X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/4452a7af2eac33dbad800bcc91f2399d62c18f53..HEAD:/bsd/kern/bsd_init.c diff --git a/bsd/kern/bsd_init.c b/bsd/kern/bsd_init.c index 862a0557a..6dd42a5f0 100644 --- a/bsd/kern/bsd_init.c +++ b/bsd/kern/bsd_init.c @@ -1,8 +1,8 @@ /* - * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved. + * Copyright (c) 2000-2020 Apple Inc. All rights reserved. * * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ - * + * * This file contains Original Code and/or Modifications of Original Code * as defined in and that are subject to the Apple Public Source License * Version 2.0 (the 'License'). You may not use this file except in @@ -11,10 +11,10 @@ * unlawful or unlicensed copies of an Apple operating system, or to * circumvent, violate, or enable the circumvention or violation of, any * terms of an Apple operating system software license agreement. - * + * * Please obtain a copy of the License at * http://www.opensource.apple.com/apsl/ and read it before using this file. - * + * * The Original Code and all software distributed under the License are * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES, @@ -22,10 +22,10 @@ * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT. * Please see the License for the specific language governing rights and * limitations under the License. - * + * * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ - */ -/* + * + * * Copyright (c) 1982, 1986, 1989, 1991, 1992, 1993 * The Regents of the University of California. All rights reserved. * (c) UNIX System Laboratories, Inc. @@ -65,13 +65,19 @@ * @(#)init_main.c 8.16 (Berkeley) 5/14/95 */ -/* +/* * * Mach Operating System * Copyright (c) 1987 Carnegie-Mellon University * All rights reserved. The CMU software License Agreement specifies * the terms and conditions for use and redistribution. */ +/* + * NOTICE: This file was modified by McAfee Research in 2004 to introduce + * support for mandatory and extensible security protections. This notice + * is included in support of clause 2.2 (b) of the Apple Public License, + * Version 2.0. + */ #include #include @@ -83,66 +89,133 @@ #include #include #include -#include #include #include #include #include +#include -#include +#include #include #include +#include #include #include #include #include +#include +#include /* for ux_handler_setup() */ #include #include #include -#include /* for ux_exception_port */ - #include -#include -#include /* for pseudo_inits */ +#include /* for pseudo_inits */ #include +#include #include #include #include #include #include +#include #include - -#include -#include +#include /* for thread_resume() */ +#include /* for mcache_init() */ +#include /* for mbinit() */ +#include /* for knote_init() */ +#include /* for eventhandler_init() */ +#include /* for memorystatus_init() */ +#include /* for memorystatus_freeze_init() */ +#include /* for aio_init() */ +#include /* for psem_cache_init() */ +#include /* for dlil_init() */ +#include /* for proto_kpi_init() */ +#include /* for iptap_init() */ +#include /* for socketinit() */ +#include /* for domaininit() */ +#include /* for thread_wakeup() */ +#include /* for ether_family_init() */ +#include /* for gif_init() */ +#include /* for devfs_kernel_mount() */ +#include /* for kmem_suballoc() */ +#include /* for log_setsize() */ +#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 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 restricted_in_port_init() */ +#include /* for assert() */ +#include /* for init_system_override() */ +#include /* for lf_init() */ +#include #include +#if CONFIG_MACF +#include +#include /* mac_init_bsd() */ +#include /* mac_update_task_label() */ +#endif + #include -extern int app_profile; /* on/off switch for pre-heat cache */ +#if CONFIG_NETBOOT +#include +#endif + +#if CONFIG_IMAGEBOOT +#include +#endif + +#if PFLOG +#include +#endif + + +#include +#include +#include + +#if CONFIG_XNUPOST +#include +#endif -char copyright[] = -"Copyright (c) 1982, 1986, 1989, 1991, 1993\n\t" -"The Regents of the University of California. " -"All rights reserved.\n\n"; +void * get_user_regs(thread_t); /* XXX kludge for */ +void IOKitInitializeTime(void); /* XXX */ +void IOSleep(unsigned int); /* XXX */ +void loopattach(void); /* XXX */ -extern void ux_handler(); +const char *const copyright = + "Copyright (c) 1982, 1986, 1989, 1991, 1993\n\t" + "The Regents of the University of California. " + "All rights reserved.\n\n"; /* Components of the first process -- never freed. */ -struct proc proc0; -struct session session0; -struct pgrp pgrp0; -struct filedesc filedesc0; -struct plimit limit0; -struct pstats pstats0; -struct sigacts sigacts0; -struct proc *kernproc, *initproc; +struct proc proc0 = { .p_comm = "kernel_task", .p_name = "kernel_task" }; +struct session session0; +struct pgrp pgrp0; +struct filedesc filedesc0; +struct plimit limit0; +struct pstats pstats0; +struct sigacts sigacts0; +SECURITY_READ_ONLY_LATE(proc_t) kernproc = &proc0; +proc_t XNU_PTRAUTH_SIGNED_PTR("initproc") initproc; long tk_cancc; long tk_nin; @@ -156,62 +229,145 @@ int nswapmap; void *swapmap; struct swdevt swdevt[1]; -dev_t rootdev; /* device of the root */ -dev_t dumpdev; /* device to take dumps on */ -long dumplo; /* offset into dumpdev */ -long hostid; -char hostname[MAXHOSTNAMELEN]; -int hostnamelen; -char domainname[MAXDOMNAMELEN]; -int domainnamelen; -#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 */ - -#ifdef KMEMSTATS -struct kmemstats kmemstats[M_LAST]; -#endif - -int lbolt; /* awoken once a second */ -struct vnode *rootvp; -int boothowto = RB_DEBUG; - -#define BSD_PAGABLE_MAP_SIZE (16 * 512 * 1024) -vm_map_t bsd_pageable_map; -vm_map_t mb_map; -semaphore_t execve_semaphore; - -int cmask = CMASK; +static LCK_GRP_DECLARE(hostname_lck_grp, "hostname"); +LCK_MTX_DECLARE(hostname_lock, &hostname_lck_grp); +LCK_MTX_DECLARE(domainname_lock, &hostname_lck_grp); + +dev_t rootdev; /* device of the root */ +dev_t dumpdev; /* device to take dumps on */ +long dumplo; /* offset into dumpdev */ +long hostid; +char hostname[MAXHOSTNAMELEN]; +char domainname[MAXDOMNAMELEN]; +char rootdevice[DEVMAXNAMESIZE]; + +struct vnode *rootvp; +bool