +#include #include @@ -119,6 +120,8 @@ #include #include /* for pseudo_inits */ #include +#include +#include #include #include @@ -133,6 +136,7 @@ #include /* for mcache_init() */ #include /* for mbinit() */ #include /* for knote_init() */ +#include /* for eventhandler_init() */ #include /* for memorystatus_init() */ #include /* for aio_init() */ #include /* for psem_cache_init() */ @@ -155,11 +159,18 @@ #include /* for tty_init() */ #include /* proc_uuid_policy_init() */ #include /* flow_divert_init() */ +#include /* for cfil_init() */ +#include /* for necp_init() */ +#include /* for netagent_init() */ +#include /* for pkt_mnglr_init() */ #include /* for utun_register_control() */ -#include /* for ipsec_register_control() */ +#include /* for ipsec_register_control() */ #include /* for net_str_id_init() */ #include /* for netsrc_init() */ #include /* for nstat_init() */ +#include /* for tcp_cc_init() */ +#include /* for mptcp_control_register() */ +#include /* for nwk_wq_init */ #include /* for assert() */ #include /* for init_system_override() */ @@ -185,10 +196,12 @@ #include #endif + #include #include #include + void * get_user_regs(thread_t); /* XXX kludge for */ void IOKitInitializeTime(void); /* XXX */ void IOSleep(unsigned int); /* XXX */ @@ -231,7 +244,7 @@ int hostnamelen; char domainname[MAXDOMNAMELEN]; int domainnamelen; -char rootdevice[16]; /* hfs device names have at least 9 chars */ +char rootdevice[DEVMAXNAMESIZE]; #if KMEMSTATS struct kmemstats kmemstats[M_LAST]; @@ -239,6 +252,18 @@ struct kmemstats kmemstats[M_LAST]; struct vnode *rootvp; int boothowto = RB_DEBUG; +int minimalboot = 0; +#if CONFIG_EMBEDDED +int darkboot = 0; +#endif + +#if PROC_REF_DEBUG +__private_extern__ int proc_ref_tracking_disabled = 0; /* disable panics on leaked proc refs across syscall boundary */ +#endif + +#if OS_REASON_DEBUG +__private_extern__ int os_reason_debug_disabled = 0; /* disable asserts for when we fail to allocate OS reasons */ +#endif extern kern_return_t IOFindBSDRoot(char *, unsigned int, dev_t *, u_int32_t *); extern void IOSecureBSDRoot(const char * rootName); @@ -247,7 +272,9 @@ extern void kminit(void); extern void file_lock_init(void); extern void kmeminit(void); extern void bsd_bufferinit(void); +extern void oslog_setsize(int size); extern void throttle_init(void); +extern void acct_init(void); extern int serverperfmode; extern int ncl; @@ -263,8 +290,18 @@ __private_extern__ vm_offset_t * execargs_cache = NULL; void bsd_exec_setup(int); +#if __arm64__ +__private_extern__ int bootarg_no64exec = 0; +#endif +#if __x86_64__ +__private_extern__ int bootarg_no32exec = 0; +#endif __private_extern__ int bootarg_vnode_cache_defeat = 0; +#if CONFIG_JETSAM && (DEVELOPMENT || DEBUG) +__private_extern__ int bootarg_no_vnode_jetsam = 0; +#endif /* CONFIG_JETSAM && (DEVELOPMENT || DEBUG) */ + /* * Prevent kernel-based ASLR from being used, for testing. */ @@ -272,24 +309,31 @@ __private_extern__ int bootarg_vnode_cache_defeat = 0; __private_extern__ int bootarg_disable_aslr = 0; #endif +/* + * Allow an alternate dyld to be used for testing. + */ + +#if DEVELOPMENT || DEBUG +char dyld_alt_path[MAXPATHLEN]; +int use_alt_dyld = 0; +#endif + int cmask = CMASK; extern int customnbuf; -void bsd_init(void); kern_return_t bsd_autoconf(void); void bsd_utaskbootstrap(void); static void parse_bsd_args(void); -extern task_t bsd_init_task; -extern char init_task_failure_data[]; +#if CONFIG_DEV_KMEM +extern void dev_kmem_init(void); +#endif extern void time_zone_slock_init(void); -extern void