X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/5c9f46613a83ebfc29a5b1f099448259e96a98f0..bca245acd4c03fd752d1a45f011ad495e60fe53d:/bsd/kern/kern_exec.c diff --git a/bsd/kern/kern_exec.c b/bsd/kern/kern_exec.c index 3ccb12119..afa0cb820 100644 --- a/bsd/kern/kern_exec.c +++ b/bsd/kern/kern_exec.c @@ -1,8 +1,8 @@ /* - * Copyright (c) 2000-2013 Apple Inc. All rights reserved. + * Copyright (c) 2000-2019 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, @@ -22,7 +22,7 @@ * 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) 1995 NeXT Computer, Inc. All Rights Reserved */ @@ -32,7 +32,7 @@ * All rights reserved. The CMU software License Agreement specifies * the terms and conditions for use and redistribution. */ - + /*- * Copyright (c) 1982, 1986, 1991, 1993 * The Regents of the University of California. All rights reserved. @@ -92,7 +92,7 @@ #include #include #include -#include +#include #include #include #include @@ -105,9 +105,9 @@ #include #include #if SYSV_SHM -#include /* shmexec() */ +#include /* shmexec() */ #endif -#include /* ubc_map() */ +#include /* ubc_map() */ #include #include #include @@ -121,6 +121,7 @@ #include +#include #include #include #include @@ -145,6 +146,14 @@ #include #endif +#if CONFIG_AUDIT +#include +#endif + +#if CONFIG_ARCADE +#include +#endif + #include #include #include @@ -154,6 +163,7 @@ #include +#include #include #include @@ -162,6 +172,12 @@ #include #endif +#include + +extern boolean_t vm_darkwake_mode; + +extern int bootarg_execfailurereports; /* bsd_init.c */ + #if CONFIG_DTRACE /* Do not include dtrace.h, it redefines kmem_[alloc/free] */ extern void dtrace_proc_exec(proc_t); @@ -177,7 +193,13 @@ static void (*dtrace_proc_waitfor_hook)(proc_t) = NULL; #endif /* support for child creation in exec after vfork */ -thread_t fork_create_child(task_t parent_task, coalition_t *parent_coalition, proc_t child_proc, int inherit_memory, int is64bit, int in_exec); +thread_t fork_create_child(task_t parent_task, + coalition_t *parent_coalition, + proc_t child_proc, + int inherit_memory, + int is_64bit_addr, + int is_64bit_data, + int in_exec); void vfork_exit(proc_t p, int rv); extern void proc_apply_task_networkbg_internal(proc_t, thread_t); extern void task_set_did_exec_flag(task_t task); @@ -188,19 +210,24 @@ boolean_t thread_is_active(thread_t thread); void thread_copy_resource_info(thread_t dst_thread, thread_t src_thread); void *ipc_importance_exec_switch_task(task_t old_task, task_t new_task); extern void ipc_importance_release(void *elem); +extern boolean_t task_has_watchports(task_t task); /* * Mach things for which prototypes are unavailable from Mach headers */ -void ipc_task_reset( - task_t task); -void ipc_thread_reset( - thread_t thread); +#define IPC_KMSG_FLAGS_ALLOW_IMMOVABLE_SEND 0x1 +void ipc_task_reset( + task_t task); +void ipc_thread_reset( + thread_t thread); kern_return_t ipc_object_copyin( - ipc_space_t space, - mach_port_name_t name, - mach_msg_type_name_t msgt_name, - ipc_object_t *objectp); + ipc_space_t space, + mach_port_name_t name, + mach_msg_type_name_t msgt_name, + ipc_object_t *objectp, + mach_port_context_t context, + mach_msg_guard_flags_t *guard_flags, + uint32_t kmsg_flags); void ipc_port_release_send(ipc_port_t); #if DEVELOPMENT || DEBUG @@ -229,7 +256,7 @@ __attribute__((noinline)) int __EXEC_WAITING_ON_TASKGATED_CODE_SIGNATURE_UPCALL_ * activator in exec_activate_image() before treating * it as malformed/corrupt. */ -#define EAI_ITERLIMIT 3 +#define EAI_ITERLIMIT 3 /* * For #! interpreter parsing @@ -239,22 +266,30 @@ __attribute__((noinline)) int __EXEC_WAITING_ON_TASKGATED_CODE_SIGNATURE_UPCALL_ extern vm_map_t bsd_pageable_map; extern const struct fileops vnops; +extern int nextpidversion; -#define USER_ADDR_ALIGN(addr, val) \ +#define USER_ADDR_ALIGN(addr, val) \ ( ( (user_addr_t)(addr) + (val) - 1) \ - & ~((val) - 1) ) + & ~((val) - 1) ) - /* Platform Code Exec Logging */ +/* Platform Code Exec Logging */ static int platform_exec_logging = 0; SYSCTL_DECL(_security_mac); SYSCTL_INT(_security_mac, OID_AUTO, platform_exec_logging, CTLFLAG_RW, &platform_exec_logging, 0, - "log cdhashes for all platform binary executions"); + "log cdhashes for all platform binary executions"); static os_log_t peLog = OS_LOG_DEFAULT; -struct image_params; /* Forward */ +struct exec_port_actions { + uint32_t portwatch_count; + uint32_t registered_count; + ipc_port_t *portwatch_array; + ipc_port_t *registered_array; +}; + +struct image_params; /* Forward */ static int exec_activate_image(struct image_params *imgp); static int exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp); static int load_return_to_errno(load_return_t lrtn); @@ -265,15 +300,17 @@ static int exec_extract_strings(struct image_params *imgp); static int exec_add_apple_strings(struct image_params *imgp, const load_result_t *load_result); static int exec_handle_sugid(struct image_params *imgp); static int sugid_scripts = 0; -SYSCTL_INT (_kern, OID_AUTO, sugid_scripts, CTLFLAG_RW | CTLFLAG_LOCKED, &sugid_scripts, 0, ""); +SYSCTL_INT(_kern, OID_AUTO, sugid_scripts, CTLFLAG_RW | CTLFLAG_LOCKED, &sugid_scripts, 0, ""); static kern_return_t create_unix_stack(vm_map_t map, load_result_t* load_result, proc_t p); static int copyoutptr(user_addr_t ua, user_addr_t ptr, int ptr_size); static void exec_resettextvp(proc_t, struct image_params *); static int check_for_signature(proc_t, struct image_params *); static void exec_prefault_data(proc_t, struct image_params *, load_result_t *); -static errno_t exec_handle_port_actions(struct image_params *imgp, boolean_t * portwatch_present, ipc_port_t * portwatch_ports); -static errno_t exec_handle_spawnattr_policy(proc_t p, int psa_apptype, uint64_t psa_qos_clamp, uint64_t psa_darwin_role, - ipc_port_t * portwatch_ports, int portwatch_count); +static errno_t exec_handle_port_actions(struct image_params *imgp, + struct exec_port_actions *port_actions); +static errno_t exec_handle_spawnattr_policy(proc_t p, thread_t thread, int psa_apptype, uint64_t psa_qos_clamp, + uint64_t psa_darwin_role, struct exec_port_actions *port_actions); +static void exec_port_actions_destroy(struct exec_port_actions *port_actions); /* * exec_add_user_string @@ -297,35 +334,36 @@ static int exec_add_user_string(struct image_params *imgp, user_addr_t str, int seg, boolean_t is_ncargs) { int error = 0; - + do { size_t len = 0; int space; - - if (is_ncargs) + + if (is_ncargs) { space = imgp->ip_argspace; /* by definition smaller than ip_strspace */ - else + } else { space = imgp->ip_strspace; - + } + if (space <= 0) { error = E2BIG; break; } - + if (!UIO_SEG_IS_USER_SPACE(seg)) { - char *kstr = CAST_DOWN(char *,str); /* SAFE */ + char *kstr = CAST_DOWN(char *, str); /* SAFE */ error = copystr(kstr, imgp->ip_strendp, space, &len); - } else { + } else { error = copyinstr(str, imgp->ip_strendp, space, &len); } imgp->ip_strendp += len; imgp->ip_strspace -= len; - if (is_ncargs) + if (is_ncargs) { imgp->ip_argspace -= len; - + } } while (error == ENAMETOOLONG); - + return error; } @@ -333,7 +371,7 @@ exec_add_user_string(struct image_params *imgp, user_addr_t str, int seg, boolea * dyld is now passed the executable path as a getenv-like variable * in the same fashion as the stack_guard and malloc_entropy keys. */ -#define EXECUTABLE_KEY "executable_path=" +#define EXECUTABLE_KEY "executable_path=" /* * exec_save_path @@ -382,13 +420,13 @@ exec_save_path(struct image_params *imgp, user_addr_t path, int seg, const char len = MIN(MAXPATHLEN, imgp->ip_strspace); - switch(seg) { + switch (seg) { case UIO_USERSPACE32: - case UIO_USERSPACE64: /* Same for copyin()... */ + case UIO_USERSPACE64: /* Same for copyin()... */ error = copyinstr(path, imgp->ip_strings + strlen(EXECUTABLE_KEY), len, &len); break; case UIO_SYSSPACE: - kpath = CAST_DOWN(char *,path); /* SAFE */ + kpath = CAST_DOWN(char *, path); /* SAFE */ error = copystr(kpath, imgp->ip_strings + strlen(EXECUTABLE_KEY), len, &len); break; default: @@ -408,7 +446,7 @@ exec_save_path(struct image_params *imgp, user_addr_t path, int seg, const char } } - return(error); + return error; } /* @@ -416,7 +454,7 @@ exec_save_path(struct image_params *imgp, user_addr_t path, int seg, const char * * If we detect a shell script, we need to reset the string area * state so that the interpreter can be saved onto the stack. - + * * Parameters; struct image_params * image parameter block * * Returns: int 0 Success @@ -433,9 +471,9 @@ exec_reset_save_path(struct image_params *imgp) { imgp->ip_strendp = imgp->ip_strings; imgp->ip_argspace = NCARGS; - imgp->ip_strspace = ( NCARGS + PAGE_SIZE ); + imgp->ip_strspace = (NCARGS + PAGE_SIZE); - return (0); + return 0; } /* @@ -474,12 +512,12 @@ exec_shell_imgact(struct image_params *imgp) if (vdata[0] != '#' || vdata[1] != '!' || (imgp->ip_flags & IMGPF_INTERPRET) != 0) { - return (-1); + return -1; } if (imgp->ip_origcputype != 0) { /* Fat header previously matched, don't allow shell script inside */ - return (-1); + return -1; } imgp->ip_flags |= IMGPF_INTERPRET; @@ -495,10 +533,10 @@ exec_shell_imgact(struct image_params *imgp) } /* Try to find the first non-whitespace character */ - for( ihp = &vdata[2]; ihp < &vdata[IMG_SHSIZE]; ihp++ ) { + for (ihp = &vdata[2]; ihp < &vdata[IMG_SHSIZE]; ihp++) { if (IS_EOL(*ihp)) { /* Did not find interpreter, "#!\n" */ - return (ENOEXEC); + return ENOEXEC; } else if (IS_WHITESPACE(*ihp)) { /* Whitespace, like "#! /bin/sh\n", keep going. */ } else { @@ -509,13 +547,13 @@ exec_shell_imgact(struct image_params *imgp) if (ihp == &vdata[IMG_SHSIZE]) { /* All whitespace, like "#! " */ - return (ENOEXEC); + return ENOEXEC; } line_startp = ihp; /* Try to find the end of the interpreter+args string */ - for ( ; ihp < &vdata[IMG_SHSIZE]; ihp++ ) { + for (; ihp < &vdata[IMG_SHSIZE]; ihp++) { if (IS_EOL(*ihp)) { /* Got it */ break; @@ -526,7 +564,7 @@ exec_shell_imgact(struct image_params *imgp) if (ihp == &vdata[IMG_SHSIZE]) { /* A long line, like "#! blah blah blah" without end */ - return (ENOEXEC); + return ENOEXEC; } /* Backtrack until we find the last non-whitespace */ @@ -546,18 +584,20 @@ exec_shell_imgact(struct image_params *imgp) /* copy the interpreter name */ interp = imgp->ip_interp_buffer; - for ( ihp = line_startp; (ihp < line_endp) && !IS_WHITESPACE(*ihp); ihp++) + for (ihp = line_startp; (ihp < line_endp) && !IS_WHITESPACE(*ihp); ihp++) { *interp++ = *ihp; + } *interp = '\0'; exec_reset_save_path(imgp); exec_save_path(imgp, CAST_USER_ADDR_T(imgp->ip_interp_buffer), - UIO_SYSSPACE, NULL); + UIO_SYSSPACE, NULL); /* Copy the entire interpreter + args for later processing into argv[] */ interp = imgp->ip_interp_buffer; - for ( ihp = line_startp; (ihp < line_endp); ihp++) + for (ihp = line_startp; (ihp < line_endp); ihp++) { *interp++ = *ihp; + } *interp = '\0'; #if !SECURE_KERNEL @@ -574,13 +614,14 @@ exec_shell_imgact(struct image_params *imgp) p = vfs_context_proc(imgp->ip_vfs_context); error = falloc(p, &fp, &fd, imgp->ip_vfs_context); - if (error) - return(error); + if (error) { + return error; + } fp->f_fglob->fg_flag = FREAD; fp->f_fglob->fg_ops = &vnops; fp->f_fglob->fg_data = (caddr_t)imgp->ip_vp; - + proc_fdlock(p); procfdtbl_releasefd(p, fd, NULL); fp_drop(p, fd, fp, 1); @@ -591,7 +632,7 @@ exec_shell_imgact(struct image_params *imgp) } #endif - return (-3); + return -3; } @@ -617,7 +658,7 @@ exec_shell_imgact(struct image_params *imgp) * activators should not be given the opportunity to attempt * to activate the image. * - * If we find an encapsulated binary, we make no assertions + * If we find an encapsulated binary, we make no assertions * about its validity; instead, we leave that up to a rescan * for an activator to claim it, and, if it is claimed by one, * that activator is responsible for determining validity. @@ -672,9 +713,10 @@ exec_fat_imgact(struct image_params *imgp) } lret = fatfile_getbestarch_for_cputype(pref, - (vm_offset_t)fat_header, - PAGE_SIZE, - &fat_arch); + (vm_offset_t)fat_header, + PAGE_SIZE, + imgp, + &fat_arch); if (lret == LOAD_SUCCESS) { goto use_arch; } @@ -688,8 +730,9 @@ exec_fat_imgact(struct image_params *imgp) regular_grading: /* Look up our preferred architecture in the fat file. */ lret = fatfile_getbestarch((vm_offset_t)fat_header, - PAGE_SIZE, - &fat_arch); + PAGE_SIZE, + imgp, + &fat_arch); if (lret != LOAD_SUCCESS) { error = load_return_to_errno(lret); goto bad; @@ -698,9 +741,9 @@ regular_grading: use_arch: /* Read the Mach-O header out of fat_arch */ error = vn_rdwr(UIO_READ, imgp->ip_vp, imgp->ip_vdata, - PAGE_SIZE, fat_arch.offset, - UIO_SYSSPACE, (IO_UNIT|IO_NODELOCKED), - cred, &resid, p); + PAGE_SIZE, fat_arch.offset, + UIO_SYSSPACE, (IO_UNIT | IO_NODELOCKED), + cred, &resid, p); if (error) { goto bad; } @@ -718,7 +761,7 @@ use_arch: bad: kauth_cred_unref(&cred); - return (error); + return error; } static int @@ -727,13 +770,13 @@ activate_exec_state(task_t task, proc_t p, thread_t thread, load_result_t *resul int ret; task_set_dyld_info(task, MACH_VM_MIN_ADDRESS, 0); - if (result->is64bit) { - task_set_64bit(task, TRUE); + task_set_64bit(task, result->is_64bit_addr, result->is_64bit_data); + if (result->is_64bit_addr) { OSBitOrAtomic(P_LP64, &p->p_flag); } else { - task_set_64bit(task, FALSE); OSBitAndAtomic(~((uint32_t)P_LP64), &p->p_flag); } + task_set_mach_header_address(task, result->mach_header); ret = thread_state_initialize(thread); if (ret != KERN_SUCCESS) { @@ -776,7 +819,7 @@ set_proc_name(struct image_params *imgp, proc_t p) } bcopy((caddr_t)imgp->ip_ndp->ni_cnd.cn_nameptr, (caddr_t)p->p_name, - (unsigned)imgp->ip_ndp->ni_cnd.cn_namelen); + (unsigned)imgp->ip_ndp->ni_cnd.cn_namelen); p->p_name[imgp->ip_ndp->ni_cnd.cn_namelen] = '\0'; if (imgp->ip_ndp->ni_cnd.cn_namelen > MAXCOMLEN) { @@ -784,19 +827,10 @@ set_proc_name(struct image_params *imgp, proc_t p) } bcopy((caddr_t)imgp->ip_ndp->ni_cnd.cn_nameptr, (caddr_t)p->p_comm, - (unsigned)imgp->ip_ndp->ni_cnd.cn_namelen); + (unsigned)imgp->ip_ndp->ni_cnd.cn_namelen); p->p_comm[imgp->ip_ndp->ni_cnd.cn_namelen] = '\0'; } -static uint64_t get_va_fsid(struct vnode_attr *vap) -{ - if (VATTR_IS_SUPPORTED(vap, va_fsid64)) { - return *(uint64_t *)&vap->va_fsid64; - } else { - return vap->va_fsid; - } -} - /* * exec_mach_imgact * @@ -824,28 +858,28 @@ static int exec_mach_imgact(struct image_params *imgp) { struct mach_header *mach_header = (struct mach_header *)imgp->ip_vdata; - proc_t p = vfs_context_proc(imgp->ip_vfs_context); - int error = 0; - task_t task; - task_t new_task = NULL; /* protected by vfexec */ - thread_t thread; - struct uthread *uthread; + proc_t p = vfs_context_proc(imgp->ip_vfs_context); + int error = 0; + task_t task; + task_t new_task = NULL; /* protected by vfexec */ + thread_t thread; + struct uthread *uthread; vm_map_t old_map = VM_MAP_NULL; vm_map_t map = VM_MAP_NULL; - load_return_t lret; - load_result_t load_result; + load_return_t lret; + load_result_t load_result = {}; struct _posix_spawnattr *psa = NULL; - int spawn = (imgp->ip_flags & IMGPF_SPAWN); - int vfexec = (imgp->ip_flags & IMGPF_VFORK_EXEC); - int exec = (imgp->ip_flags & IMGPF_EXEC); - os_reason_t exec_failure_reason = OS_REASON_NULL; + int spawn = (imgp->ip_flags & IMGPF_SPAWN); + int vfexec = (imgp->ip_flags & IMGPF_VFORK_EXEC); + int exec = (imgp->ip_flags & IMGPF_EXEC); + os_reason_t exec_failure_reason = OS_REASON_NULL; /* * make sure it's a Mach-O 1.0 or Mach-O 2.0 binary; the difference * is a reserved field on the end, so for the most part, we can * treat them as if they were identical. Reverse-endian Mach-O * binaries are recognized but not compatible. - */ + */ if ((mach_header->magic == MH_CIGAM) || (mach_header->magic == MH_CIGAM_64)) { error = EBADARCH; @@ -866,7 +900,7 @@ exec_mach_imgact(struct image_params *imgp) if (imgp->ip_origcputype != 0) { /* Fat header previously had an idea about this thin file */ if (imgp->ip_origcputype != mach_header->cputype || - imgp->ip_origcpusubtype != mach_header->cpusubtype) { + imgp->ip_origcpusubtype != mach_header->cpusubtype) { error = EBADARCH; goto bad; } @@ -879,8 +913,9 @@ exec_mach_imgact(struct image_params *imgp) thread = current_thread(); uthread = get_bsdthread_info(thread); - if ((mach_header->cputype & CPU_ARCH_ABI64) == CPU_ARCH_ABI64) - imgp->ip_flags |= IMGPF_IS_64BIT; + if ((mach_header->cputype & CPU_ARCH_ABI64) == CPU_ARCH_ABI64) { + imgp->ip_flags |= IMGPF_IS_64BIT_ADDR | IMGPF_IS_64BIT_DATA; + } /* If posix_spawn binprefs exist, respect those prefs. */ psa = (struct _posix_spawnattr *) imgp->ip_px_sa; @@ -908,17 +943,42 @@ exec_mach_imgact(struct image_params *imgp) goto bad; } grade: - if (!grade_binary(imgp->ip_origcputype, imgp->ip_origcpusubtype & ~CPU_SUBTYPE_MASK)) { + if (!grade_binary(imgp->ip_origcputype, imgp->ip_origcpusubtype & ~CPU_SUBTYPE_MASK, TRUE)) { + error = EBADARCH; + goto bad; + } + + if (validate_potential_simulator_binary(imgp->ip_origcputype, imgp, + imgp->ip_arch_offset, imgp->ip_arch_size) != LOAD_SUCCESS) { +#if __x86_64__ + const char *excpath; + error = exec_save_path(imgp, imgp->ip_user_fname, imgp->ip_seg, &excpath); + os_log_error(OS_LOG_DEFAULT, "Unsupported 32-bit executable: \"%s\"", (error) ? imgp->ip_vp->v_name : excpath); +#endif error = EBADARCH; goto bad; } +#if defined(HAS_APPLE_PAC) + assert(mach_header->cputype == CPU_TYPE_ARM64 + ); + + if (((mach_header->cputype == CPU_TYPE_ARM64 && + (mach_header->cpusubtype & ~CPU_SUBTYPE_MASK) == CPU_SUBTYPE_ARM64E) + ) && (CPU_SUBTYPE_ARM64_PTR_AUTH_VERSION(mach_header->cpusubtype) == 0)) { + imgp->ip_flags &= ~IMGPF_NOJOP; + } else { + imgp->ip_flags |= IMGPF_NOJOP; + } +#endif + /* Copy in arguments/environment from the old process */ error = exec_extract_strings(imgp); - if (error) + if (error) { goto bad; + } - AUDIT_ARG(argv, imgp->ip_startargv, imgp->ip_argc, + AUDIT_ARG(argv, imgp->ip_startargv, imgp->ip_argc, imgp->ip_endargv - imgp->ip_startargv); AUDIT_ARG(envv, imgp->ip_endargv, imgp->ip_envc, imgp->ip_endenvv - imgp->ip_endargv); @@ -931,7 +991,13 @@ grade: * new child process. */ if (vfexec) { - imgp->ip_new_thread = fork_create_child(task, NULL, p, FALSE, (imgp->ip_flags & IMGPF_IS_64BIT), FALSE); + imgp->ip_new_thread = fork_create_child(task, + NULL, + p, + FALSE, + (imgp->ip_flags & IMGPF_IS_64BIT_ADDR), + (imgp->ip_flags & IMGPF_IS_64BIT_DATA), + FALSE); /* task and thread ref returned, will be released in __mac_execve */ if (imgp->ip_new_thread == NULL) { error = ENOMEM; @@ -965,62 +1031,89 @@ grade: error = load_return_to_errno(lret); KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_BAD_MACHO, 0, 0); + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_BAD_MACHO, 0, 0); if (lret == LOAD_BADMACHO_UPX) { - /* set anything that might be useful in the crash report */ set_proc_name(imgp, p); - exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_UPX); exec_failure_reason->osr_flags |= OS_REASON_FLAG_GENERATE_CRASH_REPORT; - exec_failure_reason->osr_flags |= OS_REASON_FLAG_CONSISTENT_FAILURE; } else { exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_BAD_MACHO); + + if (bootarg_execfailurereports) { + set_proc_name(imgp, p); + exec_failure_reason->osr_flags |= OS_REASON_FLAG_GENERATE_CRASH_REPORT; + } } + exec_failure_reason->osr_flags |= OS_REASON_FLAG_CONSISTENT_FAILURE; + goto badtoolate; } proc_lock(p); p->p_cputype = imgp->ip_origcputype; p->p_cpusubtype = imgp->ip_origcpusubtype; + p->p_platform = load_result.ip_platform; + p->p_sdk = load_result.lr_sdk; proc_unlock(p); vm_map_set_user_wire_limit(map, p->p_rlimit[RLIMIT_MEMLOCK].rlim_cur); - /* + /* * Set code-signing flags if this binary is signed, or if parent has * requested them on exec. */ if (load_result.csflags & CS_VALID) { - imgp->ip_csflags |= load_result.csflags & - (CS_VALID|CS_SIGNED|CS_DEV_CODE| - CS_HARD|CS_KILL|CS_RESTRICT|CS_ENFORCEMENT|CS_REQUIRE_LV| - CS_ENTITLEMENTS_VALIDATED|CS_DYLD_PLATFORM| - CS_ENTITLEMENT_FLAGS| - CS_EXEC_SET_HARD|CS_EXEC_SET_KILL|CS_EXEC_SET_ENFORCEMENT); + imgp->ip_csflags |= load_result.csflags & + (CS_VALID | CS_SIGNED | CS_DEV_CODE | + CS_HARD | CS_KILL | CS_RESTRICT | CS_ENFORCEMENT | CS_REQUIRE_LV | + CS_FORCED_LV | CS_ENTITLEMENTS_VALIDATED | CS_DYLD_PLATFORM | CS_RUNTIME | + CS_ENTITLEMENT_FLAGS | + CS_EXEC_SET_HARD | CS_EXEC_SET_KILL | CS_EXEC_SET_ENFORCEMENT); } else { imgp->ip_csflags &= ~CS_VALID; } - if (p->p_csflags & CS_EXEC_SET_HARD) + if (p->p_csflags & CS_EXEC_SET_HARD) { imgp->ip_csflags |= CS_HARD; - if (p->p_csflags & CS_EXEC_SET_KILL) + } + if (p->p_csflags & CS_EXEC_SET_KILL) { imgp->ip_csflags |= CS_KILL; - if (p->p_csflags & CS_EXEC_SET_ENFORCEMENT) + } + if (p->p_csflags & CS_EXEC_SET_ENFORCEMENT) { imgp->ip_csflags |= CS_ENFORCEMENT; + } if (p->p_csflags & CS_EXEC_INHERIT_SIP) { - if (p->p_csflags & CS_INSTALLER) + if (p->p_csflags & CS_INSTALLER) { imgp->ip_csflags |= CS_INSTALLER; - if (p->p_csflags & CS_DATAVAULT_CONTROLLER) + } + if (p->p_csflags & CS_DATAVAULT_CONTROLLER) { imgp->ip_csflags |= CS_DATAVAULT_CONTROLLER; - if (p->p_csflags & CS_NVRAM_UNRESTRICTED) + } + if (p->p_csflags & CS_NVRAM_UNRESTRICTED) { imgp->ip_csflags |= CS_NVRAM_UNRESTRICTED; + } } /* * Set up the system reserved areas in the new address space. */ - vm_map_exec(map, task, load_result.is64bit, (void *)p->p_fd->fd_rdir, cpu_type()); + int cpu_subtype; + cpu_subtype = 0; /* all cpu_subtypes use the same shared region */ +#if defined(HAS_APPLE_PAC) + if (cpu_type() == CPU_TYPE_ARM64 && + (p->p_cpusubtype & ~CPU_SUBTYPE_MASK) == CPU_SUBTYPE_ARM64E) { + assertf(p->p_cputype == CPU_TYPE_ARM64, + "p %p cpu_type() 0x%x p->p_cputype 0x%x p->p_cpusubtype 0x%x", + p, cpu_type(), p->p_cputype, p->p_cpusubtype); + /* + * arm64e uses pointer authentication, so request a separate + * shared region for this CPU subtype. + */ + cpu_subtype = p->p_cpusubtype & ~CPU_SUBTYPE_MASK; + } +#endif /* HAS_APPLE_PAC */ + vm_map_exec(map, task, load_result.is_64bit_addr, (void *)p->p_fd->fd_rdir, cpu_type(), cpu_subtype); /* * Close file descriptors which specify close-on-exec. @@ -1035,8 +1128,14 @@ grade: vm_map_deallocate(map); KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_SUGID_FAILURE, 0, 0); + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_SUGID_FAILURE, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_SUGID_FAILURE); + if (bootarg_execfailurereports) { + set_proc_name(imgp, p); + exec_failure_reason->osr_flags |= OS_REASON_FLAG_GENERATE_CRASH_REPORT; + } + goto badtoolate; } @@ -1058,41 +1157,58 @@ grade: lret = activate_exec_state(task, p, thread, &load_result); if (lret != KERN_SUCCESS) { - KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_ACTV_THREADSTATE, 0, 0); + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_ACTV_THREADSTATE, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_ACTV_THREADSTATE); + if (bootarg_execfailurereports) { + set_proc_name(imgp, p); + exec_failure_reason->osr_flags |= OS_REASON_FLAG_GENERATE_CRASH_REPORT; + } + goto badtoolate; } /* * deal with voucher on exec-calling thread. */ - if (imgp->ip_new_thread == NULL) + if (imgp->ip_new_thread == NULL) { thread_set_mach_voucher(current_thread(), IPC_VOUCHER_NULL); + } /* Make sure we won't interrupt ourself signalling a partial process */ - if (!vfexec && !spawn && (p->p_lflag & P_LTRACED)) + if (!vfexec && !spawn && (p->p_lflag & P_LTRACED)) { psignal(p, SIGTRAP); + } if (load_result.unixproc && - create_unix_stack(get_task_map(task), - &load_result, - p) != KERN_SUCCESS) { + create_unix_stack(get_task_map(task), + &load_result, + p) != KERN_SUCCESS) { error = load_return_to_errno(LOAD_NOSPACE); KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_STACK_ALLOC, 0, 0); + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_STACK_ALLOC, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_STACK_ALLOC); + if (bootarg_execfailurereports) { + set_proc_name(imgp, p); + exec_failure_reason->osr_flags |= OS_REASON_FLAG_GENERATE_CRASH_REPORT; + } + goto badtoolate; } error = exec_add_apple_strings(imgp, &load_result); if (error) { - KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_APPLE_STRING_INIT, 0, 0); + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_APPLE_STRING_INIT, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_APPLE_STRING_INIT); + if (bootarg_execfailurereports) { + set_proc_name(imgp, p); + exec_failure_reason->osr_flags |= OS_REASON_FLAG_GENERATE_CRASH_REPORT; + } goto badtoolate; } @@ -1100,7 +1216,7 @@ grade: old_map = vm_map_switch(get_task_map(task)); if (load_result.unixproc) { - user_addr_t ap; + user_addr_t ap; /* * Copy the strings area out into the new process address @@ -1112,8 +1228,13 @@ grade: vm_map_switch(old_map); KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_COPYOUT_STRINGS, 0, 0); + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_COPYOUT_STRINGS, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_COPYOUT_STRINGS); + if (bootarg_execfailurereports) { + set_proc_name(imgp, p); + exec_failure_reason->osr_flags |= OS_REASON_FLAG_GENERATE_CRASH_REPORT; + } goto badtoolate; } /* Set the stack */ @@ -1121,8 +1242,8 @@ grade: } if (load_result.dynlinker) { - uint64_t ap; - int new_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT) ? 8 : 4; + uint64_t ap; + int new_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) ? 8 : 4; /* Adjust the stack */ ap = thread_adjuserstack(thread, -new_ptr_size); @@ -1132,8 +1253,13 @@ grade: vm_map_switch(old_map); KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_COPYOUT_DYNLINKER, 0, 0); + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_COPYOUT_DYNLINKER, 0, 0); + exec_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_COPYOUT_DYNLINKER); + if (bootarg_execfailurereports) { + set_proc_name(imgp, p); + exec_failure_reason->osr_flags |= OS_REASON_FLAG_GENERATE_CRASH_REPORT; + } goto badtoolate; } task_set_dyld_info(task, load_result.all_image_info_addr, @@ -1145,9 +1271,6 @@ grade: vm_map_switch(old_map); - /* Stop profiling */ - stopprofclock(p); - /* * Reset signal state. */ @@ -1161,8 +1284,9 @@ grade: #if SYSV_SHM /* FIXME: Till vmspace inherit is fixed: */ - if (!vfexec && p->vm_shm) + if (!vfexec && p->vm_shm) { shmexec(p); + } #endif #if SYSV_SEM /* Clean up the semaphores */ @@ -1180,20 +1304,26 @@ grade: if (secluded_for_apps && load_result.platform_binary) { if (strncmp(p->p_name, - "Camera", - sizeof (p->p_name)) == 0) { + "Camera", + sizeof(p->p_name)) == 0) { task_set_could_use_secluded_mem(task, TRUE); } else { task_set_could_use_secluded_mem(task, FALSE); } if (strncmp(p->p_name, - "mediaserverd", - sizeof (p->p_name)) == 0) { + "mediaserverd", + sizeof(p->p_name)) == 0) { task_set_could_also_use_secluded_mem(task, TRUE); } } #endif /* CONFIG_SECLUDED_MEMORY */ +#if __arm64__ + if (load_result.legacy_footprint) { + task_set_legacy_footprint(task); + } +#endif /* __arm64__ */ + pal_dbg_set_task_name(task); /* @@ -1234,16 +1364,11 @@ grade: #endif if (kdebug_enable) { - long dbg_arg1, dbg_arg2, dbg_arg3, dbg_arg4; - - /* - * Collect the pathname for tracing - */ - kdbg_trace_string(p, &dbg_arg1, &dbg_arg2, &dbg_arg3, &dbg_arg4); + long args[4] = {}; uintptr_t fsid = 0, fileid = 0; if (imgp->ip_vattr) { - uint64_t fsid64 = get_va_fsid(imgp->ip_vattr); + uint64_t fsid64 = vnode_get_va_fsid(imgp->ip_vattr); fsid = fsid64; fileid = imgp->ip_vattr->va_fileid; // check for (unexpected) overflow and trace zero in that case @@ -1251,10 +1376,15 @@ grade: fsid = fileid = 0; } } - KERNEL_DEBUG_CONSTANT1(TRACE_DATA_EXEC | DBG_FUNC_NONE, - p->p_pid , fsid, fileid, 0, (uintptr_t)thread_tid(thread)); - KERNEL_DEBUG_CONSTANT1(TRACE_STRING_EXEC | DBG_FUNC_NONE, - dbg_arg1, dbg_arg2, dbg_arg3, dbg_arg4, (uintptr_t)thread_tid(thread)); + KERNEL_DEBUG_CONSTANT_IST1(TRACE_DATA_EXEC, p->p_pid, fsid, fileid, 0, + (uintptr_t)thread_tid(thread)); + + /* + * Collect the pathname for tracing + */ + kdbg_trace_string(p, &args[0], &args[1], &args[2], &args[3]); + KERNEL_DEBUG_CONSTANT_IST1(TRACE_STRING_EXEC, args[0], args[1], + args[2], args[3], (uintptr_t)thread_tid(thread)); } /* @@ -1318,7 +1448,7 @@ badtoolate: os_reason_free(exec_failure_reason); exec_failure_reason = OS_REASON_NULL; } - + done: if (load_result.threadstate) { kfree(load_result.threadstate, load_result.threadstate_sz); @@ -1328,7 +1458,7 @@ done: bad: /* If we hit this, we likely would have leaked an exit reason */ assert(exec_failure_reason == OS_REASON_NULL); - return(error); + return error; } @@ -1342,12 +1472,12 @@ bad: * XXX hardcoded, for now; should use linker sets */ struct execsw { - int (*ex_imgact)(struct image_params *); + int(*const ex_imgact)(struct image_params *); const char *ex_name; -} execsw[] = { - { exec_mach_imgact, "Mach-o Binary" }, - { exec_fat_imgact, "Fat Binary" }, - { exec_shell_imgact, "Interpreter Script" }, +}const execsw[] = { + { exec_mach_imgact, "Mach-o Binary" }, + { exec_fat_imgact, "Fat Binary" }, + { exec_shell_imgact, "Interpreter Script" }, { NULL, NULL} }; @@ -1386,15 +1516,16 @@ exec_activate_image(struct image_params *imgp) const char *excpath; int error; int resid; - int once = 1; /* save SGUID-ness for interpreted files */ + int once = 1; /* save SGUID-ness for interpreted files */ int i; int itercount = 0; proc_t p = vfs_context_proc(imgp->ip_vfs_context); error = execargs_alloc(imgp); - if (error) + if (error) { goto bad_notrans; - + } + error = exec_save_path(imgp, imgp->ip_user_fname, imgp->ip_seg, &excpath); if (error) { goto bad_notrans; @@ -1410,14 +1541,15 @@ exec_activate_image(struct image_params *imgp) } NDINIT(ndp, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF | AUDITVNPATH1, - UIO_SYSSPACE, CAST_USER_ADDR_T(excpath), imgp->ip_vfs_context); + UIO_SYSSPACE, CAST_USER_ADDR_T(excpath), imgp->ip_vfs_context); again: error = namei(ndp); - if (error) + if (error) { goto bad_notrans; - imgp->ip_ndp = ndp; /* successful namei(); call nameidone() later */ - imgp->ip_vp = ndp->ni_vp; /* if set, need to vnode_put() at some point */ + } + imgp->ip_ndp = ndp; /* successful namei(); call nameidone() later */ + imgp->ip_vp = ndp->ni_vp; /* if set, need to vnode_put() at some point */ /* * Before we start the transition from binary A to binary B, make @@ -1434,12 +1566,14 @@ again: } error = proc_transstart(p, 1, 0); proc_unlock(p); - if (error) + if (error) { goto bad_notrans; + } error = exec_check_permissions(imgp); - if (error) + if (error) { goto bad; + } /* Copy; avoid invocation of an interpreter overwriting the original */ if (once) { @@ -1448,11 +1582,12 @@ again: } error = vn_rdwr(UIO_READ, imgp->ip_vp, imgp->ip_vdata, PAGE_SIZE, 0, - UIO_SYSSPACE, IO_NODELOCKED, - vfs_context_ucred(imgp->ip_vfs_context), - &resid, vfs_context_proc(imgp->ip_vfs_context)); - if (error) + UIO_SYSSPACE, IO_NODELOCKED, + vfs_context_ucred(imgp->ip_vfs_context), + &resid, vfs_context_proc(imgp->ip_vfs_context)); + if (error) { goto bad; + } if (resid) { memset(imgp->ip_vdata + (PAGE_SIZE - resid), 0x0, resid); @@ -1465,50 +1600,53 @@ encapsulated_binary: goto bad; } error = -1; - for(i = 0; error == -1 && execsw[i].ex_imgact != NULL; i++) { - + for (i = 0; error == -1 && execsw[i].ex_imgact != NULL; i++) { error = (*execsw[i].ex_imgact)(imgp); switch (error) { /* case -1: not claimed: continue */ - case -2: /* Encapsulated binary, imgp->ip_XXX set for next iteration */ + case -2: /* Encapsulated binary, imgp->ip_XXX set for next iteration */ goto encapsulated_binary; - case -3: /* Interpreter */ + case -3: /* Interpreter */ #if CONFIG_MACF /* * Copy the script label for later use. Note that * the label can be different when the script is * actually read by the interpreter. */ - if (imgp->ip_scriptlabelp) + if (imgp->ip_scriptlabelp) { mac_vnode_label_free(imgp->ip_scriptlabelp); + } imgp->ip_scriptlabelp = mac_vnode_label_alloc(); if (imgp->ip_scriptlabelp == NULL) { error = ENOMEM; break; } mac_vnode_label_copy(imgp->ip_vp->v_label, - imgp->ip_scriptlabelp); + imgp->ip_scriptlabelp); /* * Take a ref of the script vnode for later use. */ - if (imgp->ip_scriptvp) + if (imgp->ip_scriptvp) { vnode_put(imgp->ip_scriptvp); - if (vnode_getwithref(imgp->ip_vp) == 0) + imgp->ip_scriptvp = NULLVP; + } + if (vnode_getwithref(imgp->ip_vp) == 0) { imgp->ip_scriptvp = imgp->ip_vp; + } #endif nameidone(ndp); vnode_put(imgp->ip_vp); - imgp->ip_vp = NULL; /* already put */ + imgp->ip_vp = NULL; /* already put */ imgp->ip_ndp = NULL; /* already nameidone */ /* Use excpath, which exec_shell_imgact reset to the interpreter */ NDINIT(ndp, LOOKUP, OP_LOOKUP, FOLLOW | LOCKLEAF, - UIO_SYSSPACE, CAST_USER_ADDR_T(excpath), imgp->ip_vfs_context); + UIO_SYSSPACE, CAST_USER_ADDR_T(excpath), imgp->ip_vfs_context); proc_transend(p, 0); goto again; @@ -1518,29 +1656,62 @@ encapsulated_binary: } } - /* - * Call out to allow 3rd party notification of exec. - * Ignore result of kauth_authorize_fileop call. - */ - if (error == 0 && kauth_authorize_fileop_has_listeners()) { - kauth_authorize_fileop(vfs_context_ucred(imgp->ip_vfs_context), - KAUTH_FILEOP_EXEC, - (uintptr_t)ndp->ni_vp, 0); + if (error == 0) { + if (imgp->ip_flags & IMGPF_INTERPRET && ndp->ni_vp) { + AUDIT_ARG(vnpath, ndp->ni_vp, ARG_VNODE2); + } + + /* + * Call out to allow 3rd party notification of exec. + * Ignore result of kauth_authorize_fileop call. + */ + if (kauth_authorize_fileop_has_listeners()) { + kauth_authorize_fileop(vfs_context_ucred(imgp->ip_vfs_context), + KAUTH_FILEOP_EXEC, + (uintptr_t)ndp->ni_vp, 0); + } } bad: proc_transend(p, 0); bad_notrans: - if (imgp->ip_strings) + if (imgp->ip_strings) { execargs_free(imgp); - if (imgp->ip_ndp) + } + if (imgp->ip_ndp) { nameidone(imgp->ip_ndp); - if (ndp) + } + if (ndp) { FREE(ndp, M_TEMP); + } - return (error); + return error; } +/* + * exec_validate_spawnattr_policy + * + * Description: Validates the entitlements required to set the apptype. + * + * Parameters: int psa_apptype posix spawn attribute apptype + * + * Returns: 0 Success + * EPERM Failure + */ +static errno_t +exec_validate_spawnattr_policy(int psa_apptype) +{ + if ((psa_apptype & POSIX_SPAWN_PROC_TYPE_MASK) != 0) { + int proctype = psa_apptype & POSIX_SPAWN_PROC_TYPE_MASK; + if (proctype == POSIX_SPAWN_PROC_TYPE_DRIVER) { + if (!IOTaskHasEntitlement(current_task(), POSIX_SPAWN_ENTITLEMENT_DRIVER)) { + return EPERM; + } + } + } + + return 0; +} /* * exec_handle_spawnattr_policy @@ -1553,8 +1724,8 @@ bad_notrans: * Returns: 0 Success */ static errno_t -exec_handle_spawnattr_policy(proc_t p, int psa_apptype, uint64_t psa_qos_clamp, uint64_t psa_darwin_role, - ipc_port_t * portwatch_ports, int portwatch_count) +exec_handle_spawnattr_policy(proc_t p, thread_t thread, int psa_apptype, uint64_t psa_qos_clamp, + uint64_t psa_darwin_role, struct exec_port_actions *port_actions) { int apptype = TASK_APPTYPE_NONE; int qos_clamp = THREAD_QOS_UNSPECIFIED; @@ -1563,49 +1734,52 @@ exec_handle_spawnattr_policy(proc_t p, int psa_apptype, uint64_t psa_qos_clamp, if ((psa_apptype & POSIX_SPAWN_PROC_TYPE_MASK) != 0) { int proctype = psa_apptype & POSIX_SPAWN_PROC_TYPE_MASK; - switch(proctype) { - case POSIX_SPAWN_PROC_TYPE_DAEMON_INTERACTIVE: - apptype = TASK_APPTYPE_DAEMON_INTERACTIVE; - break; - case POSIX_SPAWN_PROC_TYPE_DAEMON_STANDARD: - apptype = TASK_APPTYPE_DAEMON_STANDARD; - break; - case POSIX_SPAWN_PROC_TYPE_DAEMON_ADAPTIVE: - apptype = TASK_APPTYPE_DAEMON_ADAPTIVE; - break; - case POSIX_SPAWN_PROC_TYPE_DAEMON_BACKGROUND: - apptype = TASK_APPTYPE_DAEMON_BACKGROUND; - break; - case POSIX_SPAWN_PROC_TYPE_APP_DEFAULT: - apptype = TASK_APPTYPE_APP_DEFAULT; - break; + switch (proctype) { + case POSIX_SPAWN_PROC_TYPE_DAEMON_INTERACTIVE: + apptype = TASK_APPTYPE_DAEMON_INTERACTIVE; + break; + case POSIX_SPAWN_PROC_TYPE_DAEMON_STANDARD: + apptype = TASK_APPTYPE_DAEMON_STANDARD; + break; + case POSIX_SPAWN_PROC_TYPE_DAEMON_ADAPTIVE: + apptype = TASK_APPTYPE_DAEMON_ADAPTIVE; + break; + case POSIX_SPAWN_PROC_TYPE_DAEMON_BACKGROUND: + apptype = TASK_APPTYPE_DAEMON_BACKGROUND; + break; + case POSIX_SPAWN_PROC_TYPE_APP_DEFAULT: + apptype = TASK_APPTYPE_APP_DEFAULT; + break; #if !CONFIG_EMBEDDED - case POSIX_SPAWN_PROC_TYPE_APP_TAL: - apptype = TASK_APPTYPE_APP_TAL; - break; + case POSIX_SPAWN_PROC_TYPE_APP_TAL: + apptype = TASK_APPTYPE_APP_TAL; + break; #endif /* !CONFIG_EMBEDDED */ - default: - apptype = TASK_APPTYPE_NONE; - /* TODO: Should an invalid value here fail the spawn? */ - break; + case POSIX_SPAWN_PROC_TYPE_DRIVER: + apptype = TASK_APPTYPE_DRIVER; + break; + default: + apptype = TASK_APPTYPE_NONE; + /* TODO: Should an invalid value here fail the spawn? */ + break; } } if (psa_qos_clamp != POSIX_SPAWN_PROC_CLAMP_NONE) { switch (psa_qos_clamp) { - case POSIX_SPAWN_PROC_CLAMP_UTILITY: - qos_clamp = THREAD_QOS_UTILITY; - break; - case POSIX_SPAWN_PROC_CLAMP_BACKGROUND: - qos_clamp = THREAD_QOS_BACKGROUND; - break; - case POSIX_SPAWN_PROC_CLAMP_MAINTENANCE: - qos_clamp = THREAD_QOS_MAINTENANCE; - break; - default: - qos_clamp = THREAD_QOS_UNSPECIFIED; - /* TODO: Should an invalid value here fail the spawn? */ - break; + case POSIX_SPAWN_PROC_CLAMP_UTILITY: + qos_clamp = THREAD_QOS_UTILITY; + break; + case POSIX_SPAWN_PROC_CLAMP_BACKGROUND: + qos_clamp = THREAD_QOS_BACKGROUND; + break; + case POSIX_SPAWN_PROC_CLAMP_MAINTENANCE: + qos_clamp = THREAD_QOS_MAINTENANCE; + break; + default: + qos_clamp = THREAD_QOS_UNSPECIFIED; + /* TODO: Should an invalid value here fail the spawn? */ + break; } } @@ -1613,33 +1787,69 @@ exec_handle_spawnattr_policy(proc_t p, int psa_apptype, uint64_t psa_qos_clamp, proc_darwin_role_to_task_role(psa_darwin_role, &role); } - if (apptype != TASK_APPTYPE_NONE || + if (apptype != TASK_APPTYPE_NONE || qos_clamp != THREAD_QOS_UNSPECIFIED || - role != TASK_UNSPECIFIED) { - proc_set_task_spawnpolicy(p->task, apptype, qos_clamp, role, - portwatch_ports, portwatch_count); + role != TASK_UNSPECIFIED || + port_actions->portwatch_count) { + proc_set_task_spawnpolicy(p->task, thread, apptype, qos_clamp, role, + port_actions->portwatch_array, port_actions->portwatch_count); + } + + if (port_actions->registered_count) { + if (mach_ports_register(p->task, port_actions->registered_array, + port_actions->registered_count)) { + return EINVAL; + } + /* mach_ports_register() consumed the array */ + port_actions->registered_array = NULL; + port_actions->registered_count = 0; } - return (0); + return 0; } +static void +exec_port_actions_destroy(struct exec_port_actions *port_actions) +{ + if (port_actions->portwatch_array) { + for (uint32_t i = 0; i < port_actions->portwatch_count; i++) { + ipc_port_t port = NULL; + if ((port = port_actions->portwatch_array[i]) != NULL) { + ipc_port_release_send(port); + } + } + kfree(port_actions->portwatch_array, + port_actions->portwatch_count * sizeof(ipc_port_t *)); + } + + if (port_actions->registered_array) { + for (uint32_t i = 0; i < port_actions->registered_count; i++) { + ipc_port_t port = NULL; + if ((port = port_actions->registered_array[i]) != NULL) { + ipc_port_release_send(port); + } + } + kfree(port_actions->registered_array, + port_actions->registered_count * sizeof(ipc_port_t *)); + } +} /* * exec_handle_port_actions * - * Description: Go through the _posix_port_actions_t contents, - * calling task_set_special_port, task_set_exception_ports - * and/or audit_session_spawnjoin for the current task. + * Description: Go through the _posix_port_actions_t contents, + * calling task_set_special_port, task_set_exception_ports + * and/or audit_session_spawnjoin for the current task. * * Parameters: struct image_params * Image parameter block * * Returns: 0 Success - * EINVAL Failure - * ENOTSUP Illegal posix_spawn attr flag was set + * EINVAL Failure + * ENOTSUP Illegal posix_spawn attr flag was set */ static errno_t -exec_handle_port_actions(struct image_params *imgp, boolean_t * portwatch_present, - ipc_port_t * portwatch_ports) +exec_handle_port_actions(struct image_params *imgp, + struct exec_port_actions *actions) { _posix_spawn_port_actions_t pacts = imgp->ip_px_spa; #if CONFIG_AUDIT @@ -1649,18 +1859,83 @@ exec_handle_port_actions(struct image_params *imgp, boolean_t * portwatch_presen task_t task = get_threadtask(imgp->ip_new_thread); ipc_port_t port = NULL; errno_t ret = 0; - int i; + int i, portwatch_i = 0, registered_i = 0; kern_return_t kr; + boolean_t task_has_watchport_boost = task_has_watchports(current_task()); + boolean_t in_exec = (imgp->ip_flags & IMGPF_EXEC); + boolean_t suid_cred_specified = FALSE; + + for (i = 0; i < pacts->pspa_count; i++) { + act = &pacts->pspa_actions[i]; + + switch (act->port_type) { + case PSPA_SPECIAL: + case PSPA_EXCEPTION: +#if CONFIG_AUDIT + case PSPA_AU_SESSION: +#endif + break; + case PSPA_IMP_WATCHPORTS: + if (++actions->portwatch_count > TASK_MAX_WATCHPORT_COUNT) { + ret = EINVAL; + goto done; + } + break; + case PSPA_REGISTERED_PORTS: + if (++actions->registered_count > TASK_PORT_REGISTER_MAX) { + ret = EINVAL; + goto done; + } + break; + + case PSPA_SUID_CRED: + /* Only a single suid credential can be specified. */ + if (suid_cred_specified) { + ret = EINVAL; + goto done; + } + suid_cred_specified = TRUE; + break; + + default: + ret = EINVAL; + goto done; + } + } + + if (actions->portwatch_count) { + if (in_exec && task_has_watchport_boost) { + ret = EINVAL; + goto done; + } + actions->portwatch_array = + kalloc(sizeof(ipc_port_t *) * actions->portwatch_count); + if (actions->portwatch_array == NULL) { + ret = ENOMEM; + goto done; + } + bzero(actions->portwatch_array, + sizeof(ipc_port_t *) * actions->portwatch_count); + } - *portwatch_present = FALSE; + if (actions->registered_count) { + actions->registered_array = + kalloc(sizeof(ipc_port_t *) * actions->registered_count); + if (actions->registered_array == NULL) { + ret = ENOMEM; + goto done; + } + bzero(actions->registered_array, + sizeof(ipc_port_t *) * actions->registered_count); + } for (i = 0; i < pacts->pspa_count; i++) { act = &pacts->pspa_actions[i]; if (MACH_PORT_VALID(act->new_port)) { kr = ipc_object_copyin(get_task_ipcspace(current_task()), - act->new_port, MACH_MSG_TYPE_COPY_SEND, - (ipc_object_t *) &port); + act->new_port, MACH_MSG_TYPE_COPY_SEND, + (ipc_object_t *) &port, 0, NULL, IPC_KMSG_FLAGS_ALLOW_IMMOVABLE_SEND); if (kr != KERN_SUCCESS) { ret = EINVAL; @@ -1675,15 +1950,17 @@ exec_handle_port_actions(struct image_params *imgp, boolean_t * portwatch_presen case PSPA_SPECIAL: kr = task_set_special_port(task, act->which, port); - if (kr != KERN_SUCCESS) + if (kr != KERN_SUCCESS) { ret = EINVAL; + } break; case PSPA_EXCEPTION: kr = task_set_exception_ports(task, act->mask, port, - act->behavior, act->flavor); - if (kr != KERN_SUCCESS) + act->behavior, act->flavor); + if (kr != KERN_SUCCESS) { ret = EINVAL; + } break; #if CONFIG_AUDIT case PSPA_AU_SESSION: @@ -1696,15 +1973,22 @@ exec_handle_port_actions(struct image_params *imgp, boolean_t * portwatch_presen break; #endif case PSPA_IMP_WATCHPORTS: - if (portwatch_ports != NULL && IPC_PORT_VALID(port)) { - *portwatch_present = TRUE; + if (actions->portwatch_array) { /* hold on to this till end of spawn */ - portwatch_ports[i] = port; + actions->portwatch_array[portwatch_i++] = port; } else { ipc_port_release_send(port); } + break; + case PSPA_REGISTERED_PORTS: + /* hold on to this till end of spawn */ + actions->registered_array[registered_i++] = port; + break; + case PSPA_SUID_CRED: + imgp->ip_sc_port = port; break; + default: ret = EINVAL; break; @@ -1718,9 +2002,10 @@ exec_handle_port_actions(struct image_params *imgp, boolean_t * portwatch_presen } done: - if (0 != ret) + if (0 != ret) { DTRACE_PROC1(spawn__port__failure, mach_port_name_t, act->new_port); - return (ret); + } + return ret; } /* @@ -1748,12 +2033,15 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) int action; proc_t p = vfs_context_proc(imgp->ip_vfs_context); _posix_spawn_file_actions_t px_sfap = imgp->ip_px_sfa; - int ival[2]; /* dummy retval for system calls) */ + int ival[2]; /* dummy retval for system calls) */ +#if CONFIG_AUDIT + struct uthread *uthread = get_bsdthread_info(current_thread()); +#endif for (action = 0; action < px_sfap->psfa_act_count; action++) { - _psfa_action_t *psfa = &px_sfap->psfa_act_acts[ action]; + _psfa_action_t *psfa = &px_sfap->psfa_act_acts[action]; - switch(psfa->psfaa_type) { + switch (psfa->psfaa_type) { case PSFA_OPEN: { /* * Open is different, in that it requires the use of @@ -1782,30 +2070,35 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) VATTR_INIT(vap); /* Mask off all but regular access permissions */ - mode = ((mode &~ p->p_fd->fd_cmask) & ALLPERMS) & ~S_ISTXT; + mode = ((mode & ~p->p_fd->fd_cmask) & ALLPERMS) & ~S_ISTXT; VATTR_SET(vap, va_mode, mode & ACCESSPERMS); + AUDIT_SUBCALL_ENTER(OPEN, p, uthread); + NDINIT(ndp, LOOKUP, OP_OPEN, FOLLOW | AUDITVNPATH1, UIO_SYSSPACE, - CAST_USER_ADDR_T(psfa->psfaa_openargs.psfao_path), - imgp->ip_vfs_context); + CAST_USER_ADDR_T(psfa->psfaa_openargs.psfao_path), + imgp->ip_vfs_context); - error = open1(imgp->ip_vfs_context, - ndp, - psfa->psfaa_openargs.psfao_oflag, - vap, - fileproc_alloc_init, NULL, - ival); + error = open1(imgp->ip_vfs_context, + ndp, + psfa->psfaa_openargs.psfao_oflag, + vap, + fileproc_alloc_init, NULL, + ival); FREE(bufp, M_TEMP); + AUDIT_SUBCALL_EXIT(uthread, error); + /* * If there's an error, or we get the right fd by * accident, then drop out here. This is easier than * reworking all the open code to preallocate fd * slots, and internally taking one as an argument. */ - if (error || ival[0] == psfa->psfaa_filedes) + if (error || ival[0] == psfa->psfaa_filedes) { break; + } origfd = ival[0]; /* @@ -1822,24 +2115,29 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) * can ignore that, since if we didn't get the * fd we wanted, the error will stop us. */ + AUDIT_SUBCALL_ENTER(DUP2, p, uthread); error = dup2(p, &dup2a, ival); - if (error) + AUDIT_SUBCALL_EXIT(uthread, error); + if (error) { break; + } /* * Finally, close the original fd. */ ca.fd = origfd; + AUDIT_SUBCALL_ENTER(CLOSE, p, uthread); error = close_nocancel(p, &ca, ival); - } - break; + AUDIT_SUBCALL_EXIT(uthread, error); + } + break; case PSFA_DUP2: { struct dup2_args dup2a; dup2a.from = psfa->psfaa_filedes; - dup2a.to = psfa->psfaa_openargs.psfao_oflag; + dup2a.to = psfa->psfaa_dup2args.psfad_newfiledes; /* * The dup2() system call implementation sets @@ -1847,18 +2145,67 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) * can ignore that, since if we didn't get the * fd we wanted, the error will stop us. */ + AUDIT_SUBCALL_ENTER(DUP2, p, uthread); error = dup2(p, &dup2a, ival); + AUDIT_SUBCALL_EXIT(uthread, error); + } + break; + + case PSFA_FILEPORT_DUP2: { + ipc_port_t port; + kern_return_t kr; + struct dup2_args dup2a; + struct close_nocancel_args ca; + + if (!MACH_PORT_VALID(psfa->psfaa_fileport)) { + error = EINVAL; + break; + } + + kr = ipc_object_copyin(get_task_ipcspace(current_task()), + psfa->psfaa_fileport, MACH_MSG_TYPE_COPY_SEND, + (ipc_object_t *) &port, 0, NULL, IPC_KMSG_FLAGS_ALLOW_IMMOVABLE_SEND); + + if (kr != KERN_SUCCESS) { + error = EINVAL; + break; + } + + error = fileport_makefd_internal(p, port, 0, ival); + + if (IPC_PORT_NULL != port) { + ipc_port_release_send(port); + } + + if (error || ival[0] == psfa->psfaa_dup2args.psfad_newfiledes) { + break; } - break; + + dup2a.from = ca.fd = ival[0]; + dup2a.to = psfa->psfaa_dup2args.psfad_newfiledes; + AUDIT_SUBCALL_ENTER(DUP2, p, uthread); + error = dup2(p, &dup2a, ival); + AUDIT_SUBCALL_EXIT(uthread, error); + if (error) { + break; + } + + AUDIT_SUBCALL_ENTER(CLOSE, p, uthread); + error = close_nocancel(p, &ca, ival); + AUDIT_SUBCALL_EXIT(uthread, error); + } + break; case PSFA_CLOSE: { struct close_nocancel_args ca; ca.fd = psfa->psfaa_filedes; + AUDIT_SUBCALL_ENTER(CLOSE, p, uthread); error = close_nocancel(p, &ca, ival); - } - break; + AUDIT_SUBCALL_EXIT(uthread, error); + } + break; case PSFA_INHERIT: { struct fcntl_nocancel_args fcntla; @@ -1872,8 +2219,9 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) */ fcntla.fd = psfa->psfaa_filedes; fcntla.cmd = F_GETFD; - if ((error = fcntl_nocancel(p, &fcntla, ival)) != 0) + if ((error = fcntl_nocancel(p, &fcntla, ival)) != 0) { break; + } if ((ival[0] & FD_CLOEXEC) == FD_CLOEXEC) { fcntla.fd = psfa->psfaa_filedes; @@ -1881,9 +2229,40 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) fcntla.arg = ival[0] & ~FD_CLOEXEC; error = fcntl_nocancel(p, &fcntla, ival); } + } + break; - } - break; + case PSFA_CHDIR: { + /* + * Chdir is different, in that it requires the use of + * a path argument, which is normally copied in from + * user space; because of this, we have to support a + * chdir from kernel space that passes an address space + * context of UIO_SYSSPACE, and casts the address + * argument to a user_addr_t. + */ + struct nameidata nd; + + AUDIT_SUBCALL_ENTER(CHDIR, p, uthread); + NDINIT(&nd, LOOKUP, OP_CHDIR, FOLLOW | AUDITVNPATH1, UIO_SYSSPACE, + CAST_USER_ADDR_T(psfa->psfaa_chdirargs.psfac_path), + imgp->ip_vfs_context); + + error = chdir_internal(p, imgp->ip_vfs_context, &nd, 0); + AUDIT_SUBCALL_EXIT(uthread, error); + } + break; + + case PSFA_FCHDIR: { + struct fchdir_args fchdira; + + fchdira.fd = psfa->psfaa_filedes; + + AUDIT_SUBCALL_ENTER(FCHDIR, p, uthread); + error = fchdir(p, &fchdira, ival); + AUDIT_SUBCALL_EXIT(uthread, error); + } + break; default: error = EINVAL; @@ -1895,7 +2274,7 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) if (error) { if (PSFA_OPEN == psfa->psfaa_type) { DTRACE_PROC1(spawn__open__failure, uintptr_t, - psfa->psfaa_openargs.psfao_path); + psfa->psfaa_openargs.psfao_path); } else { DTRACE_PROC1(spawn__fd__failure, int, psfa->psfaa_filedes); } @@ -1903,8 +2282,9 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) } } - if (error != 0 || (psa_flags & POSIX_SPAWN_CLOEXEC_DEFAULT) == 0) - return (error); + if (error != 0 || (psa_flags & POSIX_SPAWN_CLOEXEC_DEFAULT) == 0) { + return error; + } /* * If POSIX_SPAWN_CLOEXEC_DEFAULT is set, behave (during @@ -1922,20 +2302,30 @@ exec_handle_file_actions(struct image_params *imgp, short psa_flags) switch (psfa->psfaa_type) { case PSFA_DUP2: - fd = psfa->psfaa_openargs.psfao_oflag; - /*FALLTHROUGH*/ + case PSFA_FILEPORT_DUP2: + fd = psfa->psfaa_dup2args.psfad_newfiledes; + /*FALLTHROUGH*/ case PSFA_OPEN: case PSFA_INHERIT: *fdflags(p, fd) |= UF_INHERIT; break; case PSFA_CLOSE: + case PSFA_CHDIR: + case PSFA_FCHDIR: + /* + * Although PSFA_FCHDIR does have a file descriptor, it is not + * *creating* one, thus we do not automatically mark it for + * inheritance under POSIX_SPAWN_CLOEXEC_DEFAULT. A client that + * wishes it to be inherited should use the PSFA_INHERIT action + * explicitly. + */ break; } } proc_fdunlock(p); - return (0); + return 0; } #if CONFIG_MACF @@ -1948,20 +2338,23 @@ exec_spawnattr_getmacpolicyinfo(const void *macextensions, const char *policynam const struct _posix_spawn_mac_policy_extensions *psmx = macextensions; int i; - if (psmx == NULL) + if (psmx == NULL) { return NULL; + } for (i = 0; i < psmx->psmx_count; i++) { const _ps_mac_policy_extension_t *extension = &psmx->psmx_extensions[i]; if (strncmp(extension->policyname, policyname, sizeof(extension->policyname)) == 0) { - if (lenp != NULL) + if (lenp != NULL) { *lenp = extension->datalen; + } return extension->datap; } } - if (lenp != NULL) + if (lenp != NULL) { *lenp = 0; + } return NULL; } @@ -1982,8 +2375,9 @@ spawn_copyin_macpolicyinfo(const struct user__posix_spawn_args_desc *px_args, _p } MALLOC(psmx, _posix_spawn_mac_policy_extensions_t, px_args->mac_extensions_size, M_TEMP, M_WAITOK); - if ((error = copyin(px_args->mac_extensions, psmx, px_args->mac_extensions_size)) != 0) + if ((error = copyin(px_args->mac_extensions, psmx, px_args->mac_extensions_size)) != 0) { goto bad; + } size_t extsize = PS_MAC_EXTENSIONS_SIZE(psmx->psmx_count); if (extsize == 0 || extsize > px_args->mac_extensions_size) { @@ -2016,8 +2410,9 @@ spawn_copyin_macpolicyinfo(const struct user__posix_spawn_args_desc *px_args, _p bad: if (psmx != NULL) { - for (i = 0; i < copycnt; i++) + for (i = 0; i < copycnt; i++) { FREE(psmx->psmx_extensions[i].datap, M_TEMP); + } FREE(psmx, M_TEMP); } return error; @@ -2028,16 +2423,19 @@ spawn_free_macpolicyinfo(_posix_spawn_mac_policy_extensions_t psmx) { int i; - if (psmx == NULL) + if (psmx == NULL) { return; - for (i = 0; i < psmx->psmx_count; i++) + } + for (i = 0; i < psmx->psmx_count; i++) { FREE(psmx->psmx_extensions[i].datap, M_TEMP); + } FREE(psmx, M_TEMP); } #endif /* CONFIG_MACF */ #if CONFIG_COALITIONS -static inline void spawn_coalitions_release_all(coalition_t coal[COALITION_NUM_TYPES]) +static inline void +spawn_coalitions_release_all(coalition_t coal[COALITION_NUM_TYPES]) { for (int c = 0; c < COALITION_NUM_TYPES; c++) { if (coal[c]) { @@ -2049,18 +2447,22 @@ static inline void spawn_coalitions_release_all(coalition_t coal[COALITION_NUM_T #endif #if CONFIG_PERSONAS -static int spawn_validate_persona(struct _posix_spawn_persona_info *px_persona) +static int +spawn_validate_persona(struct _posix_spawn_persona_info *px_persona) { int error = 0; struct persona *persona = NULL; int verify = px_persona->pspi_flags & POSIX_SPAWN_PERSONA_FLAGS_VERIFY; - /* - * TODO: rdar://problem/19981151 - * Add entitlement check! - */ - if (!kauth_cred_issuser(kauth_cred_get())) + if (!IOTaskHasEntitlement(current_task(), PERSONA_MGMT_ENTITLEMENT)) { return EPERM; + } + + if (px_persona->pspi_flags & POSIX_SPAWN_PERSONA_GROUPS) { + if (px_persona->pspi_ngroups > NGROUPS_MAX) { + return EINVAL; + } + } persona = persona_lookup(px_persona->pspi_id); if (!persona) { @@ -2082,15 +2484,15 @@ static int spawn_validate_persona(struct _posix_spawn_persona_info *px_persona) } } if (px_persona->pspi_flags & POSIX_SPAWN_PERSONA_GROUPS) { - int ngroups = 0; + unsigned ngroups = 0; gid_t groups[NGROUPS_MAX]; if (persona_get_groups(persona, &ngroups, groups, - px_persona->pspi_ngroups) != 0) { + px_persona->pspi_ngroups) != 0) { error = EINVAL; goto out; } - if (ngroups != (int)px_persona->pspi_ngroups) { + if (ngroups != px_persona->pspi_ngroups) { error = EINVAL; goto out; } @@ -2108,29 +2510,33 @@ static int spawn_validate_persona(struct _posix_spawn_persona_info *px_persona) } out: - if (persona) + if (persona) { persona_put(persona); + } return error; } -static int spawn_persona_adopt(proc_t p, struct _posix_spawn_persona_info *px_persona) +static int +spawn_persona_adopt(proc_t p, struct _posix_spawn_persona_info *px_persona) { int ret; kauth_cred_t cred; struct persona *persona = NULL; int override = !!