rootvp_is_ssd = false; +int boothowto; +int minimalboot = 0; +#if CONFIG_DARKBOOT +int darkboot = 0; +#endif + +#if __arm64__ +int legacy_footprint_entitlement_mode = LEGACY_FOOTPRINT_ENTITLEMENT_IGNORE; +#endif /* __arm64__ */ + +#if PROC_REF_DEBUG +__private_extern__ int proc_ref_tracking_disabled = 0; /* disable panics on leaked proc refs across syscall boundary */ +#endif + +extern kern_return_t IOFindBSDRoot(char *, unsigned int, dev_t *, u_int32_t *); +extern void IOSecureBSDRoot(const char * rootName); +extern kern_return_t IOKitBSDInit(void ); +extern boolean_t IOSetRecoveryBoot(bsd_bootfail_mode_t, uuid_t, boolean_t); +extern void kminit(void); +extern void bsd_bufferinit(void); +extern void oslog_setsize(int size); +extern void throttle_init(void); + +#if CONFIG_LOCKERBOOT +#define LOCKER_PROTOBOOT_MOUNT "/protoboot" + +const char kernel_protoboot_mount[] = LOCKER_PROTOBOOT_MOUNT; +extern int mount_locker_protoboot(const char *fsname, const char *mntpoint, + const char *pbdevpath); +#endif + +extern int ncl; +#if DEVELOPMENT || DEBUG +extern int syscallfilter_disable; +#endif // DEVELOPMENT || DEBUG + +vm_map_t bsd_pageable_map; +vm_map_t mb_map; + +static int bsd_simul_execs; +static int bsd_pageable_map_size; +__private_extern__ int execargs_cache_size = 0; +__private_extern__ int execargs_free_count = 0; +__private_extern__ vm_offset_t * execargs_cache = NULL; + +void bsd_exec_setup(int); + +__private_extern__ int bootarg_execfailurereports = 0; + +#if __x86_64__ +__private_extern__ int bootarg_no32exec = 1; +#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) */ + +__private_extern__ int bootarg_no_vnode_drain = 0; + +/* + * Prevent kernel-based ASLR from being used, for testing. + */ +#if DEVELOPMENT || DEBUG +__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; +extern uint64_t dyld_flags; +#endif + +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 *); +kern_return_t bsd_autoconf(void); +void bsd_utaskbootstrap(void); + +static void parse_bsd_args(void); +#if CONFIG_DEV_KMEM +extern void dev_kmem_init(void); +#endif +extern void select_waitq_init(void); +static void process_name(const char *, proc_t); static void setconf(void); -funnel_t *kernel_flock; +#if CONFIG_BASESYSTEMROOT +static int bsd_find_basesystem_dmg(char *bsdmgpath_out, bool *rooted_dmg, bool *skip_signature_check); +static boolean_t bsdmgroot_bootable(void); +#endif // CONFIG_BASESYSTEMROOT + +static bool bsd_rooted_ramdisk(void); +#if SYSV_SHM extern void sysv_shm_lock_init(void); +#endif +#if SYSV_SEM extern void sysv_sem_lock_init(void); +#endif +#if SYSV_MSG extern void sysv_msg_lock_init(void); -extern void pshm_lock_init(); -extern void psem_lock_init(); -extern int maxprocperuid; +#endif -/* kmem access not enabled by default; can be changed with boot-args */ -int setup_kmem = 0; +extern void ulock_initialize(void); -extern void stackshot_lock_init(); +#if CONFIG_MACF +#if defined (__i386__) || defined (__x86_64__) +/* MACF policy_check configuration flags; see policy_check.c for details */ +int policy_check_flags = 0; + +extern int check_policy_init(int); +#endif +#endif /* CONFIG_MACF */ + +/* If we are using CONFIG_DTRACE */ +#if CONFIG_DTRACE +extern void dtrace_postinit(void); +#endif /* * Initialization code. @@ -229,34 +385,90 @@ extern void stackshot_lock_init(); * Sets the name for the given task. */ static void -process_name(s, p) - char *s; - struct proc *p; +process_name(const char *s, proc_t p) { - int 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 }; -struct rlimit vm_initial_limit_data = { DFLDSIZ, MAXDSIZ }; -struct rlimit vm_initial_limit_core = { DFLCSIZ, MAXCSIZ }; +struct rlimit vm_initial_limit_stack = { .rlim_cur = DFLSSIZ, .rlim_max = MAXSSIZ - PAGE_MAX_SIZE }; +struct rlimit vm_initial_limit_data = { .rlim_cur = DFLDSIZ, .rlim_max = MAXDSIZ }; +struct rlimit vm_initial_limit_core = { .rlim_cur = DFLCSIZ, .rlim_max = MAXCSIZ }; + +extern struct os_refgrp rlimit_refgrp; -extern thread_t cloneproc(struct proc *, int); -extern int (*mountroot)(void); -extern int netboot_mountroot(); /* netboot.c */ -extern int netboot_setup(struct proc * p); +extern thread_t cloneproc(task_t, coalition_t, proc_t, int, int); +extern int (*mountroot)(void); -lck_grp_t * proc_lck_grp; -lck_grp_attr_t * proc_lck_grp_attr; -lck_attr_t * proc_lck_attr; +LCK_ATTR_DECLARE(proc_lck_attr, 0, 0); +LCK_GRP_DECLARE(proc_lck_grp, "proc"); +LCK_GRP_DECLARE(proc_slock_grp, "proc-slock"); +LCK_GRP_DECLARE(proc_fdmlock_grp, "proc-fdmlock"); +LCK_GRP_DECLARE(proc_mlock_grp, "proc-mlock"); +LCK_GRP_DECLARE(proc_ucred_mlock_grp, "proc-ucred-mlock"); +LCK_GRP_DECLARE(proc_dirslock_grp, "proc-dirslock"); +LCK_GRP_DECLARE(proc_kqhashlock_grp, "proc-kqhashlock"); +LCK_GRP_DECLARE(proc_knhashlock_grp, "proc-knhashlock"); + +LCK_MTX_DECLARE_ATTR(proc_list_mlock, &proc_mlock_grp, &proc_lck_attr); + +#if XNU_TARGET_OS_OSX /* hook called after root is mounted XXX temporary hack */ void (*mountroot_post_hook)(void); +void (*unmountroot_pre_hook)(void); +#endif +void set_rootvnode(vnode_t); + +extern lck_rw_t rootvnode_rw_lock; + +/* called with an iocount and usecount on new_rootvnode */ +void +set_rootvnode(vnode_t new_rootvnode) +{ + mount_t new_mount = (new_rootvnode != NULL) ? new_rootvnode->v_mount : NULL; + vnode_t new_devvp = (new_mount != NULL) ? new_mount->mnt_devvp : NULL; + vnode_t old_rootvnode = rootvnode; + + new_rootvnode->v_flag |= VROOT; + rootvp = new_devvp; + rootvnode = new_rootvnode; + filedesc0.fd_cdir = new_rootvnode; + if (new_devvp != NULL) { + rootdev = vnode_specrdev(new_devvp); + } else if (new_mount != NULL) { + rootdev = vfs_statfs(new_mount)->f_fsid.val[0]; /* like ATTR_CMN_DEVID */ + } else { + rootdev = NODEV; + } + + if (old_rootvnode) { + vnode_rele(old_rootvnode); + } +} + +#define RAMDEV "md0" + +bool +bsd_rooted_ramdisk(void) +{ + bool is_ramdisk = false; + char *dev_path = zalloc(ZV_NAMEI); + if (dev_path == NULL) { + panic("failed to allocate devpath string! \n"); + } + + if (PE_parse_boot_argn("rd", dev_path, MAXPATHLEN)) { + if (strncmp(dev_path, RAMDEV, strlen(RAMDEV)) == 0) { + is_ramdisk = true; + } + } + + zfree(ZV_NAMEI, dev_path); + return is_ramdisk; +} /* * This