select_wait_queue_init(void); +extern void select_waitq_init(void); static void process_name(const char *, proc_t); static void setconf(void); -funnel_t *kernel_flock; - #if SYSV_SHM extern void sysv_shm_lock_init(void); #endif @@ -300,11 +344,7 @@ extern void sysv_sem_lock_init(void); extern void sysv_msg_lock_init(void); #endif -#if !defined(SECURE_KERNEL) -/* kmem access not enabled by default; can be changed with boot-args */ -/* We don't need to keep this symbol around in RELEASE kernel */ -int setup_kmem = 0; -#endif +extern void ulock_initialize(void); #if CONFIG_MACF #if defined (__i386__) || defined (__x86_64__) @@ -338,37 +378,49 @@ extern int check_policy_init(int); static void process_name(const char *s, proc_t p) { - size_t length = strlen(s); - - bcopy(s, p->p_comm, - length >= sizeof(p->p_comm) ? sizeof(p->p_comm) : - length + 1); + strlcpy(p->p_comm, s, sizeof(p->p_comm)); + strlcpy(p->p_name, s, sizeof(p->p_name)); } /* To allow these values to be patched, they're globals here */ #include -struct rlimit vm_initial_limit_stack = { DFLSSIZ, MAXSSIZ - PAGE_SIZE }; +struct rlimit vm_initial_limit_stack = { DFLSSIZ, MAXSSIZ - PAGE_MAX_SIZE }; struct rlimit vm_initial_limit_data = { DFLDSIZ, MAXDSIZ }; struct rlimit vm_initial_limit_core = { DFLCSIZ, MAXCSIZ }; -extern thread_t cloneproc(task_t, proc_t, int, int); +extern thread_t cloneproc(task_t, coalition_t, proc_t, int, int); extern int (*mountroot)(void); lck_grp_t * proc_lck_grp; lck_grp_t * proc_slock_grp; lck_grp_t * proc_fdmlock_grp; +lck_grp_t * proc_kqhashlock_grp; +lck_grp_t * proc_knhashlock_grp; +lck_grp_t * proc_ucred_mlock_grp; lck_grp_t * proc_mlock_grp; lck_grp_attr_t * proc_lck_grp_attr; lck_attr_t * proc_lck_attr; lck_mtx_t * proc_list_mlock; lck_mtx_t * proc_klist_mlock; + extern lck_mtx_t * execargs_cache_lock; /* hook called after root is mounted XXX temporary hack */ void (*mountroot_post_hook)(void); void (*unmountroot_pre_hook)(void); +/* + * This function is called before IOKit initialization, so that globals + * like the sysctl tree are initialized before kernel extensions + * are started (since they may want to register sysctls + */ +void +bsd_early_init(void) +{ + sysctl_early_init(); +} + /* * 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 @@ -384,8 +436,6 @@ void (*unmountroot_pre_hook)(void); * used like any other. */ -extern void IOServicePublishResource(const char *, boolean_t); - void bsd_init(void) { @@ -403,11 +453,6 @@ bsd_init(void) throttle_init(); - kernel_flock = funnel_alloc(KERNEL_FUNNEL); - if (kernel_flock == (funnel_t *)0 ) { - panic("bsd_init: Failed to allocate kernel funnel"); - } - printf(copyright); bsd_init_kprintf("calling kmeminit\n"); @@ -416,6 +461,11 @@ bsd_init(void) bsd_init_kprintf("calling parse_bsd_args\n"); parse_bsd_args(); +#if CONFIG_DEV_KMEM + bsd_init_kprintf("calling dev_kmem_init\n"); + dev_kmem_init(); +#endif + /* Initialize kauth subsystem before instancing the first credential */ bsd_init_kprintf("calling kauth_init\n"); kauth_init(); @@ -441,11 +491,15 @@ bsd_init(void) proc_lck_grp_attr= lck_grp_attr_alloc_init(); proc_lck_grp = lck_grp_alloc_init("proc", proc_lck_grp_attr); + #if CONFIG_FINE_LOCK_GROUPS proc_slock_grp = lck_grp_alloc_init("proc-slock", proc_lck_grp_attr); - proc_fdmlock_grp = lck_grp_alloc_init("proc-fdmlock", proc_lck_grp_attr); + proc_ucred_mlock_grp = lck_grp_alloc_init("proc-ucred-mlock", proc_lck_grp_attr); proc_mlock_grp = lck_grp_alloc_init("proc-mlock", proc_lck_grp_attr); + proc_fdmlock_grp = lck_grp_alloc_init("proc-fdmlock", proc_lck_grp_attr); #endif + proc_kqhashlock_grp = lck_grp_alloc_init("proc-kqhashlock", proc_lck_grp_attr); + proc_knhashlock_grp = lck_grp_alloc_init("proc-knhashlock", proc_lck_grp_attr); /* Allocate proc lock attribute */ proc_lck_attr = lck_attr_alloc_init(); #if 0 @@ -459,12 +513,14 @@ bsd_init(void) proc_klist_mlock = lck_mtx_alloc_init(proc_mlock_grp, proc_lck_attr); lck_mtx_init(&kernproc->p_mlock, proc_mlock_grp, proc_lck_attr); lck_mtx_init(&kernproc->p_fdmlock, proc_fdmlock_grp, proc_lck_attr); + lck_mtx_init(&kernproc->p_ucred_mlock, proc_ucred_mlock_grp, proc_lck_attr); lck_spin_init(&kernproc->p_slock, proc_slock_grp, proc_lck_attr); #else proc_list_mlock = lck_mtx_alloc_init(proc_lck_grp, proc_lck_attr); proc_klist_mlock = lck_mtx_alloc_init(proc_lck_grp, proc_lck_attr); lck_mtx_init(&kernproc->p_mlock, proc_lck_grp, proc_lck_attr); lck_mtx_init(&kernproc->p_fdmlock, proc_lck_grp, proc_lck_attr); + lck_mtx_init(&kernproc->p_ucred_mlock, proc_lck_grp, proc_lck_attr); lck_spin_init(&kernproc->p_slock, proc_lck_grp, proc_lck_attr); #endif @@ -487,7 +543,6 @@ bsd_init(void) * Initialize the MAC Framework */ mac_policy_initbsd(); - kernproc->p_mac_enforce = 0; #if defined (__i386__) || defined (__x86_64__) /* @@ -500,6 +555,8 @@ bsd_init(void) /* Initialize System Override call */ init_system_override(); + + ulock_initialize(); /* * Create process 0. @@ -534,8 +591,8 @@ bsd_init(void) LIST_INSERT_HEAD(SESSHASH(0), &session0, s_hash); proc_list_unlock(); -#if CONFIG_LCTX - kernproc->p_lctx = NULL; +#if CONFIG_PERSONAS + kernproc->p_persona = NULL; #endif kernproc->task = kernel_task; @@ -544,7 +601,11 @@ bsd_init(void) kernproc->p_flag = P_SYSTEM; kernproc->p_lflag = 0; kernproc->p_ladvflag = 0; - + +#if defined(__LP64__) + kernproc->p_flag |= P_LP64; +#endif + #if DEVELOPMENT || DEBUG if (bootarg_disable_aslr) kernproc->p_flag |= P_DISABLE_ASLR; @@ -599,7 +660,6 @@ bsd_init(void) #if CONFIG_MACF mac_cred_label_associate_kernel(kernproc->p_ucred); - mac_task_label_update_cred (kernproc->p_ucred, (struct task *) kernproc->task); #endif /* Create the file descriptor table. */ @@ -609,6 +669,8 @@ bsd_init(void) filedesc0.fd_knlist = NULL; filedesc0.fd_knhash = NULL; filedesc0.fd_knhashmask = 0; + lck_mtx_init(&filedesc0.fd_kqhashlock, proc_kqhashlock_grp, proc_lck_attr); + lck_mtx_init(&filedesc0.fd_knhashlock, proc_knhashlock_grp, proc_lck_attr); /* Create the limits structures. */ kernproc->p_limit = &limit0; @@ -627,7 +689,7 @@ bsd_init(void) kernproc->p_sigacts = &sigacts0; /* - * Charge root for two processes: init and mach_init. + * Charge root for one process: launchd. */ bsd_init_kprintf("calling chgproccnt\n"); (void)chgproccnt(0, 1); @@ -646,11 +708,16 @@ bsd_init(void) (vm_size_t)bsd_pageable_map_size, TRUE, VM_FLAGS_ANYWHERE, + VM_MAP_KERNEL_FLAGS_NONE, + VM_KERN_MEMORY_BSD, &bsd_pageable_map); if (ret != KERN_SUCCESS) panic("bsd_init: Failed to allocate bsd pageable map"); } + bsd_init_kprintf("calling fpxlog_init\n"); + fpxlog_init(); + /* * Initialize buffers and hash links for buffers * @@ -661,12 +728,6 @@ bsd_init(void) bsd_init_kprintf("calling bsd_bufferinit\n"); bsd_bufferinit(); - /* Initialize the execve() semaphore */ - bsd_init_kprintf("calling semaphore_create\n"); - - if (ret != KERN_SUCCESS) - panic("bsd_init: Failed to create execve semaphore"); - /* * Initialize the calendar. */ @@ -676,12 +737,6 @@ bsd_init(void) bsd_init_kprintf("calling ubc_init\n"); ubc_init(); - /* - * Initialize device-switches. - */ - bsd_init_kprintf("calling devsw_init() \n"); - devsw_init(); - /* Initialize