(px_persona->pspi_flags & POSIX_SPAWN_PERSONA_FLAGS_OVERRIDE); - if (!override) + if (!override) { return persona_proc_adopt_id(p, px_persona->pspi_id, NULL); + } /* * we want to spawn into the given persona, but we want to override * the kauth with a different UID/GID combo */ persona = persona_lookup(px_persona->pspi_id); - if (!persona) + if (!persona) { return ESRCH; + } cred = persona_get_cred(persona); if (!cred) { @@ -2140,24 +2546,24 @@ static int spawn_persona_adopt(proc_t p, struct _posix_spawn_persona_info *px_pe if (px_persona->pspi_flags & POSIX_SPAWN_PERSONA_UID) { cred = kauth_cred_setresuid(cred, - px_persona->pspi_uid, - px_persona->pspi_uid, - px_persona->pspi_uid, - KAUTH_UID_NONE); + px_persona->pspi_uid, + px_persona->pspi_uid, + px_persona->pspi_uid, + KAUTH_UID_NONE); } if (px_persona->pspi_flags & POSIX_SPAWN_PERSONA_GID) { cred = kauth_cred_setresgid(cred, - px_persona->pspi_gid, - px_persona->pspi_gid, - px_persona->pspi_gid); + px_persona->pspi_gid, + px_persona->pspi_gid, + px_persona->pspi_gid); } if (px_persona->pspi_flags & POSIX_SPAWN_PERSONA_GROUPS) { cred = kauth_cred_setgroups(cred, - px_persona->pspi_groups, - px_persona->pspi_ngroups, - px_persona->pspi_gmuid); + px_persona->pspi_groups, + px_persona->pspi_ngroups, + px_persona->pspi_gmuid); } ret = persona_proc_adopt(p, persona, cred); @@ -2168,6 +2574,134 @@ out: } #endif +#if __arm64__ +extern int legacy_footprint_entitlement_mode; +static inline void +proc_legacy_footprint_entitled(proc_t p, task_t task, const char *caller) +{ +#pragma unused(p, caller) + boolean_t legacy_footprint_entitled; + + switch (legacy_footprint_entitlement_mode) { + case LEGACY_FOOTPRINT_ENTITLEMENT_IGNORE: + /* the entitlement is ignored */ + break; + case LEGACY_FOOTPRINT_ENTITLEMENT_IOS11_ACCT: + /* the entitlement grants iOS11 legacy accounting */ + legacy_footprint_entitled = IOTaskHasEntitlement(task, + "com.apple.private.memory.legacy_footprint"); + if (legacy_footprint_entitled) { + task_set_legacy_footprint(task); + } + break; + case LEGACY_FOOTPRINT_ENTITLEMENT_LIMIT_INCREASE: + /* the entitlement grants a footprint limit increase */ + legacy_footprint_entitled = IOTaskHasEntitlement(task, + "com.apple.private.memory.legacy_footprint"); + if (legacy_footprint_entitled) { + task_set_extra_footprint_limit(task); + } + break; + default: + break; + } +} + +static inline void +proc_ios13extended_footprint_entitled(proc_t p, task_t task, const char *caller) +{ +#pragma unused(p, caller) + boolean_t ios13extended_footprint_entitled; + + /* the entitlement grants a footprint limit increase */ + ios13extended_footprint_entitled = IOTaskHasEntitlement(task, + "com.apple.developer.memory.ios13extended_footprint"); + if (ios13extended_footprint_entitled) { + task_set_ios13extended_footprint_limit(task); + } +} +#endif /* __arm64__ */ + +/* + * Apply a modification on the proc's kauth cred until it converges. + * + * `update` consumes its argument to return a new kauth cred. + */ +static void +apply_kauth_cred_update(proc_t p, + kauth_cred_t (^update)(kauth_cred_t orig_cred)) +{ + kauth_cred_t my_cred, my_new_cred; + + my_cred = kauth_cred_proc_ref(p); + for (;;) { + my_new_cred = update(my_cred); + if (my_cred == my_new_cred) { + kauth_cred_unref(&my_new_cred); + break; + } + + /* try update cred on proc */ + proc_ucred_lock(p); + + if (p->p_ucred == my_cred) { + /* base pointer didn't change, donate our ref */ + p->p_ucred = my_new_cred; + PROC_UPDATE_CREDS_ONPROC(p); + proc_ucred_unlock(p); + + /* drop p->p_ucred reference */ + kauth_cred_unref(&my_cred); + break; + } + + /* base pointer changed, retry */ + my_cred = p->p_ucred; + kauth_cred_ref(my_cred); + proc_ucred_unlock(p); + + kauth_cred_unref(&my_new_cred); + } +} + +static int +spawn_posix_cred_adopt(proc_t p, + struct _posix_spawn_posix_cred_info *px_pcred_info) +{ + int error = 0; + + if (px_pcred_info->pspci_flags & POSIX_SPAWN_POSIX_CRED_GID) { + struct setgid_args args = { + .gid = px_pcred_info->pspci_gid, + }; + error = setgid(p, &args, NULL); + if (error) { + return error; + } + } + + if (px_pcred_info->pspci_flags & POSIX_SPAWN_POSIX_CRED_GROUPS) { + error = setgroups_internal(p, + px_pcred_info->pspci_ngroups, + px_pcred_info->pspci_groups, + px_pcred_info->pspci_gmuid); + if (error) { + return error; + } + } + + if (px_pcred_info->pspci_flags & POSIX_SPAWN_POSIX_CRED_UID) { + struct setuid_args args = { + .uid = px_pcred_info->pspci_uid, + }; + error = setuid(p, &args, NULL); + if (error) { + return error; + } + } + return 0; +} + /* * posix_spawn * @@ -2188,6 +2722,7 @@ out: * exec_activate_image:ENAMETOOLONG Filename too long * exec_activate_image:ENOEXEC Executable file format error * exec_activate_image:ETXTBSY Text file busy [misuse of error code] + * exec_activate_image:EAUTH Image decryption failed * exec_activate_image:EBADEXEC The executable is corrupt/unknown * exec_activate_image:??? * mac_execve_enter:??? @@ -2199,14 +2734,14 @@ out: int posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) { - proc_t p = ap; /* quiet bogus GCC vfork() warning */ + proc_t p = ap; /* quiet bogus GCC vfork() warning */ user_addr_t pid = uap->pid; - int ival[2]; /* dummy retval for setpgid() */ - char *bufp = NULL; + int ival[2]; /* dummy retval for setpgid() */ + char *bufp = NULL; struct image_params *imgp; struct vnode_attr *vap; struct vnode_attr *origvap; - struct uthread *uthread = 0; /* compiler complains if not set to 0*/ + struct uthread *uthread = 0; /* compiler complains if not set to 0*/ int error, sig; int is_64 = IS_64BIT_PROCESS(p); struct vfs_context context; @@ -2218,18 +2753,19 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) boolean_t spawn_no_exec = FALSE; boolean_t proc_transit_set = TRUE; boolean_t exec_done = FALSE; - int portwatch_count = 0; - ipc_port_t * portwatch_ports = NULL; + struct exec_port_actions port_actions = { }; vm_size_t px_sa_offset = offsetof(struct _posix_spawnattr, psa_ports); + task_t old_task = current_task(); task_t new_task = NULL; boolean_t should_release_proc_ref = FALSE; void *inherit = NULL; #if CONFIG_PERSONAS struct _posix_spawn_persona_info *px_persona = NULL; #endif + struct _posix_spawn_posix_cred_info *px_pcred_info = NULL; /* - * Allocate a big chunk for locals instead of using stack since these + * Allocate a big chunk for locals instead of using stack since these * structures are pretty big. */ MALLOC(bufp, char *, (sizeof(*imgp) + sizeof(*vap) + sizeof(*origvap)), M_TEMP, M_WAITOK | M_ZERO); @@ -2248,14 +2784,16 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) imgp->ip_vattr = vap; imgp->ip_origvattr = origvap; imgp->ip_vfs_context = &context; - imgp->ip_flags = (is_64 ? IMGPF_WAS_64BIT : IMGPF_NONE); + imgp->ip_flags = (is_64 ? IMGPF_WAS_64BIT_ADDR : IMGPF_NONE); imgp->ip_seg = (is_64 ? UIO_USERSPACE64 : UIO_USERSPACE32); imgp->ip_mac_return = 0; imgp->ip_px_persona = NULL; + imgp->ip_px_pcred_info = NULL; imgp->ip_cs_error = OS_REASON_NULL; + imgp->ip_simulator_binary = IMGPF_SB_DEFAULT; if (uap->adesc != USER_ADDR_NULL) { - if(is_64) { + if (is_64) { error = copyin(uap->adesc, &px_args, sizeof(px_args)); } else { struct user32__posix_spawn_args_desc px_args32; @@ -2278,21 +2816,25 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) px_args.coal_info = CAST_USER_ADDR_T(px_args32.coal_info); px_args.persona_info_size = px_args32.persona_info_size; px_args.persona_info = CAST_USER_ADDR_T(px_args32.persona_info); + px_args.posix_cred_info_size = px_args32.posix_cred_info_size; + px_args.posix_cred_info = CAST_USER_ADDR_T(px_args32.posix_cred_info); } - if (error) + if (error) { goto bad; + } if (px_args.attr_size != 0) { - /* - * We are not copying the port_actions pointer, - * because we already have it from px_args. + /* + * We are not copying the port_actions pointer, + * because we already have it from px_args. * This is a bit fragile: */ - if ((error = copyin(px_args.attrp, &px_sa, px_sa_offset) != 0)) - goto bad; - - bzero( (void *)( (unsigned long) &px_sa + px_sa_offset), sizeof(px_sa) - px_sa_offset ); + if ((error = copyin(px_args.attrp, &px_sa, px_sa_offset)) != 0) { + goto bad; + } + + bzero((void *)((unsigned long) &px_sa + px_sa_offset), sizeof(px_sa) - px_sa_offset ); imgp->ip_px_sa = &px_sa; } @@ -2312,9 +2854,10 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) } imgp->ip_px_sfa = px_sfap; - if ((error = copyin(px_args.file_actions, px_sfap, - px_args.file_actions_size)) != 0) + if ((error = copyin(px_args.file_actions, px_sfap, + px_args.file_actions_size)) != 0) { goto bad; + } /* Verify that the action count matches the struct size */ size_t psfsize = PSF_ACTIONS_SIZE(px_sfap->psfa_act_count); @@ -2326,22 +2869,23 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) if (px_args.port_actions_size != 0) { /* Limit port_actions to one page of data */ if (px_args.port_actions_size < PS_PORT_ACTIONS_SIZE(1) || - px_args.port_actions_size > PAGE_SIZE) { + px_args.port_actions_size > PAGE_SIZE) { error = EINVAL; goto bad; } - MALLOC(px_spap, _posix_spawn_port_actions_t, - px_args.port_actions_size, M_TEMP, M_WAITOK); + MALLOC(px_spap, _posix_spawn_port_actions_t, + px_args.port_actions_size, M_TEMP, M_WAITOK); if (px_spap == NULL) { error = ENOMEM; goto bad; } imgp->ip_px_spa = px_spap; - if ((error = copyin(px_args.port_actions, px_spap, - px_args.port_actions_size)) != 0) + if ((error = copyin(px_args.port_actions, px_spap, + px_args.port_actions_size)) != 0) { goto bad; + } /* Verify that the action count matches the struct size */ size_t pasize = PS_PORT_ACTIONS_SIZE(px_spap->pspa_count); @@ -2359,7 +2903,7 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) goto bad; } - MALLOC(px_persona, struct _posix_spawn_persona_info *, px_args.persona_info_size, M_TEMP, M_WAITOK|M_ZERO); + MALLOC(px_persona, struct _posix_spawn_persona_info *, px_args.persona_info_size, M_TEMP, M_WAITOK | M_ZERO); if (px_persona == NULL) { error = ENOMEM; goto bad; @@ -2367,16 +2911,52 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) imgp->ip_px_persona = px_persona; if ((error = copyin(px_args.persona_info, px_persona, - px_args.persona_info_size)) != 0) + px_args.persona_info_size)) != 0) { goto bad; - if ((error = spawn_validate_persona(px_persona)) != 0) + } + if ((error = spawn_validate_persona(px_persona)) != 0) { goto bad; + } } #endif + /* copy in the posix cred info */ + if (px_args.posix_cred_info_size != 0 && px_args.posix_cred_info != 0) { + /* for now, we need the exact same struct in user space */ + if (px_args.posix_cred_info_size != sizeof(*px_pcred_info)) { + error = ERANGE; + goto bad; + } + + if (!kauth_cred_issuser(kauth_cred_get())) { + error = EPERM; + goto bad; + } + + MALLOC(px_pcred_info, struct _posix_spawn_posix_cred_info *, + px_args.posix_cred_info_size, M_TEMP, M_WAITOK | M_ZERO); + if (px_pcred_info == NULL) { + error = ENOMEM; + goto bad; + } + imgp->ip_px_pcred_info = px_pcred_info; + + if ((error = copyin(px_args.posix_cred_info, px_pcred_info, + px_args.posix_cred_info_size)) != 0) { + goto bad; + } + + if (px_pcred_info->pspci_flags & POSIX_SPAWN_POSIX_CRED_GROUPS) { + if (px_pcred_info->pspci_ngroups > NGROUPS_MAX) { + error = EINVAL; + goto bad; + } + } + } #if CONFIG_MACF if (px_args.mac_extensions_size != 0) { - if ((error = spawn_copyin_macpolicyinfo(&px_args, (_posix_spawn_mac_policy_extensions_t *)&imgp->ip_px_smpx)) != 0) + if ((error = spawn_copyin_macpolicyinfo(&px_args, (_posix_spawn_mac_policy_extensions_t *)&imgp->ip_px_smpx)) != 0) { goto bad; + } } #endif /* CONFIG_MACF */ } @@ -2390,8 +2970,15 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) * which depends on it. */ if (uthread->uu_flag & UT_VFORK) { - error = EINVAL; - goto bad; + error = EINVAL; + goto bad; + } + + if (imgp->ip_px_sa != NULL) { + struct _posix_spawnattr *psa = (struct _posix_spawnattr *) imgp->ip_px_sa; + if ((error = exec_validate_spawnattr_policy(psa->psa_apptype)) != 0) { + goto bad; + } } /* @@ -2401,8 +2988,7 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) * which is one of the most expensive things about using fork() * and execve(). */ - if (imgp->ip_px_sa == NULL || !(px_sa.psa_flags & POSIX_SPAWN_SETEXEC)){ - + if (imgp->ip_px_sa == NULL || !(px_sa.psa_flags & POSIX_SPAWN_SETEXEC)) { /* Set the new task's coalition, if it is requested. */ coalition_t coal[COALITION_NUM_TYPES] = { COALITION_NULL }; #if CONFIG_COALITIONS @@ -2411,17 +2997,20 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) struct _posix_spawn_coalition_info coal_info; int coal_role[COALITION_NUM_TYPES]; - if (imgp->ip_px_sa == NULL || !px_args.coal_info) + if (imgp->ip_px_sa == NULL || !px_args.coal_info) { goto do_fork1; + } memset(&coal_info, 0, sizeof(coal_info)); - if (px_args.coal_info_size > sizeof(coal_info)) + if (px_args.coal_info_size > sizeof(coal_info)) { px_args.coal_info_size = sizeof(coal_info); + } error = copyin(px_args.coal_info, - &coal_info, px_args.coal_info_size); - if (error != 0) + &coal_info, px_args.coal_info_size); + if (error != 0) { goto bad; + } ncoals = 0; for (i = 0; i < COALITION_NUM_TYPES; i++) { @@ -2434,8 +3023,8 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) */ if (!task_is_in_privileged_coalition(p->task, i)) { coal_dbg("ERROR: %d not in privilegd " - "coalition of type %d", - p->p_pid, i); + "coalition of type %d", + p->p_pid, i); spawn_coalitions_release_all(coal); error = EPERM; goto bad; @@ -2450,7 +3039,7 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) coal[i] = coalition_find_and_activate_by_id(cid); if (coal[i] == COALITION_NULL) { coal_dbg("could not find coalition id:%llu " - "(perhaps it has been terminated or reaped)", cid); + "(perhaps it has been terminated or reaped)", cid); /* * release any other coalition's we * may have a reference to @@ -2461,7 +3050,7 @@ posix_spawn(proc_t ap, struct posix_spawn_args *uap, int32_t *retval) } if (coalition_type(coal[i]) != i) { coal_dbg("coalition with id:%lld is not of type:%d" - " (it's type:%d)", cid, i, coalition_type(coal[i])); + " (it's type:%d)", cid, i, coalition_type(coal[i])); error = ESRCH; goto bad; } @@ -2502,10 +3091,11 @@ do_fork1: /* set the roles of this task within each given coalition */ if (error == 0) { kr = coalitions_set_roles(coal, new_task, coal_role); - if (kr != KERN_SUCCESS) + if (kr != KERN_SUCCESS) { error = EINVAL; + } if (kdebug_debugid_enabled(MACHDBG_CODE(DBG_MACH_COALITION, - MACH_COALITION_ADOPT))) { + MACH_COALITION_ADOPT))) { for (i = 0; i < COALITION_NUM_TYPES; i++) { if (coal[i] != COALITION_NULL) { /* @@ -2513,10 +3103,10 @@ do_fork1: * will get truncated to 32 bits */ KDBG_RELEASE(MACHDBG_CODE( - DBG_MACH_COALITION, - MACH_COALITION_ADOPT), - coalition_id(coal[i]), - get_task_uniqueid(new_task)); + DBG_MACH_COALITION, + MACH_COALITION_ADOPT), + coalition_id(coal[i]), + get_task_uniqueid(new_task)); } } } @@ -2528,33 +3118,8 @@ do_fork1: if (error != 0) { goto bad; } - imgp->ip_flags |= IMGPF_SPAWN; /* spawn w/o exec */ - spawn_no_exec = TRUE; /* used in later tests */ - -#if CONFIG_PERSONAS - /* - * If the parent isn't in a persona (launchd), and - * hasn't specified a new persona for the process, - * then we'll put the process into the system persona - * - * TODO: this will have to be re-worked because as of - * now, without any launchd adoption, the resulting - * xpcproxy process will not have sufficient - * privileges to setuid/gid. - */ -#if 0 - if (!proc_has_persona(p) && imgp->ip_px_persona == NULL) { - MALLOC(px_persona, struct _posix_spawn_persona_info *, - sizeof(*px_persona), M_TEMP, M_WAITOK|M_ZERO); - if (px_persona == NULL) { - error = ENOMEM; - goto bad; - } - px_persona->pspi_id = persona_get_id(g_system_persona); - imgp->ip_px_persona = px_persona; - } -#endif /* 0 */ -#endif /* CONFIG_PERSONAS */ + imgp->ip_flags |= IMGPF_SPAWN; /* spawn w/o exec */ + spawn_no_exec = TRUE; /* used in later tests */ } else { /* * For execve case, create a new task and thread @@ -2581,8 +3146,13 @@ do_fork1: * During exec any transition from new_task -> proc is fine, but don't allow * transition from proc->task, since it will modify old_task. */ - imgp->ip_new_thread = fork_create_child(current_task(), - NULL, p, FALSE, p->p_flag & P_LP64, TRUE); + imgp->ip_new_thread = fork_create_child(old_task, + NULL, + p, + FALSE, + p->p_flag & P_LP64, + task_get_64bit_data(old_task), + TRUE); /* task and thread ref returned by fork_create_child */ if (imgp->ip_new_thread == NULL) { error = ENOMEM; @@ -2595,7 +3165,7 @@ do_fork1: if (spawn_no_exec) { p = (proc_t)get_bsdthreadtask_info(imgp->ip_new_thread); - + /* * We had to wait until this point before firing the * proc:::create probe, otherwise p would not point to the @@ -2606,7 +3176,7 @@ do_fork1: assert(p != NULL); context.vc_thread = imgp->ip_new_thread; - context.vc_ucred = p->p_ucred; /* XXX must NOT be kauth_cred_get() */ + context.vc_ucred = p->p_ucred; /* XXX must NOT be kauth_cred_get() */ /* * Post fdcopy(), pre exec_handle_sugid() - this is where we want @@ -2621,60 +3191,47 @@ do_fork1: * The POSIX_SPAWN_CLOEXEC_DEFAULT flag * is handled in exec_handle_file_actions(). */ - if ((error = exec_handle_file_actions(imgp, - imgp->ip_px_sa != NULL ? px_sa.psa_flags : 0)) != 0) +#if CONFIG_AUDIT + /* + * The file actions auditing can overwrite the upath of + * AUE_POSIX_SPAWN audit record. Save the audit record. + */ + struct kaudit_record *save_uu_ar = uthread->uu_ar; + uthread->uu_ar = NULL; +#endif + error = exec_handle_file_actions(imgp, + imgp->ip_px_sa != NULL ? px_sa.psa_flags : 0); +#if CONFIG_AUDIT + /* Restore the AUE_POSIX_SPAWN audit record. */ + uthread->uu_ar = save_uu_ar; +#endif + if (error != 0) { goto bad; + } } /* Has spawn port actions? */ if (imgp->ip_px_spa != NULL) { - boolean_t is_adaptive = FALSE; - boolean_t portwatch_present = FALSE; - - /* Will this process become adaptive? The apptype isn't ready yet, so we can't look there. */ - if (imgp->ip_px_sa != NULL && px_sa.psa_apptype == POSIX_SPAWN_PROC_TYPE_DAEMON_ADAPTIVE) - is_adaptive = TRUE; - +#if CONFIG_AUDIT /* - * portwatch only: - * Allocate a place to store the ports we want to bind to the new task - * We can't bind them until after the apptype is set. + * Do the same for the port actions as we did for the file + * actions. Save the AUE_POSIX_SPAWN audit record. */ - if (px_spap->pspa_count != 0 && is_adaptive) { - portwatch_count = px_spap->pspa_count; - MALLOC(portwatch_ports, ipc_port_t *, (sizeof(ipc_port_t) * portwatch_count), M_TEMP, M_WAITOK | M_ZERO); - } else { - portwatch_ports = NULL; - } - - if ((error = exec_handle_port_actions(imgp, &portwatch_present, portwatch_ports)) != 0) + struct kaudit_record *save_uu_ar = uthread->uu_ar; + uthread->uu_ar = NULL; +#endif + error = exec_handle_port_actions(imgp, &port_actions); +#if CONFIG_AUDIT + /* Restore the AUE_POSIX_SPAWN audit record. */ + uthread->uu_ar = save_uu_ar; +#endif + if (error != 0) { goto bad; - - if (portwatch_present == FALSE && portwatch_ports != NULL) { - FREE(portwatch_ports, M_TEMP); - portwatch_ports = NULL; - portwatch_count = 0; } } /* Has spawn attr? */ if (imgp->ip_px_sa != NULL) { - /* - * Set the process group ID of the child process; this has - * to happen before the image activation. - */ - if (px_sa.psa_flags & POSIX_SPAWN_SETPGROUP) { - struct setpgid_args spga; - spga.pid = p->p_pid; - spga.pgid = px_sa.psa_pgroup; - /* - * Effectively, call setpgid() system call; works - * because there are no pointer arguments. - */ - if((error = setpgid(p, &spga, ival)) != 0) - goto bad; - } - /* * Reset UID/GID to parent's RUID/RGID; This works only * because the operation occurs *after* the vfork() and @@ -2688,35 +3245,33 @@ do_fork1: * proc's ucred lock. This prevents others from accessing * a garbage credential. */ - while (px_sa.psa_flags & POSIX_SPAWN_RESETIDS) { - kauth_cred_t my_cred = kauth_cred_proc_ref(p); - kauth_cred_t my_new_cred = kauth_cred_setuidgid(my_cred, kauth_cred_getruid(my_cred), kauth_cred_getrgid(my_cred)); + if (px_sa.psa_flags & POSIX_SPAWN_RESETIDS) { + apply_kauth_cred_update(p, ^kauth_cred_t (kauth_cred_t my_cred){ + return kauth_cred_setuidgid(my_cred, + kauth_cred_getruid(my_cred), + kauth_cred_getrgid(my_cred)); + }); + } - if (my_cred == my_new_cred) { - kauth_cred_unref(&my_cred); - break; + if (imgp->ip_px_pcred_info) { + if (!spawn_no_exec) { + error = ENOTSUP; + goto bad; } - /* update cred on proc */ - proc_ucred_lock(p); - - if (p->p_ucred != my_cred) { - proc_ucred_unlock(p); - kauth_cred_unref(&my_new_cred); - continue; + error = spawn_posix_cred_adopt(p, imgp->ip_px_pcred_info); + if (error != 0) { + goto bad; } - - /* donate cred reference on my_new_cred to p->p_ucred */ - p->p_ucred = my_new_cred; - PROC_UPDATE_CREDS_ONPROC(p); - proc_ucred_unlock(p); - - /* drop additional reference that was taken on the previous cred */ - kauth_cred_unref(&my_cred); } #if CONFIG_PERSONAS - if (spawn_no_exec && imgp->ip_px_persona != NULL) { + if (imgp->ip_px_persona != NULL) { + if (!spawn_no_exec) { + error = ENOTSUP; + goto bad; + } + /* * If we were asked to spawn a process into a new persona, * do the credential switch now (which may override the UID/GID @@ -2726,8 +3281,9 @@ do_fork1: * being executed. */ error = spawn_persona_adopt(p, imgp->ip_px_persona); - if (error != 0) + if (error != 0) { goto bad; + } } #endif /* CONFIG_PERSONAS */ #if !SECURE_KERNEL @@ -2741,20 +3297,30 @@ do_fork1: * useful or necessary to disable ASLR on a per-process * basis for unit testing and debugging. */ - if (px_sa.psa_flags & _POSIX_SPAWN_DISABLE_ASLR) + if (px_sa.psa_flags & _POSIX_SPAWN_DISABLE_ASLR) { OSBitOrAtomic(P_DISABLE_ASLR, &p->p_flag); + } #endif /* !SECURE_KERNEL */ /* Randomize high bits of ASLR slide */ - if (px_sa.psa_flags & _POSIX_SPAWN_HIGH_BITS_ASLR) + if (px_sa.psa_flags & _POSIX_SPAWN_HIGH_BITS_ASLR) { imgp->ip_flags |= IMGPF_HIGH_BITS_ASLR; + } +#if !SECURE_KERNEL /* * Forcibly disallow execution from data pages for the spawned process * even if it would otherwise be permitted by the architecture default. */ - if (px_sa.psa_flags & _POSIX_SPAWN_ALLOW_DATA_EXEC) + if (px_sa.psa_flags & _POSIX_SPAWN_ALLOW_DATA_EXEC) { imgp->ip_flags |= IMGPF_ALLOW_DATA_EXEC; + } +#endif /* !SECURE_KERNEL */ + + if ((px_sa.psa_apptype & POSIX_SPAWN_PROC_TYPE_MASK) == + POSIX_SPAWN_PROC_TYPE_DRIVER) { + imgp->ip_flags |= IMGPF_DRIVER; + } } /* @@ -2762,10 +3328,11 @@ do_fork1: * _POSIX_SPAWN_DISABLE_ASLR attribute was found above or if * P_DISABLE_ASLR was inherited from the parent process. */ - if (p->p_flag & P_DISABLE_ASLR) + if (p->p_flag & P_DISABLE_ASLR) { imgp->ip_flags |= IMGPF_DISABLE_ASLR; + } - /* + /* * Clear transition flag so we won't hang if exec_activate_image() causes * an automount (and launchd does a proc sysctl to service it). * @@ -2776,11 +3343,12 @@ do_fork1: proc_transit_set = 0; } -#if MAC_SPAWN /* XXX */ +#if MAC_SPAWN /* XXX */ if (uap->mac_p != USER_ADDR_NULL) { error = mac_execve_enter(uap->mac_p, imgp); - if (error) + if (error) { goto bad; + } } #endif @@ -2788,11 +3356,24 @@ do_fork1: * Activate the image */ error = exec_activate_image(imgp); - +#if defined(HAS_APPLE_PAC) + ml_task_set_disable_user_jop(new_task, imgp->ip_flags & IMGPF_NOJOP ? TRUE : FALSE); + ml_thread_set_disable_user_jop(imgp->ip_new_thread, imgp->ip_flags & IMGPF_NOJOP ? TRUE : FALSE); +#endif + if (error == 0 && !spawn_no_exec) { - p = proc_exec_switch_task(p, current_task(), new_task, imgp->ip_new_thread); + p = proc_exec_switch_task(p, old_task, new_task, imgp->ip_new_thread); /* proc ref returned */ should_release_proc_ref = TRUE; + + /* + * Need to transfer pending watch port boosts to the new task while still making + * sure that the old task remains in the importance linkage. Create an importance + * linkage from old task to new task, then switch the task importance base + * of old task and new task. After the switch the port watch boost will be + * boosting the new task and new task will be donating importance to old task. + */ + inherit = ipc_importance_exec_switch_task(old_task, new_task); } if (error == 0) { @@ -2803,26 +3384,53 @@ do_fork1: error = ENOEXEC; } - /* - * If we have a spawn attr, and it contains signal related flags, - * the we need to process them in the "context" of the new child - * process, so we have to process it following image activation, - * prior to making the thread runnable in user space. This is - * necessitated by some signal information being per-thread rather - * than per-process, and we don't have the new allocation in hand - * until after the image is activated. - */ if (!error && imgp->ip_px_sa != NULL) { thread_t child_thread = imgp->ip_new_thread; uthread_t child_uthread = get_bsdthread_info(child_thread); + /* + * Because of POSIX_SPAWN_SETEXEC, we need to handle this after image + * activation, else when image activation fails (before the point of no + * return) would leave the parent process in a modified state. + */ + if (px_sa.psa_flags & POSIX_SPAWN_SETPGROUP) { + struct setpgid_args spga; + spga.pid = p->p_pid; + spga.pgid = px_sa.psa_pgroup; + /* + * Effectively, call setpgid() system call; works + * because there are no pointer arguments. + */ + if ((error = setpgid(p, &spga, ival)) != 0) { + goto bad; + } + } + + if (px_sa.psa_flags & POSIX_SPAWN_SETSID) { + error = setsid_internal(p); + if (error != 0) { + goto bad; + } + } + + /* + * If we have a spawn attr, and it contains signal related flags, + * the we need to process them in the "context" of the new child + * process, so we have to process it following image activation, + * prior to making the thread runnable in user space. This is + * necessitated by some signal information being per-thread rather + * than per-process, and we don't have the new allocation in hand + * until after the image is activated. + */ + /* * Mask a list of signals, instead of them being unmasked, if * they were unmasked in the parent; note that some signals * are not maskable. */ - if (px_sa.psa_flags & POSIX_SPAWN_SETSIGMASK) + if (px_sa.psa_flags & POSIX_SPAWN_SETSIGMASK) { child_uthread->uu_sigmask = (px_sa.psa_sigmask & ~sigcantmask); + } /* * Default a list of signals instead of ignoring them, if * they were ignored in the parent. Note that we pass @@ -2835,9 +3443,10 @@ do_fork1: vec.sa_tramp = 0; vec.sa_mask = 0; vec.sa_flags = 0; - for (sig = 1; sig < NSIG; sig++) - if (px_sa.psa_sigdefault & (1 << (sig-1))) { + for (sig = 1; sig < NSIG; sig++) { + if (px_sa.psa_sigdefault & (1 << (sig - 1))) { error = setsigvec(p, child_thread, sig, &vec, spawn_no_exec); + } } } @@ -2855,10 +3464,19 @@ do_fork1: * whomever is turning it on could just as easily choose not to do so. */ error = proc_set_task_ruse_cpu(p->task, - TASK_POLICY_RESOURCE_ATTRIBUTE_NOTIFY_EXC, - px_sa.psa_cpumonitor_percent, - px_sa.psa_cpumonitor_interval * NSEC_PER_SEC, - 0, TRUE); + TASK_POLICY_RESOURCE_ATTRIBUTE_NOTIFY_EXC, + px_sa.psa_cpumonitor_percent, + px_sa.psa_cpumonitor_interval * NSEC_PER_SEC, + 0, TRUE); + } + + + if (px_pcred_info && + (px_pcred_info->pspci_flags & POSIX_SPAWN_POSIX_CRED_LOGIN)) { + /* + * setlogin() must happen after setsid() + */ + setlogin_internal(p, px_pcred_info->pspci_login); } } @@ -2867,30 +3485,37 @@ bad: if (error == 0) { /* reset delay idle sleep status if set */ #if !CONFIG_EMBEDDED - if ((p->p_flag & P_DELAYIDLESLEEP) == P_DELAYIDLESLEEP) + if ((p->p_flag & P_DELAYIDLESLEEP) == P_DELAYIDLESLEEP) { OSBitAndAtomic(~((uint32_t)P_DELAYIDLESLEEP), &p->p_flag); + } #endif /* !CONFIG_EMBEDDED */ /* upon successful spawn, re/set the proc control state */ if (imgp->ip_px_sa != NULL) { switch (px_sa.psa_pcontrol) { - case POSIX_SPAWN_PCONTROL_THROTTLE: - p->p_pcaction = P_PCTHROTTLE; - break; - case POSIX_SPAWN_PCONTROL_SUSPEND: - p->p_pcaction = P_PCSUSP; - break; - case POSIX_SPAWN_PCONTROL_KILL: - p->p_pcaction = P_PCKILL; - break; - case POSIX_SPAWN_PCONTROL_NONE: - default: - p->p_pcaction = 0; - break; - }; + case POSIX_SPAWN_PCONTROL_THROTTLE: + p->p_pcaction = P_PCTHROTTLE; + break; + case POSIX_SPAWN_PCONTROL_SUSPEND: + p->p_pcaction = P_PCSUSP; + break; + case POSIX_SPAWN_PCONTROL_KILL: + p->p_pcaction = P_PCKILL; + break; + case POSIX_SPAWN_PCONTROL_NONE: + default: + p->p_pcaction = 0; + break; + } + ; } exec_resettextvp(p, imgp); - + #if CONFIG_MEMORYSTATUS + /* Set jetsam priority for DriverKit processes */ + if (px_sa.psa_apptype == POSIX_SPAWN_PROC_TYPE_DRIVER) { + px_sa.psa_priority = JETSAM_PRIORITY_DRIVER_APPLE; + } + /* Has jetsam attributes? */ if (imgp->ip_px_sa != NULL && (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_SET)) { /* @@ -2901,24 +3526,52 @@ bad: * we attempt to mimic previous behavior by forcing the BG limit data into the * inactive/non-fatal mode and force the active slots to hold system_wide/fatal mode. */ + if (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_HIWATER_BACKGROUND) { - memorystatus_update(p, px_sa.psa_priority, 0, - (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_USE_EFFECTIVE_PRIORITY), - TRUE, - -1, TRUE, - px_sa.psa_memlimit_inactive, FALSE); + memorystatus_update(p, px_sa.psa_priority, 0, FALSE, /* assertion priority */ + (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_USE_EFFECTIVE_PRIORITY), + TRUE, + -1, TRUE, + px_sa.psa_memlimit_inactive, FALSE); } else { - memorystatus_update(p, px_sa.psa_priority, 0, - (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_USE_EFFECTIVE_PRIORITY), - TRUE, - px_sa.psa_memlimit_active, - (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_MEMLIMIT_ACTIVE_FATAL), - px_sa.psa_memlimit_inactive, - (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_MEMLIMIT_INACTIVE_FATAL)); + memorystatus_update(p, px_sa.psa_priority, 0, FALSE, /* assertion priority */ + (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_USE_EFFECTIVE_PRIORITY), + TRUE, + px_sa.psa_memlimit_active, + (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_MEMLIMIT_ACTIVE_FATAL), + px_sa.psa_memlimit_inactive, + (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_MEMLIMIT_INACTIVE_FATAL)); } + } + /* Has jetsam relaunch behavior? */ + if (imgp->ip_px_sa != NULL && (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_RELAUNCH_BEHAVIOR_MASK)) { + /* + * Launchd has passed in data indicating the behavior of this process in response to jetsam. + * This data would be used by the jetsam subsystem to determine the position and protection + * offered to this process on dirty -> clean transitions. + */ + int relaunch_flags = P_MEMSTAT_RELAUNCH_UNKNOWN; + switch (px_sa.psa_jetsam_flags & POSIX_SPAWN_JETSAM_RELAUNCH_BEHAVIOR_MASK) { + case POSIX_SPAWN_JETSAM_RELAUNCH_BEHAVIOR_LOW: + relaunch_flags = P_MEMSTAT_RELAUNCH_LOW; + break; + case POSIX_SPAWN_JETSAM_RELAUNCH_BEHAVIOR_MED: + relaunch_flags = P_MEMSTAT_RELAUNCH_MED; + break; + case POSIX_SPAWN_JETSAM_RELAUNCH_BEHAVIOR_HIGH: + relaunch_flags = P_MEMSTAT_RELAUNCH_HIGH; + break; + default: + break; + } + memorystatus_relaunch_flags_update(p, relaunch_flags); } + #endif /* CONFIG_MEMORYSTATUS */ + if (imgp->ip_px_sa != NULL && px_sa.psa_thread_limit > 0) { + task_set_thread_limit(new_task, (uint16_t)px_sa.psa_thread_limit); + } } /* @@ -2930,8 +3583,9 @@ bad: * before check_for_signature(), which uses psignal. */ if (spawn_no_exec) { - if (proc_transit_set) + if (proc_transit_set) { proc_transend(p, 0); + } /* * Drop the signal lock on the child which was taken on our @@ -2945,8 +3599,9 @@ bad: } /* flag exec has occurred, notify only if it has not failed due to FP Key error */ - if (!error && ((p->p_lflag & P_LTERM_DECRYPTFAIL) == 0)) + if (!error && ((p->p_lflag & P_LTERM_DECRYPTFAIL) == 0)) { proc_knote(p, NOTE_EXEC); + } if (error == 0) { @@ -2959,14 +3614,34 @@ bad: error = proc_transstart(p, 0, 0); if (error == 0) { - task_bank_init(get_threadtask(imgp->ip_new_thread)); + task_bank_init(new_task); proc_transend(p, 0); } + +#if __arm64__ + proc_legacy_footprint_entitled(p, new_task, __FUNCTION__); + proc_ios13extended_footprint_entitled(p, new_task, __FUNCTION__); +#endif /* __arm64__ */ } /* Inherit task role from old task to new task for exec */ if (error == 0 && !spawn_no_exec) { - proc_inherit_task_role(get_threadtask(imgp->ip_new_thread), current_task()); + proc_inherit_task_role(new_task, old_task); + } + +#if CONFIG_ARCADE + if (error == 0) { + /* + * Check to see if we need to trigger an arcade upcall AST now + * that the vnode has been reset on the task. + */ + arcade_prepare(new_task, imgp->ip_new_thread); + } +#endif /* CONFIG_ARCADE */ + + /* Clear the initial wait on the thread before handling spawn policy */ + if (imgp && imgp->ip_new_thread) { + task_clear_return_wait(get_threadtask(imgp->ip_new_thread), TCRW_CLEAR_INITIAL_WAIT); } /* @@ -2981,25 +3656,30 @@ bad: if (error == 0 && imgp->ip_px_sa != NULL) { struct _posix_spawnattr *psa = (struct _posix_spawnattr *) imgp->ip_px_sa; - exec_handle_spawnattr_policy(p, psa->psa_apptype, psa->psa_qos_clamp, psa->psa_darwin_role, - portwatch_ports, portwatch_count); + error = exec_handle_spawnattr_policy(p, imgp->ip_new_thread, psa->psa_apptype, psa->psa_qos_clamp, + psa->psa_darwin_role, &port_actions); + } + + /* Transfer the turnstile watchport boost to new task if in exec */ + if (error == 0 && !spawn_no_exec) { + task_transfer_turnstile_watchports(old_task, new_task, imgp->ip_new_thread); } /* - * Need to transfer pending watch port boosts to the new task while still making - * sure that the old task remains in the importance linkage. Create an importance - * linkage from old task to new task, then switch the task importance base - * of old task and new task. After the switch the port watch boost will be - * boosting the new task and new task will be donating importance to old task. + * Apply the requested maximum address. */ - if (error == 0 && task_did_exec(current_task())) { - inherit = ipc_importance_exec_switch_task(current_task(), get_threadtask(imgp->ip_new_thread)); + if (error == 0 && imgp->ip_px_sa != NULL) { + struct _posix_spawnattr *psa = (struct _posix_spawnattr *) imgp->ip_px_sa; + + if (psa->psa_max_addr) { + vm_map_set_max_addr(get_task_map(new_task), psa->psa_max_addr); + } } if (error == 0) { - /* Apply the main thread qos */ + /* Apply the main thread qos */ thread_t main_thread = imgp->ip_new_thread; - task_set_main_thread_qos(get_threadtask(imgp->ip_new_thread), main_thread); + task_set_main_thread_qos(new_task, main_thread); #if CONFIG_MACF /* @@ -3007,7 +3687,8 @@ bad: * a jumbo-size map. */ if (mac_proc_check_map_anon(p, 0, 0, 0, MAP_JIT, NULL) == 0) { - vm_map_set_jumbo(get_task_map(p->task)); + vm_map_set_jumbo(get_task_map(new_task)); + vm_map_set_jit_entitled(get_task_map(new_task)); } #endif /* CONFIG_MACF */ } @@ -3016,16 +3697,8 @@ bad: * Release any ports we kept around for binding to the new task * We need to release the rights even if the posix_spawn has failed. */ - if (portwatch_ports != NULL) { - for (int i = 0; i < portwatch_count; i++) { - ipc_port_t port = NULL; - if ((port = portwatch_ports[i]) != NULL) { - ipc_port_release_send(port); - } - } - FREE(portwatch_ports, M_TEMP); - portwatch_ports = NULL; - portwatch_count = 0; + if (imgp->ip_px_spa != NULL) { + exec_port_actions_destroy(&port_actions); } /* @@ -3045,39 +3718,56 @@ bad: psignal_vfork(p, p->task, imgp->ip_new_thread, SIGTRAP); } - if (error == 0 && !spawn_no_exec) - KDBG(BSDDBG_CODE(DBG_BSD_PROC,BSD_PROC_EXEC), - p->p_pid); + if (error == 0 && !spawn_no_exec) { + KDBG(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXEC), + p->p_pid); + } } if (imgp != NULL) { - if (imgp->ip_vp) + if (imgp->ip_vp) { vnode_put(imgp->ip_vp); - if (imgp->ip_scriptvp) + } + if (imgp->ip_scriptvp) { vnode_put(imgp->ip_scriptvp); - if (imgp->ip_strings) + } + if (imgp->ip_strings) { execargs_free(imgp); - if (imgp->ip_px_sfa != NULL) + } + if (imgp->ip_px_sfa != NULL) { FREE(imgp->ip_px_sfa, M_TEMP); - if (imgp->ip_px_spa != NULL) + } + if (imgp->ip_px_spa != NULL) { FREE(imgp->ip_px_spa, M_TEMP); + } #if CONFIG_PERSONAS - if (imgp->ip_px_persona != NULL) + if (imgp->ip_px_persona != NULL) { FREE(imgp->ip_px_persona, M_TEMP); + } #endif + if (imgp->ip_px_pcred_info != NULL) { + FREE(imgp->ip_px_pcred_info, M_TEMP); + } #if CONFIG_MACF - if (imgp->ip_px_smpx != NULL) + if (imgp->ip_px_smpx != NULL) { spawn_free_macpolicyinfo(imgp->ip_px_smpx); - if (imgp->ip_execlabelp) + } + if (imgp->ip_execlabelp) { mac_cred_label_free(imgp->ip_execlabelp); - if (imgp->ip_scriptlabelp) + } + if (imgp->ip_scriptlabelp) { mac_vnode_label_free(imgp->ip_scriptlabelp); + } if (imgp->ip_cs_error != OS_REASON_NULL) { os_reason_free(imgp->ip_cs_error); imgp->ip_cs_error = OS_REASON_NULL; } #endif + if (imgp->ip_sc_port != NULL) { + ipc_port_release_send(imgp->ip_sc_port); + imgp->ip_sc_port = NULL; + } } #if CONFIG_DTRACE @@ -3122,11 +3812,22 @@ bad: (*dtrace_proc_waitfor_hook)(p); } #endif + +#if CONFIG_AUDIT + if (!error && AUDIT_ENABLED() && p) { + /* Add the CDHash of the new process to the audit record */ + uint8_t *cdhash = cs_get_cdhash(p); + if (cdhash) { + AUDIT_ARG(data, cdhash, sizeof(uint8_t), CS_CDHASH_LEN); + } + } +#endif + /* * clear bsd_info from old task if it did exec. */ - if (task_did_exec(current_task())) { - set_bsdtask_info(current_task(), NULL); + if (task_did_exec(old_task)) { + set_bsdtask_info(old_task, NULL); } /* clear bsd_info from new task and terminate it if exec failed */ @@ -3140,8 +3841,15 @@ bad: /* * If the parent wants the pid, copy it out */ - if (pid != USER_ADDR_NULL) - (void)suword(pid, p->p_pid); + if (pid != USER_ADDR_NULL) { + _Static_assert(sizeof(p->p_pid) == 4, "posix_spawn() assumes a 32-bit pid_t"); + bool aligned = (pid & 3) == 0; + if (aligned) { + (void)copyout_atomic32(p->p_pid, pid); + } else { + (void)suword(pid, p->p_pid); + } + } retval[0] = error; /* @@ -3170,15 +3878,15 @@ bad: * switch the tasks, terminating the current task without the switch would * result in loosing the SIGKILL status. */ - if (task_did_exec(current_task())) { + if (task_did_exec(old_task)) { /* Terminate the current task, since exec will start in new task */ - task_terminate_internal(current_task()); + task_terminate_internal(old_task); } /* Release the thread ref returned by fork_create_child/fork1 */ if (imgp != NULL && imgp->ip_new_thread) { /* wake up the new thread */ - task_clear_return_wait(get_threadtask(imgp->ip_new_thread)); + task_clear_return_wait(get_threadtask(imgp->ip_new_thread), TCRW_CLEAR_FINAL_WAIT); thread_deallocate(imgp->ip_new_thread); imgp->ip_new_thread = NULL; } @@ -3200,8 +3908,8 @@ bad: if (inherit != NULL) { ipc_importance_release(inherit); } - - return(error); + + return error; } /* @@ -3277,6 +3985,7 @@ proc_exec_switch_task(proc_t p, task_t old_task, task_t new_task, thread_t new_t /* Clear dispatchqueue and workloop ast offset */ p->p_dispatchqueue_offset = 0; p->p_dispatchqueue_serialno_offset = 0; + p->p_dispatchqueue_label_offset = 0; p->p_return_to_kernel_offset = 0; /* Copy the signal state, dtrace state and set bsd ast on new thread */ @@ -3366,7 +4075,7 @@ execve(proc_t p, struct execve_args *uap, int32_t *retval) muap.mac_p = USER_ADDR_NULL; err = __mac_execve(p, &muap, retval); - return(err); + return err; } /* @@ -3398,14 +4107,15 @@ execve(proc_t p, struct execve_args *uap, int32_t *retval) int __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) { - char *bufp = NULL; + char *bufp = NULL; struct image_params *imgp; struct vnode_attr *vap; struct vnode_attr *origvap; int error; int is_64 = IS_64BIT_PROCESS(p); struct vfs_context context; - struct uthread *uthread; + struct uthread *uthread; + task_t old_task = current_task(); task_t new_task = NULL; boolean_t should_release_proc_ref = FALSE; boolean_t exec_done = FALSE; @@ -3413,9 +4123,9 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) void *inherit = NULL; context.vc_thread = current_thread(); - context.vc_ucred = kauth_cred_proc_ref(p); /* XXX must NOT be kauth_cred_get() */ + context.vc_ucred = kauth_cred_proc_ref(p); /* XXX must NOT be kauth_cred_get() */ - /* Allocate a big chunk for locals instead of using stack since these + /* Allocate a big chunk for locals instead of using stack since these * structures a pretty big. */ MALLOC(bufp, char *, (sizeof(*imgp) + sizeof(*vap) + sizeof(*origvap)), M_TEMP, M_WAITOK | M_ZERO); @@ -3426,7 +4136,7 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) } vap = (struct vnode_attr *) (bufp + sizeof(*imgp)); origvap = (struct vnode_attr *) (bufp + sizeof(*imgp) + sizeof(*vap)); - + /* Initialize the common data in the image_params structure */ imgp->ip_user_fname = uap->fname; imgp->ip_user_argv = uap->argp; @@ -3434,10 +4144,11 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) imgp->ip_vattr = vap; imgp->ip_origvattr = origvap; imgp->ip_vfs_context = &context; - imgp->ip_flags = (is_64 ? IMGPF_WAS_64BIT : IMGPF_NONE) | ((p->p_flag & P_DISABLE_ASLR) ? IMGPF_DISABLE_ASLR : IMGPF_NONE); + imgp->ip_flags = (is_64 ? IMGPF_WAS_64BIT_ADDR : IMGPF_NONE) | ((p->p_flag & P_DISABLE_ASLR) ? IMGPF_DISABLE_ASLR : IMGPF_NONE); imgp->ip_seg = (is_64 ? UIO_USERSPACE64 : UIO_USERSPACE32); imgp->ip_mac_return = 0; imgp->ip_cs_error = OS_REASON_NULL; + imgp->ip_simulator_binary = IMGPF_SB_DEFAULT; #if CONFIG_MACF if (uap->mac_p != USER_ADDR_NULL) { @@ -3480,8 +4191,13 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) * During exec any transition from new_task -> proc is fine, but don't allow * transition from proc->task, since it will modify old_task. */ - imgp->ip_new_thread = fork_create_child(current_task(), - NULL, p, FALSE, p->p_flag & P_LP64, TRUE); + imgp->ip_new_thread = fork_create_child(old_task, + NULL, + p, + FALSE, + p->p_flag & P_LP64, + task_get_64bit_data(old_task), + TRUE); /* task and thread ref returned by fork_create_child */ if (imgp->ip_new_thread == NULL) { error = ENOMEM; @@ -3501,19 +4217,33 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) * for vfexec. */ new_task = get_threadtask(imgp->ip_new_thread); +#if defined(HAS_APPLE_PAC) + ml_task_set_disable_user_jop(new_task, imgp->ip_flags & IMGPF_NOJOP ? TRUE : FALSE); + ml_thread_set_disable_user_jop(imgp->ip_new_thread, imgp->ip_flags & IMGPF_NOJOP ? TRUE : FALSE); +#endif } if (!error && !in_vfexec) { - p = proc_exec_switch_task(p, current_task(), new_task, imgp->ip_new_thread); + p = proc_exec_switch_task(p, old_task, new_task, imgp->ip_new_thread); /* proc ref returned */ should_release_proc_ref = TRUE; + + /* + * Need to transfer pending watch port boosts to the new task while still making + * sure that the old task remains in the importance linkage. Create an importance + * linkage from old task to new task, then switch the task importance base + * of old task and new task. After the switch the port watch boost will be + * boosting the new task and new task will be donating importance to old task. + */ + inherit = ipc_importance_exec_switch_task(old_task, new_task); } kauth_cred_unref(&context.vc_ucred); - + /* Image not claimed by any activator? */ - if (error == -1) + if (error == -1) { error = ENOEXEC; + } if (!error) { exec_done = TRUE; @@ -3524,20 +4254,26 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) } /* flag exec has occurred, notify only if it has not failed due to FP Key error */ - if (exec_done && ((p->p_lflag & P_LTERM_DECRYPTFAIL) == 0)) + if (exec_done && ((p->p_lflag & P_LTERM_DECRYPTFAIL) == 0)) { proc_knote(p, NOTE_EXEC); + } - if (imgp->ip_vp != NULLVP) + if (imgp->ip_vp != NULLVP) { vnode_put(imgp->ip_vp); - if (imgp->ip_scriptvp != NULLVP) + } + if (imgp->ip_scriptvp != NULLVP) { vnode_put(imgp->ip_scriptvp); - if (imgp->ip_strings) + } + if (imgp->ip_strings) { execargs_free(imgp); + } #if CONFIG_MACF - if (imgp->ip_execlabelp) + if (imgp->ip_execlabelp) { mac_cred_label_free(imgp->ip_execlabelp); - if (imgp->ip_scriptlabelp) + } + if (imgp->ip_scriptlabelp) { mac_vnode_label_free(imgp->ip_scriptlabelp); + } #endif if (imgp->ip_cs_error != OS_REASON_NULL) { os_reason_free(imgp->ip_cs_error); @@ -3555,21 +4291,34 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) } if (!error) { - task_bank_init(get_threadtask(imgp->ip_new_thread)); + task_bank_init(new_task); proc_transend(p, 0); +#if __arm64__ + proc_legacy_footprint_entitled(p, new_task, __FUNCTION__); + proc_ios13extended_footprint_entitled(p, new_task, __FUNCTION__); +#endif /* __arm64__ */ + /* Sever any extant thread affinity */ thread_affinity_exec(current_thread()); /* Inherit task role from old task to new task for exec */ if (!in_vfexec) { - proc_inherit_task_role(get_threadtask(imgp->ip_new_thread), current_task()); + proc_inherit_task_role(new_task, old_task); } thread_t main_thread = imgp->ip_new_thread; task_set_main_thread_qos(new_task, main_thread); +#if CONFIG_ARCADE + /* + * Check to see if we need to trigger an arcade upcall AST now + * that the vnode has been reset on the task. + */ + arcade_prepare(new_task, imgp->ip_new_thread); +#endif /* CONFIG_ARCADE */ + #if CONFIG_MACF /* * Processes with the MAP_JIT entitlement are permitted to have @@ -3577,15 +4326,35 @@ __mac_execve(proc_t p, struct __mac_execve_args *uap, int32_t *retval) */ if (mac_proc_check_map_anon(p, 0, 0, 0, MAP_JIT, NULL) == 0) { vm_map_set_jumbo(get_task_map(new_task)); + vm_map_set_jit_entitled(get_task_map(new_task)); } #endif /* CONFIG_MACF */ + if (vm_darkwake_mode == TRUE) { + /* + * This process is being launched when the system + * is in darkwake. So mark it specially. This will + * cause all its pages to be entered in the background Q. + */ + task_set_darkwake_mode(new_task, vm_darkwake_mode); + } #if CONFIG_DTRACE dtrace_thread_didexec(imgp->ip_new_thread); - if ((dtrace_proc_waitfor_hook = dtrace_proc_waitfor_exec_ptr) != NULL) + if ((dtrace_proc_waitfor_hook = dtrace_proc_waitfor_exec_ptr) != NULL) { (*dtrace_proc_waitfor_hook)(p); + } +#endif + +#if CONFIG_AUDIT + if (!error && AUDIT_ENABLED() && p) { + /* Add the CDHash of the new process to the audit record */ + uint8_t *cdhash = cs_get_cdhash(p); + if (cdhash) { + AUDIT_ARG(data, cdhash, sizeof(uint8_t), CS_CDHASH_LEN); + } + } #endif if (in_vfexec) { @@ -3600,8 +4369,8 @@ exit_with_error: /* * clear bsd_info from old task if it did exec. */ - if (task_did_exec(current_task())) { - set_bsdtask_info(current_task(), NULL); + if (task_did_exec(old_task)) { + set_bsdtask_info(old_task, NULL); } /* clear bsd_info from new task and terminate it if exec failed */ @@ -3610,32 +4379,31 @@ exit_with_error: task_terminate_internal(new_task); } - /* - * Need to transfer pending watch port boosts to the new task while still making - * sure that the old task remains in the importance linkage. Create an importance - * linkage from old task to new task, then switch the task importance base - * of old task and new task. After the switch the port watch boost will be - * boosting the new task and new task will be donating importance to old task. - */ - if (error == 0 && task_did_exec(current_task())) { - inherit = ipc_importance_exec_switch_task(current_task(), get_threadtask(imgp->ip_new_thread)); - } - if (imgp != NULL) { + /* Clear the initial wait on the thread transferring watchports */ + if (imgp->ip_new_thread) { + task_clear_return_wait(get_threadtask(imgp->ip_new_thread), TCRW_CLEAR_INITIAL_WAIT); + } + + /* Transfer the watchport boost to new task */ + if (!error && !in_vfexec) { + task_transfer_turnstile_watchports(old_task, + new_task, imgp->ip_new_thread); + } /* * Do not terminate the current task, if proc_exec_switch_task did not * switch the tasks, terminating the current task without the switch would * result in loosing the SIGKILL status. */ - if (task_did_exec(current_task())) { + if (task_did_exec(old_task)) { /* Terminate the current task, since exec will start in new task */ - task_terminate_internal(current_task()); + task_terminate_internal(old_task); } /* Release the thread ref returned by fork_create_child */ if (imgp->ip_new_thread) { /* wake up the new exec thread */ - task_clear_return_wait(get_threadtask(imgp->ip_new_thread)); + task_clear_return_wait(get_threadtask(imgp->ip_new_thread), TCRW_CLEAR_FINAL_WAIT); thread_deallocate(imgp->ip_new_thread); imgp->ip_new_thread = NULL; } @@ -3658,8 +4426,8 @@ exit_with_error: if (inherit != NULL) { ipc_importance_release(inherit); } - - return(error); + + return error; } @@ -3686,14 +4454,14 @@ copyinptr(user_addr_t froma, user_addr_t *toptr, int ptr_size) if (ptr_size == 4) { /* 64 bit value containing 32 bit address */ - unsigned int i; + unsigned int i = 0; error = copyin(froma, &i, 4); - *toptr = CAST_USER_ADDR_T(i); /* SAFE */ + *toptr = CAST_USER_ADDR_T(i); /* SAFE */ } else { error = copyin(froma, toptr, 8); } - return (error); + return error; } @@ -3718,13 +4486,13 @@ copyoutptr(user_addr_t ua, user_addr_t ptr, int ptr_size) if (ptr_size == 4) { /* 64 bit value containing 32 bit address */ - unsigned int i = CAST_DOWN_EXPLICIT(unsigned int,ua); /* SAFE */ + unsigned int i = CAST_DOWN_EXPLICIT(unsigned int, ua); /* SAFE */ error = copyout(&i, ptr, 4); } else { error = copyout(&ua, ptr, 8); } - return (error); + return error; } @@ -3806,25 +4574,25 @@ static int exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) { proc_t p = vfs_context_proc(imgp->ip_vfs_context); - int ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT) ? 8 : 4; - int ptr_area_size; + int ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) ? 8 : 4; + int ptr_area_size; void *ptr_buffer_start, *ptr_buffer; int string_size; - user_addr_t string_area; /* *argv[], *env[] */ - user_addr_t ptr_area; /* argv[], env[], applev[] */ - user_addr_t argc_area; /* argc */ - user_addr_t stack; + user_addr_t string_area; /* *argv[], *env[] */ + user_addr_t ptr_area; /* argv[], env[], applev[] */ + user_addr_t argc_area; /* argc */ + user_addr_t stack; int error; unsigned i; struct copyout_desc { - char *start_string; - int count; + char *start_string; + int count; #if CONFIG_DTRACE - user_addr_t *dtrace_cookie; + user_addr_t *dtrace_cookie; #endif - boolean_t null_term; + boolean_t null_term; } descriptors[] = { { .start_string = imgp->ip_startargv, @@ -3880,8 +4648,7 @@ exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) * Need room for one pointer for each string, plus * one for the NULLs terminating the argv, envv, and apple areas. */ - ptr_area_size = (imgp->ip_argc + imgp->ip_envc + imgp->ip_applec + 3) * - ptr_size; + ptr_area_size = (imgp->ip_argc + imgp->ip_envc + imgp->ip_applec + 3) * ptr_size; stack -= ptr_area_size; ptr_area = stack; @@ -3913,11 +4680,12 @@ exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) * Copy out the entire strings area. */ error = copyout(imgp->ip_strings, string_area, - string_size); - if (error) + string_size); + if (error) { goto bad; + } - for (i = 0; i < sizeof(descriptors)/sizeof(descriptors[0]); i++) { + for (i = 0; i < sizeof(descriptors) / sizeof(descriptors[0]); i++) { char *cur_string = descriptors[i].start_string; int j; @@ -3935,14 +4703,14 @@ exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) */ for (j = 0; j < descriptors[i].count; j++) { user_addr_t cur_address = string_area + (cur_string - imgp->ip_strings); - + /* Copy out the pointer to the current string. Alignment has been verified */ if (ptr_size == 8) { *(uint64_t *)ptr_buffer = (uint64_t)cur_address; } else { *(uint32_t *)ptr_buffer = (uint32_t)cur_address; } - + ptr_buffer = (void *)((uintptr_t)ptr_buffer + ptr_size); cur_string += strlen(cur_string) + 1; /* Only a NUL between strings in the same area */ } @@ -3953,7 +4721,7 @@ exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) } else { *(uint32_t *)ptr_buffer = 0; } - + ptr_buffer = (void *)((uintptr_t)ptr_buffer + ptr_size); } } @@ -3962,17 +4730,19 @@ exec_copyout_strings(struct image_params *imgp, user_addr_t *stackp) * Copy out all our pointer arrays in bulk. */ error = copyout(ptr_buffer_start, ptr_area, - ptr_area_size); - if (error) + ptr_area_size); + if (error) { goto bad; + } /* argc (int32, stored in a ptr_size area) */ error = copyoutptr((user_addr_t)imgp->ip_argc, argc_area, ptr_size); - if (error) + if (error) { goto bad; + } bad: - return(error); + return error; } @@ -4007,14 +4777,14 @@ static int exec_extract_strings(struct image_params *imgp) { int error = 0; - int ptr_size = (imgp->ip_flags & IMGPF_WAS_64BIT) ? 8 : 4; - int new_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT) ? 8 : 4; - user_addr_t argv = imgp->ip_user_argv; - user_addr_t envv = imgp->ip_user_envv; + int ptr_size = (imgp->ip_flags & IMGPF_WAS_64BIT_ADDR) ? 8 : 4; + int new_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) ? 