function is called very early on in the Mach startup, from the @@ -266,153 +478,236 @@ void (*mountroot_post_hook)(void); * 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. + * uu_context 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 contents + * of the uu_context.vc_ucred field so that the uthread structure can be + * used like any other. */ + void -bsd_init() +bsd_init(void) { - register struct proc *p; struct uthread *ut; - extern kauth_cred_t rootcred; - register int i; - int s; - int error; - thread_t th; + unsigned int i; struct vfs_context context; - void lightning_bolt(void ); - kern_return_t ret; - boolean_t funnel_state; + kern_return_t ret; struct ucred temp_cred; - extern void file_lock_init(void); + struct posix_cred temp_pcred; + vnode_t init_rootvnode = NULLVP; +#if CONFIG_NETBOOT || CONFIG_IMAGEBOOT + boolean_t netboot = FALSE; +#endif +#if CONFIG_LOCKERBOOT + vnode_t pbvn = NULLVP; + mount_t pbmnt = NULL; + char *pbdevp = NULL; + char pbdevpath[64]; + char pbfsname[MFSNAMELEN]; + const char *slash_dev = NULL; +#endif - kernel_flock = funnel_alloc(KERNEL_FUNNEL); - if (kernel_flock == (funnel_t *)0 ) { - panic("bsd_init: Failed to allocate kernel funnel"); - } - - funnel_state = thread_funnel_set(kernel_flock, TRUE); +#define DEBUG_BSDINIT 0 + +#if DEBUG_BSDINIT +#define bsd_init_kprintf(x, ...) kprintf("bsd_init: " x, ## __VA_ARGS__) +#else +#define bsd_init_kprintf(x, ...) +#endif + + throttle_init(); printf(copyright); - - kmeminit(); - + + bsd_init_kprintf("calling parse_bsd_args\n"); parse_bsd_args(); - /* Initialize the uthread zone */ - //uthread_zone_init(); /* XXX redundant: previous uthread_alloc() */ +#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(); /* Initialize process and pgrp structures. */ + bsd_init_kprintf("calling procinit\n"); procinit(); - kernproc = &proc0; - - p = kernproc; - /* kernel_task->proc = kernproc; */ - set_bsdtask_info(kernel_task,(void *)p); - p->p_pid = 0; + set_bsdtask_info(kernel_task, (void *)kernproc); /* give kernproc a name */ - 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); + bsd_init_kprintf("calling process_name\n"); + process_name("kernel_task", kernproc); /* 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); + 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); + lck_rw_init(&kernproc->p_dirs_lock, &proc_dirslock_grp, &proc_lck_attr); - if (current_task() != kernel_task) + assert(bsd_simul_execs != 0); + execargs_cache_size = bsd_simul_execs; + execargs_free_count = bsd_simul_execs; + execargs_cache = zalloc_permanent(bsd_simul_execs * sizeof(vm_offset_t), + ZALIGN(vm_offset_t)); + + if (current_task() != kernel_task) { printf("bsd_init: We have a problem, " - "current task is not kernel task\n"); - + "current task is not kernel task\n"); + } + + bsd_init_kprintf("calling get_bsdthread_info\n"); ut = (uthread_t)get_bsdthread_info(current_thread()); +#if CONFIG_MACF + /* + * Initialize the MAC Framework + */ + mac_policy_initbsd(); + +#if defined (__i386__) || defined (__x86_64__) + /* + * We currently only support this on i386/x86_64, as that is the + * only lock code we have instrumented so far. + */ + check_policy_init(policy_check_flags); +#endif +#endif /* MAC */ + + ulock_initialize(); + /* * Create process 0. */ - LIST_INSERT_HEAD(&allproc, p, p_list); - p->p_pgrp = &pgrp0; + proc_list_lock(); + LIST_INSERT_HEAD(&allproc, kernproc, p_list); + kernproc->p_pgrp = &pgrp0; LIST_INSERT_HEAD(PGRPHASH(0), &pgrp0, pg_hash); LIST_INIT(&pgrp0.pg_members); - LIST_INSERT_HEAD(&pgrp0.pg_members, p, p_pglist); + lck_mtx_init(&pgrp0.pg_mlock, &proc_mlock_grp, &proc_lck_attr); + /* There is no other bsd thread this point and is safe without pgrp lock */ + LIST_INSERT_HEAD(&pgrp0.pg_members, kernproc, p_pglist); + kernproc->p_listflag |= P_LIST_INPGRP; + kernproc->p_pgrpid = 0; + kernproc->p_uniqueid = 0; pgrp0.pg_session = &session0; + pgrp0.pg_membercnt = 1; + session0.s_count = 1; - session0.s_leader = p; - - p->task = kernel_task; - - p->p_stat = SRUN; - p->p_flag = P_SYSTEM; - p->p_nice = NZERO; - p->p_pptr = p; - lockinit(&p->signal_lock, PVM, "signal", 0, 0); - TAILQ_INIT(&p->p_uthlist); - p->sigwait = FALSE; - p->sigwait_thread = THREAD_NULL; - p->exit_thread = THREAD_NULL; + session0.s_leader = kernproc; + session0.s_listflags = 0; + lck_mtx_init(&session0.s_mlock, &proc_mlock_grp, &proc_lck_attr); + LIST_INSERT_HEAD(SESSHASH(0), &session0, s_hash); + proc_list_unlock(); + +#if CONFIG_PERSONAS + kernproc->p_persona = NULL; +#endif + + kernproc->task = kernel_task; + + kernproc->p_stat = SRUN; + 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; + } +#endif + + kernproc->p_nice = NZERO; + kernproc->p_pptr = kernproc; + + TAILQ_INIT(&kernproc->p_uthlist); + TAILQ_INSERT_TAIL(&kernproc->p_uthlist, ut, uu_list); + + kernproc->sigwait = FALSE; + kernproc->sigwait_thread = THREAD_NULL; + kernproc->exit_thread = THREAD_NULL; + kernproc->p_csflags = CS_VALID; /* * 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.. */ + bsd_init_kprintf("calling bzero\n"); bzero(&temp_cred, sizeof(temp_cred)); - temp_cred.cr_ngroups = 1; + bzero(&temp_pcred, sizeof(temp_pcred)); + temp_pcred.cr_ngroups = 1; + /* kern_proc, shouldn't call up to DS for group membership */ + temp_pcred.cr_flags = CRF_NOMEMBERD; + temp_cred.cr_audit.as_aia_p = audit_default_aia_p; + + bsd_init_kprintf("calling kauth_cred_create\n"); + /* + * We have to label the temp cred before we create from it to + * properly set cr_ngroups, or the create will fail. + */ + posix_cred_label(&temp_cred, &temp_pcred); + kernproc->p_ucred = kauth_cred_create(&temp_cred); - p->p_ucred = kauth_cred_create(&temp_cred); + /* update cred on proc */ + PROC_UPDATE_CREDS_ONPROC(kernproc); /* 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; + bsd_init_kprintf("calling kauth_cred_ref\n"); + kauth_cred_ref(kernproc->p_ucred); + ut->uu_context.vc_ucred = kernproc->p_ucred; + ut->uu_context.vc_thread = current_thread(); - file_lock_init(); + vfs_set_context_kernel(&ut->uu_context); + + TAILQ_INIT(&kernproc->p_aio_activeq); + TAILQ_INIT(&kernproc->p_aio_doneq); + kernproc->p_aio_total_count = 0; + +#if CONFIG_MACF + mac_cred_label_associate_kernel(kernproc->p_ucred); +#endif /* Create the file descriptor table. */ - filedesc0.fd_refcnt = 1+1; /* +1 so shutdown will not _FREE_ZONE */ - p->p_fd = &filedesc0; - filedesc0.fd_cmask = cmask; - filedesc0.fd_knlistsize = -1; + kernproc->p_fd = &filedesc0; + filedesc0.fd_cmask = (mode_t)cmask; + filedesc0.fd_knlistsize = 0; 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. */ - p->p_limit = &limit0; - for (i = 0; i < sizeof(p->p_rlimit)/sizeof(p->p_rlimit[0]); i++) - limit0.pl_rlimit[i].rlim_cur = - limit0.pl_rlimit[i].rlim_max = RLIM_INFINITY; + kernproc->p_limit = &limit0; + for (i = 0; i < sizeof(kernproc->p_limit->pl_rlimit) / sizeof(kernproc->p_limit->pl_rlimit[0]); i++) { + 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 = 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; limit0.pl_rlimit[RLIMIT_CORE] = vm_initial_limit_core; - limit0.p_refcnt = 1; + os_ref_init_count(&limit0.pl_refcnt, &rlimit_refgrp, 1); - p->p_stats = &pstats0; - p->p_sigacts = &sigacts0; + kernproc->p_stats = &pstats0; + kernproc->p_sigacts = &sigacts0; + kernproc->p_subsystem_root_path = NULL; /* - * 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); /* @@ -420,16 +715,21 @@ bsd_init() * for temporary copying (execve()). */ { - vm_offset_t min; + vm_offset_t minimum; + bsd_init_kprintf("calling kmem_suballoc\n"); + assert(bsd_pageable_map_size != 0); ret = kmem_suballoc(kernel_map, - &min, - (vm_size_t)BSD_PAGABLE_MAP_SIZE, - TRUE, - VM_FLAGS_ANYWHERE, - &bsd_pageable_map); - if (ret != KERN_SUCCESS) + &minimum, + (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"); + } } /* @@ -439,83 +739,127 @@ bsd_init() * happen after a credential has been associated with * the kernel task. */ + bsd_init_kprintf("calling bsd_bufferinit\n"); bsd_bufferinit(); - /* Initialize the execve() semaphore */ - ret = semaphore_create(kernel_task, &execve_semaphore, - SYNC_POLICY_FIFO, (BSD_PAGABLE_MAP_SIZE / NCARGS)); - if (ret != KERN_SUCCESS) - panic("bsd_init: Failed to create execve semaphore"); - /* * Initialize the calendar. */ - IOKitResetTime(); - - ubc_init(); + bsd_init_kprintf("calling IOKitInitializeTime\n"); + IOKitInitializeTime(); /* Initialize the file systems. */ + bsd_init_kprintf("calling vfsinit\n"); vfsinit(); +#if CONFIG_PROC_UUID_POLICY + /* Initial proc_uuid_policy subsystem */ + bsd_init_kprintf("calling proc_uuid_policy_init()\n"); + proc_uuid_policy_init(); +#endif + +#if SOCKETS + /* Initialize per-CPU cache allocator */ + mcache_init(); + /* Initialize mbuf's. */ + bsd_init_kprintf("calling mbinit\n"); mbinit(); + net_str_id_init(); /* for mbuf tags */ + restricted_in_port_init(); +#endif /* SOCKETS */ /* * Initializes security event auditing. * XXX: Should/could this occur later? */ - audit_init(); +#if CONFIG_AUDIT + bsd_init_kprintf("calling audit_init\n"); + audit_init(); +#endif /* Initialize kqueues */ + 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(); - /* 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(); - + pthread_init(); /* POSIX Shm and Sem */ + bsd_init_kprintf("calling pshm_cache_init\n"); pshm_cache_init(); + bsd_init_kprintf("calling psem_cache_init\n"); psem_cache_init(); - time_zone_slock_init(); + bsd_init_kprintf("calling select_waitq_init\n"); + select_waitq_init(); - /* Stack snapshot facility lock */ - stackshot_lock_init(); /* * Initialize protocols. Block reception of incoming packets * until everything is ready. */ - sysctl_register_fixed(); - 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"); proto_kpi_init(); +#endif /* NETWORKING */ +#if SOCKETS + bsd_init_kprintf("calling socketinit\n"); socketinit(); + bsd_init_kprintf("calling domaininit\n"); domaininit(); - - p->p_fd->fd_cdir = NULL; - p->p_fd->fd_rdir = NULL; - -#ifdef GPROF - /* Initialize kernel profiling. */ - kmstartup(); + iptap_init(); +#if FLOW_DIVERT + flow_divert_init(); +#endif /* FLOW_DIVERT */ +#endif /* SOCKETS */ +#if NETWORKING +#if NECP + /* Initialize Network Extension Control Policies */ + necp_init(); +#endif + netagent_init(); +#endif /* NETWORKING */ + kernproc->p_fd->fd_cdir = NULL; + kernproc->p_fd->fd_rdir = NULL; + +#if defined (__x86_64__) && (DEBUG || DEVELOPMENT) + hvg_bsd_init(); +#endif /* DEBUG || DEVELOPMENT */ + +#if CONFIG_FREEZE +#ifndef CONFIG_MEMORYSTATUS + #error "CONFIG_FREEZE defined without matching CONFIG_MEMORYSTATUS" +#endif + /* Initialise background freezing */ + bsd_init_kprintf("calling memorystatus_freeze_init\n"); + memorystatus_freeze_init(); #endif - /* kick off timeout driven events by calling first time */ - thread_wakeup(&lbolt); - timeout((void (*)(void *))lightning_bolt, 0, hz); +#if CONFIG_MEMORYSTATUS + /* Initialize kernel memory status notifications */ + bsd_init_kprintf("calling memorystatus_init\n"); + memorystatus_init(); +#endif /* CONFIG_MEMORYSTATUS */ + bsd_init_kprintf("calling sysctl_mib_init\n"); + sysctl_mib_init(); + + bsd_init_kprintf("calling bsd_autoconf\n"); bsd_autoconf(); +#if CONFIG_DTRACE + dtrace_postinit(); +#endif + /* * We attach the loopback interface *way* down here to ensure * it happens after autoconf(), otherwise it becomes the @@ -523,180 +867,346 @@ bsd_init() */ #include #if NLOOP > 0 - loopattach(); /* XXX */ + bsd_init_kprintf("calling loopattach\n"); + loopattach(); /* XXX */ +#endif +#if NGIF + /* Initialize gif interface (after lo0) */ + gif_init(); #endif - - /* Register the built-in dlil ethernet interface family */ + +#if PFLOG + /* Initialize packet filter log interface */ + pfloginit(); +#endif /* PFLOG */ + +#if NETHER > 0 + /* Register the built-in dlil ethernet interface family */ + bsd_init_kprintf("calling ether_family_init\n"); ether_family_init(); +#endif /* ETHER */ - /* Call any kext code that wants to run just after network init */ - net_init_run(); +#if NETWORKING +#if CONTENT_FILTER + cfil_init(); +#endif - vnode_pager_bootstrap(); -#if 0 - /* XXX Hack for early debug stop */ - printf("\nabout to sleep for 10 seconds\n"); - IOSleep( 10 * 1000 ); - /* Debugger("hello"); */ +#if PACKET_MANGLER + pkt_mnglr_init(); #endif + /* + * Register subsystems with kernel control handlers + */ + utun_register_control(); +#if IPSEC + ipsec_register_control(); +#endif /* IPSEC */ + netsrc_init(); + nstat_init(); + tcp_cc_init(); +#if MPTCP + mptcp_control_register(); +#endif /* MPTCP */ + + /* + * The the networking stack is now initialized so it is a good time to call + * the clients that are waiting for the networking stack to be usable. + */ + bsd_init_kprintf("calling net_init_run\n"); + net_init_run(); +#endif /* NETWORKING */ + + bsd_init_kprintf("calling inittodr\n"); inittodr(0); /* Mount the root file system. */ - while( TRUE) { + while (TRUE) { int err; + bsd_init_kprintf("calling setconf\n"); setconf(); - bsd_hardclockinit = -1; /* start ticking */ +#if CONFIG_NETBOOT + netboot = (mountroot == netboot_mountroot); +#endif - if (0 == (err = vfs_mountroot())) + bsd_init_kprintf("vfs_mountroot\n"); + 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; + } + rootdevice[0] = '\0'; +#if CONFIG_NETBOOT + if (netboot) { + PE_display_icon( 0, "noroot"); /* XXX a netboot-specific icon would be nicer */ + vc_progress_set(FALSE, 0); + for (i = 1; 1; i *= 2) { + printf("bsd_init: failed to mount network root, error %d, %s\n", + err, PE_boot_args()); + printf("We are hanging here...\n"); + IOSleep(i * 60 * 1000); + } + /*NOTREACHED*/ } #endif printf("cannot mount root, errno = %d\n", err); boothowto |= RB_ASKNAME; } - context.vc_proc = p; - context.vc_ucred = p->p_ucred; + IOSecureBSDRoot(rootdevice); + + context.vc_thread = current_thread(); + context.vc_ucred = kernproc->p_ucred; mountlist.tqh_first->mnt_flag |= MNT_ROOTFS; + bsd_init_kprintf("calling VFS_ROOT\n"); /* Get the vnode for '/'. Set fdp->fd_fd.fd_cdir to reference it. */ - if (VFS_ROOT(mountlist.tqh_first, &rootvnode, &context)) - panic("bsd_init: cannot find root vnode"); - rootvnode->v_flag |= VROOT; - (void)vnode_ref(rootvnode); - (void)vnode_put(rootvnode); - filedesc0.fd_cdir = rootvnode; - -#if NFSCLIENT - if (mountroot == netboot_mountroot) { + if (VFS_ROOT(mountlist.tqh_first, &init_rootvnode, &context)) { + panic("bsd_init: cannot find root vnode: %s", PE_boot_args()); + } + (void)vnode_ref(init_rootvnode); + (void)vnode_put(init_rootvnode); + + lck_rw_lock_exclusive(&rootvnode_rw_lock); + set_rootvnode(init_rootvnode); + lck_rw_unlock_exclusive(&rootvnode_rw_lock); + init_rootvnode = NULLVP; /* use rootvnode after this point */ + + + if (!bsd_rooted_ramdisk()) { +#if CONFIG_IMAGEBOOT +#if XNU_TARGET_OS_OSX && defined(__arm64__) + /* Apple Silicon MacOS */ + if (!imageboot_desired()) { + /* enforce sealedness */ + int autherr = VNOP_IOCTL(rootvnode, FSIOC_KERNEL_ROOTAUTH, NULL, 0, vfs_context_kernel()); + if (autherr) { + panic("rootvp not authenticated after mounting \n"); + } + } +#endif // TARGET_OS_OSX && arm64 +#endif // config_imageboot + /* Otherwise, noop */ + } + + +#if CONFIG_NETBOOT + if (netboot) { int err; + + netboot = TRUE; /* post mount setup */ - if (err = netboot_setup(p)) { - panic("bsd_init: NetBoot could not find root, %d", err); + if ((err = netboot_setup()) != 0) { + PE_display_icon( 0, "noroot"); /* XXX a netboot-specific icon would be nicer */ + vc_progress_set(FALSE, 0); + for (i = 1; 1; i *= 2) { + printf("bsd_init: NetBoot could not find root, error %d: %s\n", + err, PE_boot_args()); + printf("We are hanging here...\n"); + IOSleep(i * 60 * 1000); + } + /*NOTREACHED*/ } } #endif - - microtime(&p->p_stats->p_start); - p->p_rtime.tv_sec = p->p_rtime.tv_usec = 0; + +#if CONFIG_IMAGEBOOT +#if CONFIG_LOCKERBOOT + /* + * Stash the protoboot vnode, mount, filesystem name, and device name for + * later use. Note that the mount-from name may not have the "/dev/" + * component, so we must sniff out this condition and add it as needed. + */ + pbvn = rootvnode; + pbmnt = pbvn->v_mount; + pbdevp = vfs_statfs(pbmnt)->f_mntfromname; + slash_dev = strnstr(pbdevp, "/dev/", strlen(pbdevp)); + if (slash_dev) { + /* + * If the old root is a snapshot mount, it will have the form: + * + * com.apple.os.update-@ + * + * So we just search the mntfromname for any occurrence of "/dev/" and + * grab that as the device path. The image boot code needs a dev node to + * do the re-mount, so we cannot directly mount the snapshot as the + * protoboot volume currently. + */ + strlcpy(pbdevpath, slash_dev, sizeof(pbdevpath)); + } else { + snprintf(pbdevpath, sizeof(pbdevpath), "/dev/%s", pbdevp); + } + + bsd_init_kprintf("protoboot mount-from: %s\n", pbdevp); + bsd_init_kprintf("protoboot dev path: %s\n", pbdevpath); + + strlcpy(pbfsname, pbmnt->mnt_vtable->vfc_name, sizeof(pbfsname)); +#endif + + /* + * See if a system disk image is present. If so, mount it and + * switch the root vnode to point to it + */ + imageboot_type_t imageboot_type = imageboot_needed(); + if (netboot == FALSE && imageboot_type) { + /* + * An image was found. No turning back: we're booted + * with a kernel from the disk image. + */ + bsd_init_kprintf("doing image boot: type = %d\n", imageboot_type); + imageboot_setup(imageboot_type); + } + +#if CONFIG_LOCKERBOOT + if (imageboot_type == IMAGEBOOT_LOCKER) { + bsd_init_kprintf("booting from locker\n"); + if (vnode_tag(rootvnode) != VT_LOCKERFS) { + panic("root filesystem not a locker: fsname = %s", + rootvnode->v_mount->mnt_vtable->vfc_name); + } + } +#endif /* CONFIG_LOCKERBOOT */ +#endif /* CONFIG_IMAGEBOOT */ + + /* set initial time; all other resource data is already zero'ed */ + microtime_with_abstime(&kernproc->p_start, &kernproc->p_stats->ps_start); #if DEVFS { - extern void devfs_kernel_mount(char * str); - - devfs_kernel_mount("/dev"); + char mounthere[] = "/dev"; /* !const because of internal casting */ + + bsd_init_kprintf("calling devfs_kernel_mount\n"); + devfs_kernel_mount(mounthere); } #endif /* DEVFS */ - + +#if CONFIG_BASESYSTEMROOT +#if CONFIG_IMAGEBOOT + if (bsdmgroot_bootable()) { + int error; + bool rooted_dmg = false; + bool skip_signature_check = false; + + printf("trying to find and mount BaseSystem dmg as root volume\n"); +#if DEVELOPMENT || DEBUG + printf("(set boot-arg -nobsdmgroot to avoid this)\n"); +#endif // DEVELOPMENT || DEBUG + + char *dmgpath = NULL; + dmgpath = zalloc_flags(ZV_NAMEI, Z_ZERO | Z_WAITOK); + if (dmgpath == NULL) { + panic("%s: M_NAMEI zone exhausted", __FUNCTION__); + } + + error = bsd_find_basesystem_dmg(dmgpath, &rooted_dmg, &skip_signature_check); + if (error) { + bsd_init_kprintf("failed to to find BaseSystem dmg: error = %d\n", error); + } else { + PE_parse_boot_argn("bsdmgpath", dmgpath, sizeof(dmgpath)); + + bsd_init_kprintf("found BaseSystem dmg at: %s\n", dmgpath); + + error = imageboot_pivot_image(dmgpath, IMAGEBOOT_DMG, "/System/Volumes/BaseSystem", "System/Volumes/macOS", rooted_dmg, skip_signature_check); + if (error) { + bsd_init_kprintf("couldn't mount BaseSystem dmg: error = %d", error); + } + } + zfree(ZV_NAMEI, dmgpath); + } +#else /* CONFIG_IMAGEBOOT */ +#error CONFIG_BASESYSTEMROOT requires CONFIG_IMAGEBOOT +#endif /* CONFIG_IMAGEBOOT */ +#endif /* CONFIG_BASESYSTEMROOT */ + +#if CONFIG_LOCKERBOOT + /* + * We need to wait until devfs is up before remounting the protoboot volume + * within the locker so that it can have a real devfs vnode backing it. + */ + if (imageboot_type == IMAGEBOOT_LOCKER) { + bsd_init_kprintf("re-mounting protoboot volume\n"); + int error = mount_locker_protoboot(pbfsname, LOCKER_PROTOBOOT_MOUNT, + pbdevpath); + if (error) { + panic("failed to mount protoboot volume: dev path = %s, error = %d", + pbdevpath, error); + } + } +#endif /* CONFIG_LOCKERBOOT */ + /* Initialize signal state for process 0. */ - siginit(p); + bsd_init_kprintf("calling siginit\n"); + siginit(kernproc); + bsd_init_kprintf("calling bsd_utaskbootstrap\n"); 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 - + pal_kernel_announce(); + + bsd_init_kprintf("calling mountroot_post_hook\n"); + +#if XNU_TARGET_OS_OSX /* invoke post-root-mount hook */ - if (mountroot_post_hook != NULL) + if (mountroot_post_hook != NULL) { mountroot_post_hook(); - - (void) thread_funnel_set(kernel_flock, funnel_state); + } +#endif + +#if 0 /* not yet */ + consider_zone_gc(FALSE); +#endif + + bsd_init_kprintf("done\n"); } -/* Called with kernel funnel held */ void bsdinit_task(void) { - struct proc *p = current_proc(); - struct uthread *ut; - kern_return_t kr; - thread_t th_act; - shared_region_mapping_t system_region; + proc_t p = current_proc(); process_name("init", p); - ux_handler_init(); - - th_act = current_thread(); - (void) host_set_exception_ports(host_priv_self(), - EXC_MASK_ALL & ~(EXC_MASK_SYSCALL | - EXC_MASK_MACH_SYSCALL | - EXC_MASK_RPC_ALERT), - ux_exception_port, - EXCEPTION_DEFAULT, 0); - - (void) task_set_exception_ports(get_threadtask(th_act), - EXC_MASK_ALL & ~(EXC_MASK_SYSCALL | - EXC_MASK_MACH_SYSCALL | - EXC_MASK_RPC_ALERT), - ux_exception_port, - EXCEPTION_DEFAULT, 0); + /* Set up exception-to-signal reflection */ + ux_handler_setup(); +#if CONFIG_MACF + mac_cred_label_associate_user(p->p_ucred); +#endif + vm_init_before_launchd(); +#if CONFIG_XNUPOST + int result = bsd_list_tests(); + result = bsd_do_post(); + if (result != 0) { + panic("bsd_do_post: Tests failed with result = 0x%08x\n", result); + } +#endif - ut = (uthread_t)get_bsdthread_info(th_act); + bsd_init_kprintf("bsd_do_post - done"); - 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, cpu_type()); - if (system_region == NULL) { - 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 -lightning_bolt() -{ - boolean_t funnel_state; - extern void klogwakeup(void); - - funnel_state = thread_funnel_set(kernel_flock, TRUE); - - thread_wakeup(&lbolt); - timeout(lightning_bolt,0,hz); - klogwakeup(); - - (void) thread_funnel_set(kernel_flock, FALSE); -} - -bsd_autoconf() +kern_return_t +bsd_autoconf(void) { - extern kern_return_t IOKitBSDInit( void ); - + kprintf("bsd_autoconf: calling kminit\n"); kminit(); - /* + /* * Early startup for bsd pseudodevices. */ { - struct pseudo_init *pi; - - for (pi = pseudo_inits; pi->ps_func; pi++) - (*pi->ps_func) (pi->ps_count); + struct pseudo_init *pi; + + for (pi = pseudo_inits; pi->ps_func; pi++) { + (*pi->ps_func)(pi->ps_count); + } } - return( IOKitBSDInit()); + return IOKitBSDInit(); } @@ -704,131 +1214,553 @@ bsd_autoconf() static void setconf(void) -{ - extern kern_return_t IOFindBSDRoot( char * rootName, - dev_t * root, u_int32_t * flags ); - u_int32_t flags; - kern_return_t err; +{ + 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 - */ - thread_funnel_set(kernel_flock, FALSE); - err = IOFindBSDRoot( rootdevice, &rootdev, &flags ); - thread_funnel_set(kernel_flock, TRUE); - if( err) { + err = IOFindBSDRoot(rootdevice, sizeof(rootdevice), &rootdev, &flags); + if (err) { printf("setconf: IOFindBSDRoot returned an error (%d);" - "setting rootdevice to 'sd0a'.\n", err); /* XXX DEBUG TEMP */ + "setting rootdevice to 'sd0a'.\n", err); /* XXX DEBUG TEMP */ rootdev = makedev( 6, 0 ); - strcpy( rootdevice, "sd0a" ); + strlcpy(rootdevice, "sd0a", sizeof(rootdevice)); flags = 0; } -#if NFSCLIENT - if( flags & 1 ) { +#if CONFIG_NETBOOT + if (flags & 1) { /* network device */ mountroot = netboot_mountroot; } else { #endif - /* otherwise have vfs determine root filesystem */ - mountroot = NULL; -#if NFSCLIENT - } + /* otherwise have vfs determine root filesystem */ + mountroot = NULL; +#if CONFIG_NETBOOT +} #endif +} +/* + * Boot into the flavor of Recovery dictated by `mode`. + */ +boolean_t +bsd_boot_to_recovery(bsd_bootfail_mode_t mode, uuid_t volume_uuid, boolean_t reboot) +{ + return IOSetRecoveryBoot(mode, volume_uuid, reboot); } -bsd_utaskbootstrap() +void +bsd_utaskbootstrap(void) { - thread_t th_act; + thread_t thread; struct uthread *ut; - th_act = cloneproc(kernproc, 0); - initproc = pfind(1); - /* Set the launch time for init */ - microtime(&initproc->p_stats->p_start); + /* + * Clone the bootstrap process from the kernel process, without + * inheriting either task characteristics or memory from the kernel; + */ + thread = cloneproc(TASK_NULL, COALITION_NULL, kernproc, FALSE, TRUE); + + /* Hold the reference as it will be dropped during shutdown */ + initproc = proc_find(1); +#if __PROC_INTERNAL_DEBUG + if (initproc == PROC_NULL) { + panic("bsd_utaskbootstrap: initproc not set\n"); + } +#endif + + zalloc_first_proc_made(); + + /* + * Since we aren't going back out the normal way to our parent, + * we have to drop the transition locks explicitly. + */ + proc_signalend(initproc, 0); + proc_transend(initproc, 0); - ut = (struct uthread *)get_bsdthread_info(th_act); + ut = (struct uthread *)get_bsdthread_info(thread); ut->uu_sigmask = 0; - act_set_astbsd(th_act); - (void) thread_resume(th_act); + act_set_astbsd(thread); + task_clear_return_wait(get_threadtask(thread), TCRW_CLEAR_ALL_WAIT); } -parse_bsd_args() +static void +parse_bsd_args(void) { - extern char init_args[]; - char namep[16]; - extern int boothowto; - extern int srv; - extern int ncl; - - int len; + char namep[48]; + int msgbuf; - if (PE_parse_boot_arg("-s", namep)) { + if (PE_parse_boot_argn("-s", namep, sizeof(namep))) { boothowto |= RB_SINGLE; - len = strlen(init_args); - if(len != 0) - strcat(init_args," -s"); - else - strcat(init_args,"-s"); } - if (PE_parse_boot_arg("-b", namep)) { - boothowto |= RB_NOBOOTRC; - len = strlen(init_args); - if(len != 0) - strcat(init_args," -b"); - else - strcat(init_args,"-b"); + if (PE_parse_boot_argn("-x", namep, sizeof(namep))) { /* safe boot */ + boothowto |= RB_SAFEBOOT; } - if (PE_parse_boot_arg("-F", namep)) { - len = strlen(init_args); - if(len != 0) - strcat(init_args," -F"); - else - strcat(init_args,"-F"); + 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 (PE_parse_boot_arg("-v", namep)) { - len = strlen(init_args); - if(len != 0) - strcat(init_args," -v"); - else - strcat(init_args,"-v"); +#if __x86_64__ + int no32exec; + + /* disable 32 bit grading */ + if (PE_parse_boot_argn("no32exec", &no32exec, sizeof(no32exec))) { + bootarg_no32exec = !!no32exec; + } +#endif + + int execfailure_crashreports; + /* enable crash reports on various exec failures */ + if (PE_parse_boot_argn("execfailurecrashes", &execfailure_crashreports, sizeof(execfailure_crashreports))) { + bootarg_execfailurereports = !!execfailure_crashreports; } - if (PE_parse_boot_arg("-x", namep)) { /* safe boot */ - len = strlen(init_args); - if(len != 0) - strcat(init_args," -x"); - else - strcat(init_args,"-x"); + /* 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; } - if (PE_parse_boot_arg("-d", namep)) { - len = strlen(init_args); - if(len != 0) - strcat(init_args," -d"); - else - strcat(init_args,"-d"); +#if DEVELOPMENT || DEBUG + if (PE_parse_boot_argn("-disable_aslr", namep, sizeof(namep))) { + bootarg_disable_aslr = 1; } +#endif + - PE_parse_boot_arg("srv", &srv); - PE_parse_boot_arg("ncl", &ncl); - if (PE_parse_boot_arg("nbuf", &max_nbuf_headers)) + + PE_parse_boot_argn("ncl", &ncl, sizeof(ncl)); + if (PE_parse_boot_argn("nbuf", &max_nbuf_headers, + sizeof(max_nbuf_headers))) { customnbuf = 1; - PE_parse_boot_arg("kmem", &setup_kmem); + } - return 0; +#if CONFIG_MACF +#if defined (__i386__) || defined (__x86_64__) + PE_parse_boot_argn("policy_check", &policy_check_flags, sizeof(policy_check_flags)); +#endif +#endif /* CONFIG_MACF */ + + 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 (PE_parse_boot_argn("-no_vnode_drain", namep, sizeof(namep))) { + bootarg_no_vnode_drain = 1; + } + +#if CONFIG_DARKBOOT + /* + * 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 + + 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; + } + + if (PE_parse_boot_argn("alt-dyld", dyld_alt_path, sizeof(dyld_alt_path))) { + if (strlen(dyld_alt_path) > 0) { + use_alt_dyld = 1; + } + } + PE_parse_boot_argn("dyld_flags", &dyld_flags, sizeof(dyld_flags)); + + if (PE_parse_boot_argn("-disable_syscallfilter", &namep, sizeof(namep))) { + syscallfilter_disable = 1; + } + +#if __arm64__ + if (PE_parse_boot_argn("legacy_footprint_entitlement_mode", &legacy_footprint_entitlement_mode, sizeof(legacy_footprint_entitlement_mode))) { + /* + * legacy_footprint_entitlement_mode specifies the behavior we want associated + * with the entitlement. The supported modes are: + * + * LEGACY_FOOTPRINT_ENTITLEMENT_IGNORE: + * Indicates that we want every process to have the memory accounting + * that is available in iOS 12.0 and beyond. + * + * LEGACY_FOOTPRINT_ENTITLEMENT_IOS11_ACCT: + * Indicates that for every process that has the 'legacy footprint entitlement', + * we want to give it the old iOS 11.0 accounting behavior which accounted some + * of the process's memory to the kernel. + * + * LEGACY_FOOTPRINT_ENTITLEMENT_LIMIT_INCREASE: + * Indicates that for every process that has the 'legacy footprint entitlement', + * we want it to have a higher memory limit which will help them acclimate to the + * iOS 12.0 (& beyond) accounting behavior that does the right accounting. + * The bonus added to the system-wide task limit to calculate this higher memory limit + * is available in legacy_footprint_bonus_mb. + */ + + if (legacy_footprint_entitlement_mode < LEGACY_FOOTPRINT_ENTITLEMENT_IGNORE || + legacy_footprint_entitlement_mode > LEGACY_FOOTPRINT_ENTITLEMENT_LIMIT_INCREASE) { + legacy_footprint_entitlement_mode = LEGACY_FOOTPRINT_ENTITLEMENT_LIMIT_INCREASE; + } + } +#endif /* __arm64__ */ +#endif /* DEVELOPMENT || DEBUG */ +} + +#if CONFIG_BASESYSTEMROOT + +extern const char* IOGetBootUUID(void); +extern const char* IOGetApfsPrebootUUID(void); + +// Get the UUID of the Preboot (and Recovery) folder associated with the +// current boot volume, if applicable. The meaning of the UUID can be +// filesystem-dependent and not all kinds of boots will have a UUID. +// If available, the string will be returned. It does not need to be +// deallocate. (Future: if we need to return the string as a copy that the +// caller must free, we'll introduce a new functcion for that.) +// NULL will be returned if the current boot has no applicable Preboot UUID. +static +const char * +get_preboot_uuid(void) +{ + const char *maybe_uuid_string; + + // try IOGetApfsPrebootUUID + maybe_uuid_string = IOGetApfsPrebootUUID(); + if (maybe_uuid_string) { + uuid_t maybe_uuid; + int error = uuid_parse(maybe_uuid_string, maybe_uuid); + if (error == 0) { + return maybe_uuid_string; + } + } + + // try IOGetBootUUID + maybe_uuid_string = IOGetBootUUID(); + if (maybe_uuid_string) { + uuid_t maybe_uuid; + int error = uuid_parse(maybe_uuid_string, maybe_uuid); + if (error == 0) { + return maybe_uuid_string; + } + } + + // didn't find it + return NULL; +} + +#if defined(__arm64__) +extern const char *IOGetBootObjectsPath(void); +#endif + +// Find the BaseSystem.dmg to be used as the initial root volume during certain +// kinds of boots. +// This may mount