the file systems. */ bsd_init_kprintf("calling vfsinit\n"); vfsinit(); @@ -715,6 +770,10 @@ bsd_init(void) bsd_init_kprintf("calling knote_init\n"); knote_init(); + /* Initialize event handler */ + bsd_init_kprintf("calling eventhandler_init\n"); + eventhandler_init(); + /* Initialize for async IO */ bsd_init_kprintf("calling aio_init\n"); aio_init(); @@ -751,18 +810,16 @@ bsd_init(void) psem_cache_init(); bsd_init_kprintf("calling time_zone_slock_init\n"); time_zone_slock_init(); - bsd_init_kprintf("calling select_wait_queue_init\n"); - select_wait_queue_init(); + bsd_init_kprintf("calling select_waitq_init\n"); + select_waitq_init(); /* * Initialize protocols. Block reception of incoming packets * until everything is ready. */ - bsd_init_kprintf("calling sysctl_register_fixed\n"); - sysctl_register_fixed(); - bsd_init_kprintf("calling sysctl_mib_init\n"); - sysctl_mib_init(); #if NETWORKING + bsd_init_kprintf("calling nwk_wq_init\n"); + nwk_wq_init(); bsd_init_kprintf("calling dlil_init\n"); dlil_init(); bsd_init_kprintf("calling proto_kpi_init\n"); @@ -778,7 +835,6 @@ bsd_init(void) flow_divert_init(); #endif /* FLOW_DIVERT */ #endif /* SOCKETS */ - kernproc->p_fd->fd_cdir = NULL; kernproc->p_fd->fd_rdir = NULL; @@ -797,14 +853,23 @@ bsd_init(void) memorystatus_init(); #endif /* CONFIG_MEMORYSTATUS */ + bsd_init_kprintf("calling acct_init\n"); + acct_init(); + #ifdef GPROF /* Initialize kernel profiling. */ kmstartup(); #endif + bsd_init_kprintf("calling sysctl_mib_init\n"); + sysctl_mib_init() + bsd_init_kprintf("calling bsd_autoconf\n"); bsd_autoconf(); + bsd_init_kprintf("calling os_reason_init\n"); + os_reason_init(); + #if CONFIG_DTRACE dtrace_postinit(); #endif @@ -840,6 +905,21 @@ bsd_init(void) bsd_init_kprintf("calling net_init_run\n"); net_init_run(); +#if CONTENT_FILTER + cfil_init(); +#endif + +#if PACKET_MANGLER + pkt_mnglr_init(); +#endif + +#if NECP + /* Initialize Network Extension Control Policies */ + necp_init(); +#endif + + netagent_init(); + /* register user tunnel kernel control handler */ utun_register_control(); #if IPSEC @@ -847,16 +927,14 @@ bsd_init(void) #endif /* IPSEC */ netsrc_init(); nstat_init(); + tcp_cc_init(); +#if MPTCP + mptcp_control_register(); +#endif /* MPTCP */ #endif /* NETWORKING */ bsd_init_kprintf("calling vnode_pager_bootstrap\n"); 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 bsd_init_kprintf("calling inittodr\n"); inittodr(0); @@ -953,7 +1031,7 @@ bsd_init(void) devfs_kernel_mount(mounthere); } #endif /* DEVFS */ - + /* Initialize signal state for process 0. */ bsd_init_kprintf("calling siginit\n"); siginit(kernproc); @@ -961,10 +1039,6 @@ bsd_init(void) bsd_init_kprintf("calling bsd_utaskbootstrap\n"); bsd_utaskbootstrap(); -#if defined(__LP64__) - kernproc->p_flag |= P_LP64; -#endif - pal_kernel_announce(); bsd_init_kprintf("calling mountroot_post_hook\n"); @@ -978,10 +1052,8 @@ bsd_init(void) #endif bsd_init_kprintf("done\n"); - } -/* Called with kernel funnel held */ void bsdinit_task(void) { @@ -1002,13 +1074,15 @@ bsdinit_task(void) ut = (uthread_t)get_bsdthread_info(thread); - bsd_init_task = get_threadtask(thread); - init_task_failure_data[0] = 0; - #if CONFIG_MACF mac_cred_label_associate_user(p->p_ucred); - mac_task_label_update_cred (p->p_ucred, (struct task *) p->task); #endif + + vm_init_before_launchd(); + + + bsd_init_kprintf("bsd_do_post - done"); + load_init_program(p); lock_trace = 1; } @@ -1041,11 +1115,6 @@ setconf(void) u_int32_t flags; kern_return_t err; - /* - * calls into IOKit can generate