volumes and lookup vnodes. +// The DEVELOPMENT kernel will look for BaseSystem.rooted.dmg first. +// If it returns 0 (no error), then it also writes the absolute path to the +// BaseSystem.dmg into its argument (which must be a char[MAXPATHLEN]). +static +int +bsd_find_basesystem_dmg(char *bsdmgpath_out, bool *rooted_dmg, bool *skip_signature_check) +{ + int error; + size_t len; + char *dmgbasepath; + char *dmgpath; + bool allow_rooted_dmg = false; + + dmgbasepath = zalloc_flags(ZV_NAMEI, Z_ZERO | Z_WAITOK); + dmgpath = zalloc_flags(ZV_NAMEI, Z_ZERO | Z_WAITOK); + vnode_t imagevp = NULLVP; + +#if DEVELOPMENT || DEBUG + allow_rooted_dmg = true; +#endif + + //must provide output bool + if (rooted_dmg && skip_signature_check) { + *rooted_dmg = false; + *skip_signature_check = false; + } else { + error = EINVAL; + goto done; + } + + error = vfs_mount_recovery(); + if (error) { + goto done; + } + + len = strlcpy(dmgbasepath, "/System/Volumes/Recovery/", MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + if (csr_check(CSR_ALLOW_ANY_RECOVERY_OS) == 0) { + *skip_signature_check = true; + allow_rooted_dmg = true; + } + +#if defined(__arm64__) + const char *boot_obj_path = IOGetBootObjectsPath(); + if (boot_obj_path) { + if (boot_obj_path[0] == '/') { + dmgbasepath[len - 1] = '\0'; + } + + len = strlcat(dmgbasepath, boot_obj_path, MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + len = strlcat(dmgbasepath, "/usr/standalone/firmware/", MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + if (allow_rooted_dmg) { + len = strlcpy(dmgpath, dmgbasepath, MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + len = strlcat(dmgpath, "arm64eBaseSystem.rooted.dmg", MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + error = vnode_lookup(dmgpath, 0, &imagevp, vfs_context_kernel()); + if (error == 0) { + *rooted_dmg = true; + *skip_signature_check = true; + goto done; + } + memset(dmgpath, 0, MAXPATHLEN); + } + + len = strlcpy(dmgpath, dmgbasepath, MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + len = strlcat(dmgpath, "arm64eBaseSystem.dmg", MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + error = vnode_lookup(dmgpath, 0, &imagevp, vfs_context_kernel()); + if (error == 0) { + goto done; + } + memset(dmgpath, 0, MAXPATHLEN); + dmgbasepath[strlen("/System/Volumes/Recovery/")] = '\0'; + } +#endif // __arm64__ + + const char *preboot_uuid = get_preboot_uuid(); + if (preboot_uuid == NULL) { + // no preboot? bail out + return EINVAL; + } + + len = strlcat(dmgbasepath, preboot_uuid, MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + if (allow_rooted_dmg) { + // Try BaseSystem.rooted.dmg + len = strlcpy(dmgpath, dmgbasepath, MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + len = strlcat(dmgpath, "/BaseSystem.rooted.dmg", MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + error = vnode_lookup(dmgpath, 0, &imagevp, vfs_context_kernel()); + if (error == 0) { + // we found it! success! + *rooted_dmg = true; + *skip_signature_check = true; + goto done; + } + } + + // Try BaseSystem.dmg + len = strlcpy(dmgpath, dmgbasepath, MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + len = strlcat(dmgpath, "/BaseSystem.dmg", MAXPATHLEN); + if (len > MAXPATHLEN) { + error = ENAMETOOLONG; + goto done; + } + + error = vnode_lookup(dmgpath, 0, &imagevp, vfs_context_kernel()); + if (error == 0) { + // success! + goto done; + } + +done: + if (error == 0) { + strlcpy(bsdmgpath_out, dmgpath, MAXPATHLEN); + } else { + bsd_init_kprintf("%s: error %d\n", __func__, error); + } + if (imagevp != NULLVP) { + vnode_put(imagevp); + } + zfree(ZV_NAMEI, dmgpath); + zfree(ZV_NAMEI, dmgbasepath); + return error; +} + +static boolean_t +bsdmgroot_bootable(void) +{ +#if defined(__arm64__) +#define BSDMGROOT_DEFAULT true +#else +#define BSDMGROOT_DEFAULT false +#endif + + boolean_t resolved = BSDMGROOT_DEFAULT; + + boolean_t boot_arg_bsdmgroot = false; + boolean_t boot_arg_nobsdmgroot = false; + int error; + mount_t mp; + boolean_t root_part_of_volume_group = false; + struct vfs_attr vfsattr; + + mp = rootvnode->v_mount; + VFSATTR_INIT(&vfsattr); + VFSATTR_WANTED(&vfsattr, f_capabilities); + + boot_arg_bsdmgroot = PE_parse_boot_argn("-bsdmgroot", NULL, 0); + boot_arg_nobsdmgroot = PE_parse_boot_argn("-nobsdmgroot", NULL, 0); + + error = vfs_getattr(mp, &vfsattr, vfs_context_kernel()); + if (!error && VFSATTR_IS_SUPPORTED(&vfsattr, f_capabilities)) { + if ((vfsattr.f_capabilities.capabilities[VOL_CAPABILITIES_FORMAT] & VOL_CAP_FMT_VOL_GROUPS) && + (vfsattr.f_capabilities.valid[VOL_CAPABILITIES_FORMAT] & VOL_CAP_FMT_VOL_GROUPS)) { + root_part_of_volume_group = true; + } + } + + boolean_t singleuser = (boothowto & RB_SINGLE) != 0; + + // Start with the #defined default above. + // If booting to single-user mode, default to false, because single- + // user mode inside the BaseSystem is probably not what's wanted. + // If the 'yes' boot-arg is set, we'll allow that even in single-user + // mode, we'll assume you know what you're doing. + // The 'no' boot-arg overpowers the 'yes' boot-arg. + // In any case, we will not attempt to root from BaseSystem if the + // original (booter-chosen) root volume isn't in a volume group. + // This is just out of an abundance of caution: if the boot environment + // seems to be "something other than a standard install", + // we'll be conservative in messing with the root volume. + + if (singleuser) { + resolved = false; + } + + if (boot_arg_bsdmgroot) { + resolved = true; + } + + if (boot_arg_nobsdmgroot) { + resolved = false; + } + + if (!root_part_of_volume_group) { + resolved = false; + } + + return resolved; } +#endif // CONFIG_BASESYSTEMROOT -#if !NFSCLIENT -int +void +bsd_exec_setup(int scale) +{ + switch (scale) { + case 0: + case 1: + bsd_simul_execs = BSD_SIMUL_EXECS; + break; + case 2: + case 3: + bsd_simul_execs = 65; + break; + case 4: + case 5: + bsd_simul_execs = 129; + break; + case 6: + case 7: + bsd_simul_execs = 257; + break; + default: + bsd_simul_execs = 513; + break; + } + bsd_pageable_map_size = (bsd_simul_execs * BSD_PAGEABLE_SIZE_PER_EXEC); +} + +#if !CONFIG_NETBOOT +int +netboot_root(void); + +int netboot_root(void) { - return(0); + return 0; } #endif