networking registrations - * which needs to be under network funnel. Right thing to do - * here is to drop the funnel alltogether and regrab it afterwards - */ err = IOFindBSDRoot(rootdevice, sizeof(rootdevice), &rootdev, &flags); if( err) { printf("setconf: IOFindBSDRoot returned an error (%d);" @@ -1079,7 +1148,7 @@ bsd_utaskbootstrap(void) * Clone the bootstrap process from the kernel process, without * inheriting either task characteristics or memory from the kernel; */ - thread = cloneproc(TASK_NULL, kernproc, FALSE, TRUE); + thread = cloneproc(TASK_NULL, COALITION_NULL, kernproc, FALSE, TRUE); /* Hold the reference as it will be dropped during shutdown */ initproc = proc_find(1); @@ -1097,7 +1166,7 @@ bsd_utaskbootstrap(void) ut = (struct uthread *)get_bsdthread_info(thread); ut->uu_sigmask = 0; act_set_astbsd(thread); - (void) thread_resume(thread); + task_clear_return_wait(get_threadtask(thread)); } static void @@ -1115,6 +1184,26 @@ parse_bsd_args(void) if (PE_parse_boot_argn("-x", namep, sizeof (namep))) /* safe boot */ boothowto |= RB_SAFEBOOT; + if (PE_parse_boot_argn("-minimalboot", namep, sizeof(namep))) { + /* + * -minimalboot indicates that we want userspace to be bootstrapped to a + * minimal environment. What constitutes minimal is up to the bootstrap + * process. + */ + minimalboot = 1; + } + +#if __arm64__ + /* disable 64 bit grading */ + if (PE_parse_boot_argn("-no64exec", namep, sizeof (namep))) + bootarg_no64exec = 1; +#endif +#if __x86_64__ + /* disable 32 bit grading */ + if (PE_parse_boot_argn("-no32exec", namep, sizeof (namep))) + bootarg_no32exec = 1; +#endif + /* disable vnode_cache_is_authorized() by setting vnode_cache_defeat */ if (PE_parse_boot_argn("-vnode_cache_defeat", namep, sizeof (namep))) bootarg_vnode_cache_defeat = 1; @@ -1129,9 +1218,6 @@ parse_bsd_args(void) sizeof (max_nbuf_headers))) { customnbuf = 1; } -#if !defined(SECURE_KERNEL) - PE_parse_boot_argn("kmem", &setup_kmem, sizeof (setup_kmem)); -#endif #if CONFIG_MACF #if defined (__i386__) || defined (__x86_64__) @@ -1141,12 +1227,59 @@ parse_bsd_args(void) if (PE_parse_boot_argn("msgbuf", &msgbuf, sizeof (msgbuf))) { log_setsize(msgbuf); + oslog_setsize(msgbuf); } if (PE_parse_boot_argn("-novfscache", namep, sizeof(namep))) { nc_disabled = 1; } +#if CONFIG_JETSAM && (DEVELOPMENT || DEBUG) + if (PE_parse_boot_argn("-no_vnode_jetsam", namep, sizeof(namep))) + bootarg_no_vnode_jetsam = 1; +#endif /* CONFIG_JETSAM && (DEVELOPMENT || DEBUG) */ + + +#if CONFIG_EMBEDDED + /* + * The darkboot flag is specified by the bootloader and is stored in + * boot_args->bootFlags. This flag is available starting revision 2. + */ + boot_args *args = (boot_args *) PE_state.bootArgs; + if ((args != NULL) && (args->Revision >= kBootArgsRevision2)) { + darkboot = (args->bootFlags & kBootFlagsDarkBoot) ? 1 : 0; + } else { + darkboot = 0; + } +#endif + +#if PROC_REF_DEBUG + if (PE_parse_boot_argn("-disable_procref_tracking", namep, sizeof(namep))) { + proc_ref_tracking_disabled = 1; + } +#endif + +#if OS_REASON_DEBUG + if (PE_parse_boot_argn("-disable_osreason_debug", namep, sizeof(namep))) { + os_reason_debug_disabled = 1; + } +#endif + + PE_parse_boot_argn("sigrestrict", &sigrestrict_arg, sizeof(sigrestrict_arg)); + +#if DEVELOPMENT|| DEBUG + if (PE_parse_boot_argn("-no_sigsys", namep, sizeof(namep))) { + send_sigsys = false; + } +#endif + +#if (DEVELOPMENT|| DEBUG) + if (PE_parse_boot_argn("alt-dyld", dyld_alt_path, sizeof(dyld_alt_path))) { + if (strlen(dyld_alt_path) > 0) { + use_alt_dyld = 1; + } + } +#endif } void