8 : 4; + user_addr_t argv = imgp->ip_user_argv; + user_addr_t envv = imgp->ip_user_envv; /* * Adjust space reserved for the path name by however much padding it - * needs. Doing this here since we didn't know if this would be a 32- + * needs. Doing this here since we didn't know if this would be a 32- * or 64-bit process back in exec_save_path. */ while (imgp->ip_strspace % new_ptr_size != 0) { @@ -4029,8 +4799,8 @@ exec_extract_strings(struct image_params *imgp) imgp->ip_startargv = imgp->ip_strendp; imgp->ip_argc = 0; - if((imgp->ip_flags & IMGPF_INTERPRET) != 0) { - user_addr_t arg; + if ((imgp->ip_flags & IMGPF_INTERPRET) != 0) { + user_addr_t arg; char *argstart, *ch; /* First, the arguments in the "#!" string are tokenized and extracted. */ @@ -4061,8 +4831,9 @@ exec_extract_strings(struct image_params *imgp) } /* Error-check, regardless of whether this is the last interpreter arg or not */ - if (error) + if (error) { goto bad; + } if (imgp->ip_argspace < new_ptr_size) { error = E2BIG; goto bad; @@ -4079,8 +4850,9 @@ exec_extract_strings(struct image_params *imgp) * to locate their script arguments. */ error = copyinptr(argv, &arg, ptr_size); - if (error) + if (error) { goto bad; + } if (arg != 0LL) { argv += ptr_size; /* consume without using */ } @@ -4093,9 +4865,10 @@ exec_extract_strings(struct image_params *imgp) } else { error = exec_add_user_string(imgp, imgp->ip_user_fname, imgp->ip_seg, TRUE); } - - if (error) + + if (error) { goto bad; + } if (imgp->ip_argspace < new_ptr_size) { error = E2BIG; goto bad; @@ -4105,11 +4878,12 @@ exec_extract_strings(struct image_params *imgp) } while (argv != 0LL) { - user_addr_t arg; + user_addr_t arg; error = copyinptr(argv, &arg, ptr_size); - if (error) + if (error) { goto bad; + } if (arg == 0LL) { break; @@ -4118,18 +4892,19 @@ exec_extract_strings(struct image_params *imgp) argv += ptr_size; /* - * av[n...] = arg[n] - */ + * av[n...] = arg[n] + */ error = exec_add_user_string(imgp, arg, imgp->ip_seg, TRUE); - if (error) + if (error) { goto bad; + } if (imgp->ip_argspace < new_ptr_size) { error = E2BIG; goto bad; } imgp->ip_argspace -= new_ptr_size; /* to hold argv[] entry */ imgp->ip_argc++; - } + } /* Save space for argv[] NULL terminator */ if (imgp->ip_argspace < new_ptr_size) { @@ -4137,29 +4912,31 @@ exec_extract_strings(struct image_params *imgp) goto bad; } imgp->ip_argspace -= new_ptr_size; - + /* Note where the args ends and env begins. */ imgp->ip_endargv = imgp->ip_strendp; imgp->ip_envc = 0; /* Now, get the environment */ while (envv != 0LL) { - user_addr_t env; + user_addr_t env; error = copyinptr(envv, &env, ptr_size); - if (error) + if (error) { goto bad; + } envv += ptr_size; if (env == 0LL) { break; } /* - * av[n...] = env[n] - */ + * av[n...] = env[n] + */ error = exec_add_user_string(imgp, env, imgp->ip_seg, TRUE); - if (error) + if (error) { goto bad; + } if (imgp->ip_argspace < new_ptr_size) { error = E2BIG; goto bad; @@ -4185,7 +4962,7 @@ exec_extract_strings(struct image_params *imgp) imgp->ip_strspace--; imgp->ip_argspace--; } - + /* Note where the envv ends and applev begins. */ imgp->ip_endenvv = imgp->ip_strendp; @@ -4204,15 +4981,21 @@ bad: * 8-byte guard. Until somebody needs more than an 8-byte guard value, don't * do the work to construct them. */ -#define GUARD_VALUES 1 -#define GUARD_KEY "stack_guard=" +#define GUARD_VALUES 1 +#define GUARD_KEY "stack_guard=" /* * System malloc needs some entropy when it is initialized. */ -#define ENTROPY_VALUES 2 +#define ENTROPY_VALUES 2 #define ENTROPY_KEY "malloc_entropy=" +/* + * libplatform needs a random pointer-obfuscation value when it is initialized. + */ +#define PTR_MUNGE_VALUES 1 +#define PTR_MUNGE_KEY "ptr_munge=" + /* * System malloc engages nanozone for UIAPP. */ @@ -4228,6 +5011,7 @@ extern user64_addr_t commpage_text64_location; #define FSID_KEY "executable_file=" #define DYLD_FSID_KEY "dyld_file=" #define CDHASH_KEY "executable_cdhash=" +#define DYLD_FLAGS_KEY "dyld_flags=" #define FSID_MAX_STRING "0x1234567890abcdef,0x1234567890abcdef" @@ -4235,9 +5019,9 @@ extern user64_addr_t commpage_text64_location; static int exec_add_entropy_key(struct image_params *imgp, - const char *key, - int values, - boolean_t embedNUL) + const char *key, + int values, + boolean_t embedNUL) { const int limit = 8; uint64_t entropy[limit]; @@ -4246,17 +5030,17 @@ exec_add_entropy_key(struct image_params *imgp, values = limit; } - read_random(entropy, sizeof(entropy[0]) * values); + read_random(entropy, sizeof(entropy[0]) * values); if (embedNUL) { entropy[0] &= ~(0xffull << 8); } - int len = snprintf(str, sizeof(str), "%s0x%llx", key, entropy[0]); + int len = scnprintf(str, sizeof(str), "%s0x%llx", key, entropy[0]); int remaining = sizeof(str) - len; for (int i = 1; i < values && remaining > 0; ++i) { int start = sizeof(str) - remaining; - len = snprintf(&str[start], remaining, ",0x%llx", entropy[i]); + len = scnprintf(&str[start], remaining, ",0x%llx", entropy[i]); remaining -= len; } @@ -4266,12 +5050,16 @@ exec_add_entropy_key(struct image_params *imgp, /* * Build up the contents of the apple[] string vector */ +#if (DEVELOPMENT || DEBUG) +uint64_t dyld_flags = 0; +#endif + static int exec_add_apple_strings(struct image_params *imgp, - const load_result_t *load_result) + const load_result_t *load_result) { int error; - int img_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT) ? 8 : 4; + int img_ptr_size = (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) ? 8 : 4; /* exec_save_path stored the first string */ imgp->ip_applec = 1; @@ -4299,7 +5087,7 @@ exec_add_apple_strings(struct image_params *imgp, if ((proc_flags & _POSIX_SPAWN_NANO_ALLOCATOR) == _POSIX_SPAWN_NANO_ALLOCATOR) { const char *nano_string = NANO_ENGAGE_KEY; error = exec_add_user_string(imgp, CAST_USER_ADDR_T(nano_string), UIO_SYSSPACE, FALSE); - if (error){ + if (error) { goto bad; } imgp->ip_applec++; @@ -4328,7 +5116,17 @@ exec_add_apple_strings(struct image_params *imgp, } imgp->ip_applec++; - /* + /* + * Supply libpthread & libplatform with a random value to use for pointer + * obfuscation. + */ + error = exec_add_entropy_key(imgp, PTR_MUNGE_KEY, PTR_MUNGE_VALUES, FALSE); + if (error) { + goto bad; + } + imgp->ip_applec++; + + /* * Add MAIN_STACK_KEY: Supplies the address and size of the main thread's * stack if it was allocated by the kernel. * @@ -4338,11 +5136,11 @@ exec_add_apple_strings(struct image_params *imgp, if (load_result->unixproc) { char stack_string[strlen(MAIN_STACK_KEY) + (HEX_STR_LEN + 1) * MAIN_STACK_VALUES + 1]; snprintf(stack_string, sizeof(stack_string), - MAIN_STACK_KEY "0x%llx,0x%llx,0x%llx,0x%llx", - (uint64_t)load_result->user_stack, - (uint64_t)load_result->user_stack_size, - (uint64_t)load_result->user_stack_alloc, - (uint64_t)load_result->user_stack_alloc_size); + MAIN_STACK_KEY "0x%llx,0x%llx,0x%llx,0x%llx", + (uint64_t)load_result->user_stack, + (uint64_t)load_result->user_stack_size, + (uint64_t)load_result->user_stack_alloc, + (uint64_t)load_result->user_stack_alloc_size); error = exec_add_user_string(imgp, CAST_USER_ADDR_T(stack_string), UIO_SYSSPACE, FALSE); if (error) { goto bad; @@ -4351,12 +5149,12 @@ exec_add_apple_strings(struct image_params *imgp, } if (imgp->ip_vattr) { - uint64_t fsid = get_va_fsid(imgp->ip_vattr); + uint64_t fsid = vnode_get_va_fsid(imgp->ip_vattr); uint64_t fsobjid = imgp->ip_vattr->va_fileid; char fsid_string[strlen(FSID_KEY) + strlen(FSID_MAX_STRING) + 1]; snprintf(fsid_string, sizeof(fsid_string), - FSID_KEY "0x%llx,0x%llx", fsid, fsobjid); + FSID_KEY "0x%llx,0x%llx", fsid, fsobjid); error = exec_add_user_string(imgp, CAST_USER_ADDR_T(fsid_string), UIO_SYSSPACE, FALSE); if (error) { goto bad; @@ -4364,26 +5162,26 @@ exec_add_apple_strings(struct image_params *imgp, imgp->ip_applec++; } - if (imgp->ip_dyld_fsid || imgp->ip_dyld_fsobjid ) { + if (imgp->ip_dyld_fsid || imgp->ip_dyld_fsobjid) { char fsid_string[strlen(DYLD_FSID_KEY) + strlen(FSID_MAX_STRING) + 1]; snprintf(fsid_string, sizeof(fsid_string), - DYLD_FSID_KEY "0x%llx,0x%llx", imgp->ip_dyld_fsid, imgp->ip_dyld_fsobjid); + DYLD_FSID_KEY "0x%llx,0x%llx", imgp->ip_dyld_fsid, imgp->ip_dyld_fsobjid); error = exec_add_user_string(imgp, CAST_USER_ADDR_T(fsid_string), UIO_SYSSPACE, FALSE); if (error) { goto bad; } imgp->ip_applec++; } - - uint8_t cdhash[SHA1_RESULTLEN]; + + uint8_t cdhash[SHA1_RESULTLEN]; int cdhash_errror = ubc_cs_getcdhash(imgp->ip_vp, imgp->ip_arch_offset, cdhash); if (cdhash_errror == 0) { - char hash_string[strlen(CDHASH_KEY) + 2*SHA1_RESULTLEN + 1]; + char hash_string[strlen(CDHASH_KEY) + 2 * SHA1_RESULTLEN + 1]; strncpy(hash_string, CDHASH_KEY, sizeof(hash_string)); - char *p = hash_string + sizeof(CDHASH_KEY) - 1; - for (int i = 0; i < SHA1_RESULTLEN; i++) { + char *p = hash_string + sizeof(CDHASH_KEY) - 1; + for (int i = 0; i < SHA1_RESULTLEN; i++) { snprintf(p, 3, "%02x", (int) cdhash[i]); - p += 2; + p += 2; } error = exec_add_user_string(imgp, CAST_USER_ADDR_T(hash_string), UIO_SYSSPACE, FALSE); if (error) { @@ -4391,6 +5189,17 @@ exec_add_apple_strings(struct image_params *imgp, } imgp->ip_applec++; } +#if (DEVELOPMENT || DEBUG) + if (dyld_flags) { + char dyld_flags_string[strlen(DYLD_FLAGS_KEY) + HEX_STR_LEN + 1]; + snprintf(dyld_flags_string, sizeof(dyld_flags_string), DYLD_FLAGS_KEY "0x%llx", dyld_flags); + error = exec_add_user_string(imgp, CAST_USER_ADDR_T(dyld_flags_string), UIO_SYSSPACE, FALSE); + if (error) { + goto bad; + } + imgp->ip_applec++; + } +#endif /* Align the tail of the combined applev area */ while (imgp->ip_strspace % img_ptr_size != 0) { @@ -4402,7 +5211,7 @@ bad: return error; } -#define unix_stack_size(p) (p->p_rlimit[RLIMIT_STACK].rlim_cur) +#define unix_stack_size(p) (p->p_rlimit[RLIMIT_STACK].rlim_cur) /* * exec_check_permissions @@ -4430,9 +5239,10 @@ exec_check_permissions(struct image_params *imgp) kauth_action_t action; /* Only allow execution of regular files */ - if (!vnode_isreg(vp)) - return (EACCES); - + if (!vnode_isreg(vp)) { + return EACCES; + } + /* Get the file attributes that we will be using here and elsewhere */ VATTR_INIT(vap); VATTR_WANTED(vap, va_uid); @@ -4442,48 +5252,56 @@ exec_check_permissions(struct image_params *imgp) VATTR_WANTED(vap, va_fsid64); VATTR_WANTED(vap, va_fileid); VATTR_WANTED(vap, va_data_size); - if ((error = vnode_getattr(vp, vap, imgp->ip_vfs_context)) != 0) - return (error); + if ((error = vnode_getattr(vp, vap, imgp->ip_vfs_context)) != 0) { + return error; + } /* * Ensure that at least one execute bit is on - otherwise root * will always succeed, and we don't want to happen unless the * file really is executable. */ - if (!vfs_authopaque(vnode_mount(vp)) && ((vap->va_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) - return (EACCES); + if (!vfs_authopaque(vnode_mount(vp)) && ((vap->va_mode & (S_IXUSR | S_IXGRP | S_IXOTH)) == 0)) { + return EACCES; + } /* Disallow zero length files */ - if (vap->va_data_size == 0) - return (ENOEXEC); + if (vap->va_data_size == 0) { + return ENOEXEC; + } imgp->ip_arch_offset = (user_size_t)0; imgp->ip_arch_size = vap->va_data_size; /* Disable setuid-ness for traced programs or if MNT_NOSUID */ - if ((vp->v_mount->mnt_flag & MNT_NOSUID) || (p->p_lflag & P_LTRACED)) + if ((vp->v_mount->mnt_flag & MNT_NOSUID) || (p->p_lflag & P_LTRACED)) { vap->va_mode &= ~(VSUID | VSGID); + } /* * Disable _POSIX_SPAWN_ALLOW_DATA_EXEC and _POSIX_SPAWN_DISABLE_ASLR * flags for setuid/setgid binaries. */ - if (vap->va_mode & (VSUID | VSGID)) + if (vap->va_mode & (VSUID | VSGID)) { imgp->ip_flags &= ~(IMGPF_ALLOW_DATA_EXEC | IMGPF_DISABLE_ASLR); + } #if CONFIG_MACF error = mac_vnode_check_exec(imgp->ip_vfs_context, vp, imgp); - if (error) - return (error); + if (error) { + return error; + } #endif - /* Check for execute permission */ - action = KAUTH_VNODE_EXECUTE; - /* Traced images must also be readable */ - if (p->p_lflag & P_LTRACED) - action |= KAUTH_VNODE_READ_DATA; - if ((error = vnode_authorize(vp, NULL, action, imgp->ip_vfs_context)) != 0) - return (error); + /* Check for execute permission */ + action = KAUTH_VNODE_EXECUTE; + /* Traced images must also be readable */ + if (p->p_lflag & P_LTRACED) { + action |= KAUTH_VNODE_READ_DATA; + } + if ((error = vnode_authorize(vp, NULL, action, imgp->ip_vfs_context)) != 0) { + return error; + } #if 0 /* Don't let it run if anyone had it open for writing */ @@ -4491,7 +5309,7 @@ exec_check_permissions(struct image_params *imgp) if (vp->v_writecount) { panic("going to return ETXTBSY %x", vp); vnode_unlock(vp); - return (ETXTBSY); + return ETXTBSY; } vnode_unlock(vp); #endif @@ -4499,7 +5317,7 @@ exec_check_permissions(struct image_params *imgp) /* XXX May want to indicate to underlying FS that vnode is open */ - return (error); + return error; } @@ -4532,17 +5350,16 @@ exec_check_permissions(struct image_params *imgp) static int exec_handle_sugid(struct image_params *imgp) { - proc_t p = vfs_context_proc(imgp->ip_vfs_context); - kauth_cred_t cred = vfs_context_ucred(imgp->ip_vfs_context); - kauth_cred_t my_cred, my_new_cred; - int i; - int leave_sugid_clear = 0; - int mac_reset_ipc = 0; - int error = 0; - task_t task = NULL; + proc_t p = vfs_context_proc(imgp->ip_vfs_context); + kauth_cred_t cred = vfs_context_ucred(imgp->ip_vfs_context); + int i; + int leave_sugid_clear = 0; + int mac_reset_ipc = 0; + int error = 0; + task_t task = NULL; #if CONFIG_MACF - int mac_transition, disjoint_cred = 0; - int label_update_return = 0; + int mac_transition, disjoint_cred = 0; + int label_update_return = 0; /* * Determine whether a call to update the MAC label will result in the @@ -4554,14 +5371,14 @@ exec_handle_sugid(struct image_params *imgp) * slow down the exec fast path for normal binaries. */ mac_transition = mac_cred_check_label_update_execve( - imgp->ip_vfs_context, - imgp->ip_vp, - imgp->ip_arch_offset, - imgp->ip_scriptvp, - imgp->ip_scriptlabelp, - imgp->ip_execlabelp, - p, - imgp->ip_px_smpx); + imgp->ip_vfs_context, + imgp->ip_vp, + imgp->ip_arch_offset, + imgp->ip_scriptvp, + imgp->ip_scriptlabelp, + imgp->ip_execlabelp, + p, + imgp->ip_px_smpx); #endif OSBitAndAtomic(~((uint32_t)P_SUGID), &p->p_flag); @@ -4581,11 +5398,11 @@ exec_handle_sugid(struct image_params *imgp) * such a call. */ if (((imgp->ip_origvattr->va_mode & VSUID) != 0 && - kauth_cred_getuid(cred) != imgp->ip_origvattr->va_uid) || + kauth_cred_getuid(cred) != imgp->ip_origvattr->va_uid) || ((imgp->ip_origvattr->va_mode & VSGID) != 0 && - ((kauth_cred_ismember_gid(cred, imgp->ip_origvattr->va_gid, &leave_sugid_clear) || !leave_sugid_clear) || - (kauth_cred_getgid(cred) != imgp->ip_origvattr->va_gid)))) { - + ((kauth_cred_ismember_gid(cred, imgp->ip_origvattr->va_gid, &leave_sugid_clear) || !leave_sugid_clear) || + (kauth_cred_getgid(cred) != imgp->ip_origvattr->va_gid))) || + (imgp->ip_sc_port != NULL)) { #if CONFIG_MACF /* label for MAC transition and neither VSUID nor VSGID */ handle_mac_transition: @@ -4612,73 +5429,61 @@ handle_mac_transition: * proc's ucred lock. This prevents others from accessing * a garbage credential. */ - while (imgp->ip_origvattr->va_mode & VSUID) { - my_cred = kauth_cred_proc_ref(p); - my_new_cred = kauth_cred_setresuid(my_cred, KAUTH_UID_NONE, imgp->ip_origvattr->va_uid, imgp->ip_origvattr->va_uid, KAUTH_UID_NONE); - - if (my_new_cred == my_cred) { - kauth_cred_unref(&my_cred); - break; - } - /* update cred on proc */ - proc_ucred_lock(p); + if (imgp->ip_sc_port != NULL) { + extern int suid_cred_verify(ipc_port_t, vnode_t, uint32_t *); + int ret = -1; + uid_t uid = UINT32_MAX; - if (p->p_ucred != my_cred) { - proc_ucred_unlock(p); - kauth_cred_unref(&my_new_cred); - continue; + /* + * Check that the vnodes match. If a script is being + * executed check the script's vnode rather than the + * interpreter's. + */ + struct vnode *vp = imgp->ip_scriptvp != NULL ? imgp->ip_scriptvp : imgp->ip_vp; + + ret = suid_cred_verify(imgp->ip_sc_port, vp, &uid); + if (ret == 0) { + apply_kauth_cred_update(p, ^kauth_cred_t (kauth_cred_t my_cred) { + return kauth_cred_setresuid(my_cred, + KAUTH_UID_NONE, + uid, + uid, + KAUTH_UID_NONE); + }); + } else { + error = EPERM; } - - /* donate cred reference on my_new_cred to p->p_ucred */ - p->p_ucred = my_new_cred; - PROC_UPDATE_CREDS_ONPROC(p); - proc_ucred_unlock(p); - - /* drop additional reference that was taken on the previous cred */ - kauth_cred_unref(&my_cred); - - break; } - while (imgp->ip_origvattr->va_mode & VSGID) { - my_cred = kauth_cred_proc_ref(p); - my_new_cred = kauth_cred_setresgid(my_cred, KAUTH_GID_NONE, imgp->ip_origvattr->va_gid, imgp->ip_origvattr->va_gid); - - if (my_new_cred == my_cred) { - kauth_cred_unref(&my_cred); - break; - } - - /* update cred on proc */ - proc_ucred_lock(p); - - if (p->p_ucred != my_cred) { - proc_ucred_unlock(p); - kauth_cred_unref(&my_new_cred); - continue; - } - - /* donate cred reference on my_new_cred to p->p_ucred */ - p->p_ucred = my_new_cred; - PROC_UPDATE_CREDS_ONPROC(p); - proc_ucred_unlock(p); - - /* drop additional reference that was taken on the previous cred */ - kauth_cred_unref(&my_cred); + if (imgp->ip_origvattr->va_mode & VSUID) { + apply_kauth_cred_update(p, ^kauth_cred_t (kauth_cred_t my_cred) { + return kauth_cred_setresuid(my_cred, + KAUTH_UID_NONE, + imgp->ip_origvattr->va_uid, + imgp->ip_origvattr->va_uid, + KAUTH_UID_NONE); + }); + } - break; + if (imgp->ip_origvattr->va_mode & VSGID) { + apply_kauth_cred_update(p, ^kauth_cred_t (kauth_cred_t my_cred) { + return kauth_cred_setresgid(my_cred, + KAUTH_GID_NONE, + imgp->ip_origvattr->va_gid, + imgp->ip_origvattr->va_gid); + }); } #endif /* !SECURE_KERNEL */ #if CONFIG_MACF - /* + /* * If a policy has indicated that it will transition the label, * before making the call into the MAC policies, get a new * duplicate credential, so they can modify it without * modifying any others sharing it. */ - if (mac_transition) { + if (mac_transition) { /* * This hook may generate upcalls that require * importance donation from the kernel. @@ -4687,16 +5492,16 @@ handle_mac_transition: thread_t thread = current_thread(); thread_enable_send_importance(thread, TRUE); kauth_proc_label_update_execve(p, - imgp->ip_vfs_context, - imgp->ip_vp, - imgp->ip_arch_offset, - imgp->ip_scriptvp, - imgp->ip_scriptlabelp, - imgp->ip_execlabelp, - &imgp->ip_csflags, - imgp->ip_px_smpx, - &disjoint_cred, /* will be non zero if disjoint */ - &label_update_return); + imgp->ip_vfs_context, + imgp->ip_vp, + imgp->ip_arch_offset, + imgp->ip_scriptvp, + imgp->ip_scriptlabelp, + imgp->ip_execlabelp, + &imgp->ip_csflags, + imgp->ip_px_smpx, + &disjoint_cred, /* will be non zero if disjoint */ + &label_update_return); thread_enable_send_importance(thread, FALSE); if (disjoint_cred) { @@ -4711,13 +5516,13 @@ handle_mac_transition: */ leave_sugid_clear = 0; } - + imgp->ip_mac_return = label_update_return; } - + mac_reset_ipc = mac_proc_check_inherit_ipc_ports(p, p->p_textvp, p->p_textoff, imgp->ip_vp, imgp->ip_arch_offset, imgp->ip_scriptvp); -#endif /* CONFIG_MACF */ +#endif /* CONFIG_MACF */ /* * If 'leave_sugid_clear' is non-zero, then we passed the @@ -4725,7 +5530,7 @@ handle_mac_transition: * the previous cred was a member of the VSGID group, but * that it was not the default at the time of the execve, * and that the post-labelling credential was not disjoint. - * So we don't set the P_SUGID or reset mach ports and fds + * So we don't set the P_SUGID or reset mach ports and fds * on the basis of simply running this code. */ if (mac_reset_ipc || !leave_sugid_clear) { @@ -4736,9 +5541,9 @@ handle_mac_transition: * task/thread after. */ ipc_task_reset((imgp->ip_new_thread != NULL) ? - get_threadtask(imgp->ip_new_thread) : p->task); + get_threadtask(imgp->ip_new_thread) : p->task); ipc_thread_reset((imgp->ip_new_thread != NULL) ? - imgp->ip_new_thread : current_thread()); + imgp->ip_new_thread : current_thread()); } if (!leave_sugid_clear) { @@ -4756,17 +5561,17 @@ handle_mac_transition: * to libc. */ for (i = 0; i < 3; i++) { - - if (p->p_fd->fd_ofiles[i] != NULL) + if (p->p_fd->fd_ofiles[i] != NULL) { continue; + } /* * Do the kernel equivalent of * - * if i == 0 - * (void) open("/dev/null", O_RDONLY); - * else - * (void) open("/dev/null", O_WRONLY); + * if i == 0 + * (void) open("/dev/null", O_RDONLY); + * else + * (void) open("/dev/null", O_WRONLY); */ struct fileproc *fp; @@ -4774,14 +5579,16 @@ handle_mac_transition: int flag; struct nameidata *ndp = NULL; - if (i == 0) + if (i == 0) { flag = FREAD; - else + } else { flag = FWRITE; + } if ((error = falloc(p, - &fp, &indx, imgp->ip_vfs_context)) != 0) + &fp, &indx, imgp->ip_vfs_context)) != 0) { continue; + } MALLOC(ndp, struct nameidata *, sizeof(*ndp), M_TEMP, M_WAITOK | M_ZERO); if (ndp == NULL) { @@ -4832,7 +5639,7 @@ handle_mac_transition: } } -#endif /* CONFIG_MACF */ +#endif /* CONFIG_MACF */ /* * Implement the semantic where the effective user and group become @@ -4842,38 +5649,14 @@ handle_mac_transition: * proc's ucred lock. This prevents others from accessing * a garbage credential. */ - for (;;) { - my_cred = kauth_cred_proc_ref(p); - my_new_cred = kauth_cred_setsvuidgid(my_cred, kauth_cred_getuid(my_cred), kauth_cred_getgid(my_cred)); - - if (my_new_cred == my_cred) { - kauth_cred_unref(&my_cred); - break; - } - - /* update cred on proc */ - proc_ucred_lock(p); - - if (p->p_ucred != my_cred) { - proc_ucred_unlock(p); - kauth_cred_unref(&my_new_cred); - continue; - } - - /* donate cred reference on my_new_cred to p->p_ucred */ - p->p_ucred = my_new_cred; - PROC_UPDATE_CREDS_ONPROC(p); - proc_ucred_unlock(p); - - /* drop additional reference that was taken on the previous cred */ - kauth_cred_unref(&my_cred); - - break; - } - + apply_kauth_cred_update(p, ^kauth_cred_t (kauth_cred_t my_cred) { + return kauth_cred_setsvuidgid(my_cred, + kauth_cred_getuid(my_cred), + kauth_cred_getgid(my_cred)); + }); /* Update the process' identity version and set the security token */ - p->p_idversion++; + p->p_idversion = OSIncrementAtomic(&nextpidversion); if (imgp->ip_new_thread != NULL) { task = get_threadtask(imgp->ip_new_thread); @@ -4882,7 +5665,7 @@ handle_mac_transition: } set_security_token_task_internal(p, task); - return(error); + return error; } @@ -4904,17 +5687,20 @@ handle_mac_transition: * !KERN_SUCCESS Mach failure code */ static kern_return_t -create_unix_stack(vm_map_t map, load_result_t* load_result, - proc_t p) +create_unix_stack(vm_map_t map, load_result_t* load_result, + proc_t p) { - mach_vm_size_t size, prot_size; - mach_vm_offset_t addr, prot_addr; - kern_return_t kr; + mach_vm_size_t size, prot_size; + mach_vm_offset_t addr, prot_addr; + kern_return_t kr; + + mach_vm_address_t user_stack = load_result->user_stack; - mach_vm_address_t user_stack = load_result->user_stack; - proc_lock(p); p->user_stack = user_stack; + if (load_result->custom_stack) { + p->p_lflag |= P_LCUSTOM_STACK; + } proc_unlock(p); if (load_result->user_stack_alloc_size > 0) { @@ -4930,12 +5716,12 @@ create_unix_stack(vm_map_t map, load_result_t* load_result, } addr = mach_vm_trunc_page(load_result->user_stack - size); kr = mach_vm_allocate_kernel(map, &addr, size, - VM_FLAGS_FIXED, VM_MEMORY_STACK); + VM_FLAGS_FIXED, VM_MEMORY_STACK); if (kr != KERN_SUCCESS) { // Can't allocate at default location, try anywhere addr = 0; kr = mach_vm_allocate_kernel(map, &addr, size, - VM_FLAGS_ANYWHERE, VM_MEMORY_STACK); + VM_FLAGS_ANYWHERE, VM_MEMORY_STACK); if (kr != KERN_SUCCESS) { return kr; } @@ -4955,7 +5741,9 @@ create_unix_stack(vm_map_t map, load_result_t* load_result, * size limit for this process. */ if (load_result->user_stack_size == 0) { + proc_list_lock(); load_result->user_stack_size = unix_stack_size(p); + proc_list_unlock(); prot_size = mach_vm_trunc_page(size - load_result->user_stack_size); } else { prot_size = PAGE_SIZE; @@ -4963,10 +5751,10 @@ create_unix_stack(vm_map_t map, load_result_t* load_result, prot_addr = addr; kr = mach_vm_protect(map, - prot_addr, - prot_size, - FALSE, - VM_PROT_NONE); + prot_addr, + prot_size, + FALSE, + VM_PROT_NONE); if (kr != KERN_SUCCESS) { (void)mach_vm_deallocate(map, addr, size); return kr; @@ -4989,7 +5777,7 @@ create_unix_stack(vm_map_t map, load_result_t* load_result, * path NULL terminated path * * Returns: KERN_SUCCESS Success - * !KERN_SUCCESS See execve/mac_execve for error codes + * !KERN_SUCCESS See execve/mac_execve for error codes * * Notes: The process that is passed in is the first manufactured * process on the system, and gets here via bsd_ast() firing @@ -5021,8 +5809,9 @@ load_init_program_at_path(proc_t p, user_addr_t scratch_addr, const char* path) size_t path_length = strlen(path) + 1; argv0 = scratch_addr; error = copyout(path, argv0, path_length); - if (error) + if (error) { return error; + } scratch_addr = USER_ADDR_ALIGN(scratch_addr + path_length, sizeof(user_addr_t)); @@ -5032,36 +5821,39 @@ load_init_program_at_path(proc_t p, user_addr_t scratch_addr, const char* path) */ if (boothowto & RB_SINGLE) { const char *init_args = "-s"; - size_t init_args_length = strlen(init_args)+1; + size_t init_args_length = strlen(init_args) + 1; argv1 = scratch_addr; error = copyout(init_args, argv1, init_args_length); - if (error) + if (error) { return error; + } scratch_addr = USER_ADDR_ALIGN(scratch_addr + init_args_length, sizeof(user_addr_t)); } if (proc_is64bit(p)) { - user64_addr_t argv64bit[3]; + user64_addr_t argv64bit[3] = {}; argv64bit[0] = argv0; argv64bit[1] = argv1; argv64bit[2] = USER_ADDR_NULL; error = copyout(argv64bit, scratch_addr, sizeof(argv64bit)); - if (error) + if (error) { return error; + } } else { - user32_addr_t argv32bit[3]; + user32_addr_t argv32bit[3] = {}; argv32bit[0] = (user32_addr_t)argv0; argv32bit[1] = (user32_addr_t)argv1; argv32bit[2] = USER_ADDR_NULL; error = copyout(argv32bit, scratch_addr, sizeof(argv32bit)); - if (error) + if (error) { return error; + } } /* @@ -5138,13 +5930,14 @@ load_init_program(proc_t p) if (PE_parse_boot_argn("launchdsuffix", launchd_suffix, sizeof(launchd_suffix))) { char launchd_path[128]; boolean_t is_release_suffix = ((launchd_suffix[0] == 0) || - (strcmp(launchd_suffix, "release") == 0)); + (strcmp(launchd_suffix, "release") == 0)); if (is_release_suffix) { printf("load_init_program: attempting to load /sbin/launchd\n"); error = load_init_program_at_path(p, (user_addr_t)scratch_addr, "/sbin/launchd"); - if (!error) + if (!error) { return; + } panic("Process 1 exec of launchd.release failed, errno %d", error); } else { @@ -5163,7 +5956,7 @@ load_init_program(proc_t p) #endif error = ENOENT; - for (i = 0; i < sizeof(init_programs)/sizeof(init_programs[0]); i++) { + for (i = 0; i < sizeof(init_programs) / sizeof(init_programs[0]); i++) { printf("load_init_program: attempting to load %s\n", init_programs[i]); error = load_init_program_at_path(p, (user_addr_t)scratch_addr, init_programs[i]); if (!error) { @@ -5173,7 +5966,7 @@ load_init_program(proc_t p) } } - panic("Process 1 exec of %s failed, errno %d", ((i == 0) ? "" : init_programs[i-1]), error); + panic("Process 1 exec of %s failed, errno %d", ((i == 0) ? "" : init_programs[i - 1]), error); } /* @@ -5195,7 +5988,7 @@ load_init_program(proc_t p) * EIO An I/O error occurred * EBADEXEC The executable is corrupt/unknown */ -static int +static int load_return_to_errno(load_return_t lrtn) { switch (lrtn) { @@ -5217,8 +6010,9 @@ load_return_to_errno(load_return_t lrtn) return ENOENT; case LOAD_IOERROR: return EIO; - case LOAD_FAILURE: case LOAD_DECRYPTFAIL: + return EAUTH; + case LOAD_FAILURE: default: return EBADEXEC; } @@ -5269,29 +6063,34 @@ static int execargs_waiters = 0; lck_mtx_t *execargs_cache_lock; static void -execargs_lock_lock(void) { +execargs_lock_lock(void) +{ lck_mtx_lock_spin(execargs_cache_lock); } static void -execargs_lock_unlock(void) { +execargs_lock_unlock(void) +{ lck_mtx_unlock(execargs_cache_lock); } static wait_result_t -execargs_lock_sleep(void) { - return(lck_mtx_sleep(execargs_cache_lock, LCK_SLEEP_DEFAULT, &execargs_free_count, THREAD_INTERRUPTIBLE)); +execargs_lock_sleep(void) +{ + return lck_mtx_sleep(execargs_cache_lock, LCK_SLEEP_DEFAULT, &execargs_free_count, THREAD_INTERRUPTIBLE); } static kern_return_t -execargs_purgeable_allocate(char **execarg_address) { +execargs_purgeable_allocate(char **execarg_address) +{ kern_return_t kr = vm_allocate_kernel(bsd_pageable_map, (vm_offset_t *)execarg_address, BSD_PAGEABLE_SIZE_PER_EXEC, VM_FLAGS_ANYWHERE | VM_FLAGS_PURGABLE, VM_KERN_MEMORY_NONE); assert(kr == KERN_SUCCESS); return kr; } static kern_return_t -execargs_purgeable_reference(void *execarg_address) { +execargs_purgeable_reference(void *execarg_address) +{ int state = VM_PURGABLE_NONVOLATILE; kern_return_t kr = vm_purgable_control(bsd_pageable_map, (vm_offset_t) execarg_address, VM_PURGABLE_SET_STATE, &state); @@ -5300,7 +6099,8 @@ execargs_purgeable_reference(void *execarg_address) { } static kern_return_t -execargs_purgeable_volatilize(void *execarg_address) { +execargs_purgeable_volatilize(void *execarg_address) +{ int state = VM_PURGABLE_VOLATILE | VM_PURGABLE_ORDERING_OBSOLETE; kern_return_t kr; kr = vm_purgable_control(bsd_pageable_map, (vm_offset_t) execarg_address, VM_PURGABLE_SET_STATE, &state); @@ -5311,7 +6111,8 @@ execargs_purgeable_volatilize(void *execarg_address) { } static void -execargs_wakeup_waiters(void) { +execargs_wakeup_waiters(void) +{ thread_wakeup(&execargs_free_count); } @@ -5330,7 +6131,7 @@ execargs_alloc(struct image_params *imgp) execargs_waiters--; if (res != THREAD_AWAKENED) { execargs_lock_unlock(); - return (EINTR); + return EINTR; } } @@ -5349,25 +6150,25 @@ execargs_alloc(struct image_params *imgp) assert(execargs_free_count >= 0); execargs_lock_unlock(); - + if (cache_index == -1) { kret = execargs_purgeable_allocate(&imgp->ip_strings); - } - else + } else { kret = execargs_purgeable_reference(imgp->ip_strings); + } assert(kret == KERN_SUCCESS); if (kret != KERN_SUCCESS) { - return (ENOMEM); + return ENOMEM; } /* last page used to read in file headers */ - imgp->ip_vdata = imgp->ip_strings + ( NCARGS + PAGE_SIZE ); + imgp->ip_vdata = imgp->ip_strings + (NCARGS + PAGE_SIZE); imgp->ip_strendp = imgp->ip_strings; imgp->ip_argspace = NCARGS; - imgp->ip_strspace = ( NCARGS + PAGE_SIZE ); + imgp->ip_strspace = (NCARGS + PAGE_SIZE); - return (0); + return 0; } /* @@ -5389,7 +6190,7 @@ execargs_free(struct image_params *imgp) kern_return_t kret; int i; boolean_t needs_wakeup = FALSE; - + kret = execargs_purgeable_volatilize(imgp->ip_strings); execargs_lock_lock(); @@ -5406,15 +6207,17 @@ execargs_free(struct image_params *imgp) assert(imgp->ip_strings == NULL); - if (execargs_waiters > 0) + if (execargs_waiters > 0) { needs_wakeup = TRUE; - + } + execargs_lock_unlock(); - if (needs_wakeup == TRUE) + if (needs_wakeup == TRUE) { execargs_wakeup_waiters(); + } - return ((kret == KERN_SUCCESS ? 0 : EINVAL)); + return kret == KERN_SUCCESS ? 0 : EINVAL; } static void @@ -5428,8 +6231,9 @@ exec_resettextvp(proc_t p, struct image_params *imgp) vp = imgp->ip_vp; offset = imgp->ip_arch_offset; - if (vp == NULLVP) + if (vp == NULLVP) { panic("exec_resettextvp: expected valid vp"); + } ret = vnode_ref(vp); proc_lock(p); @@ -5437,107 +6241,47 @@ exec_resettextvp(proc_t p, struct image_params *imgp) p->p_textvp = vp; p->p_textoff = offset; } else { - p->p_textvp = NULLVP; /* this is paranoia */ + p->p_textvp = NULLVP; /* this is paranoia */ p->p_textoff = 0; } proc_unlock(p); - if ( tvp != NULLVP) { + if (tvp != NULLVP) { if (vnode_getwithref(tvp) == 0) { vnode_rele(tvp); vnode_put(tvp); } - } - + } } // Includes the 0-byte (therefore "SIZE" instead of "LEN"). static const size_t CS_CDHASH_STRING_SIZE = CS_CDHASH_LEN * 2 + 1; -static void cdhash_to_string(char str[CS_CDHASH_STRING_SIZE], uint8_t const * const cdhash) { - static char const nibble[] = "0123456789abcdef"; - - /* Apparently still the safest way to get a hex representation - * of binary data. - * xnu's printf routines have %*D/%20D in theory, but "not really", see: - * confusion around %*D/%nD in printf - */ - for (int i = 0; i < CS_CDHASH_LEN; ++i) { - str[i*2] = nibble[(cdhash[i] & 0xf0) >> 4]; - str[i*2+1] = nibble[cdhash[i] & 0x0f]; - } - str[CS_CDHASH_STRING_SIZE - 1] = 0; -} - -/* - * If the process is not signed or if it contains entitlements, we - * need to communicate through the task_access_port to taskgated. - * - * taskgated will provide a detached code signature if present, and - * will enforce any restrictions on entitlements. - */ - -static boolean_t -taskgated_required(proc_t p, boolean_t *require_success) +static void +cdhash_to_string(char str[CS_CDHASH_STRING_SIZE], uint8_t const * const cdhash) { - size_t length; - void *blob; - int error; - - if (cs_debug > 2) - csvnode_print_debug(p->p_textvp); - -#if !CONFIG_EMBEDDED - const int can_skip_taskgated = csproc_get_platform_binary(p) && !csproc_get_platform_path(p); -#else - const int can_skip_taskgated = csproc_get_platform_binary(p); -#endif - if (can_skip_taskgated) { - if (cs_debug) printf("taskgated not required for: %s\n", p->p_name); - *require_success = FALSE; - return FALSE; - } - - if ((p->p_csflags & CS_VALID) == 0) { - *require_success = FALSE; - return TRUE; - } - - error = cs_entitlements_blob_get(p, &blob, &length); - if (error == 0 && blob != NULL) { -#if !CONFIG_EMBEDDED - /* - * fatal on the desktop when entitlements are present, - * unless we started in single-user mode - */ - if ((boothowto & RB_SINGLE) == 0) - *require_success = TRUE; - /* - * Allow initproc to run without causing taskgated to launch - */ - if (p == initproc) { - *require_success = FALSE; - return FALSE; - } + static char const nibble[] = "0123456789abcdef"; -#endif - if (cs_debug) printf("taskgated required for: %s\n", p->p_name); - - return TRUE; + /* Apparently still the safest way to get a hex representation + * of binary data. + * xnu's printf routines have %*D/%20D in theory, but "not really", see: + * confusion around %*D/%nD in printf + */ + for (int i = 0; i < CS_CDHASH_LEN; ++i) { + str[i * 2] = nibble[(cdhash[i] & 0xf0) >> 4]; + str[i * 2 + 1] = nibble[cdhash[i] & 0x0f]; } - - *require_success = FALSE; - return FALSE; + str[CS_CDHASH_STRING_SIZE - 1] = 0; } /* * __EXEC_WAITING_ON_TASKGATED_CODE_SIGNATURE_UPCALL__ - * + * * Description: Waits for the userspace daemon to respond to the request - * we made. Function declared non inline to be visible in + * we made. Function declared non inline to be visible in * stackshots and spindumps as well as debugging. */ -__attribute__((noinline)) int +__attribute__((noinline)) int __EXEC_WAITING_ON_TASKGATED_CODE_SIGNATURE_UPCALL__(mach_port_t task_access_port, int32_t new_pid) { return find_code_signature(task_access_port, new_pid); @@ -5546,11 +6290,11 @@ __EXEC_WAITING_ON_TASKGATED_CODE_SIGNATURE_UPCALL__(mach_port_t task_access_port static int check_for_signature(proc_t p, struct image_params *imgp) { - mach_port_t port = NULL; + mach_port_t port = IPC_PORT_NULL; kern_return_t kr = KERN_FAILURE; int error = EACCES; boolean_t unexpected_failure = FALSE; - unsigned char hash[CS_CDHASH_LEN]; + struct cs_blob *csb; boolean_t require_success = FALSE; int spawn = (imgp->ip_flags & IMGPF_SPAWN); int vfexec = (imgp->ip_flags & IMGPF_VFORK_EXEC); @@ -5566,7 +6310,7 @@ check_for_signature(proc_t p, struct image_params *imgp) proc_unlock(p); /* Set the switch_protect flag on the map */ - if(p->p_csflags & (CS_HARD|CS_KILL)) { + if (p->p_csflags & (CS_HARD | CS_KILL)) { vm_map_switch_protect(get_task_map(p->task), TRUE); } @@ -5576,9 +6320,8 @@ check_for_signature(proc_t p, struct image_params *imgp) * approve of exec, kill and return immediately. */ if (imgp->ip_mac_return != 0) { - KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_SECURITY_POLICY, 0, 0); + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_SECURITY_POLICY, 0, 0); signature_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_SECURITY_POLICY); error = imgp->ip_mac_return; unexpected_failure = TRUE; @@ -5592,18 +6335,23 @@ check_for_signature(proc_t p, struct image_params *imgp) goto done; } - /* check if callout to taskgated is needed */ - if (!taskgated_required(p, &require_success)) { + /* If the code signature came through the image activation path, we skip the + * taskgated / externally attached path. */ + if (imgp->ip_csflags & CS_SIGNED) { error = 0; goto done; } + /* The rest of the code is for signatures that either already have been externally + * attached (likely, but not necessarily by a previous run through the taskgated + * path), or that will now be attached by taskgated. */ + kr = task_get_task_access_port(p->task, &port); if (KERN_SUCCESS != kr || !IPC_PORT_VALID(port)) { error = 0; if (require_success) { KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_CODESIGNING, CODESIGNING_EXIT_REASON_TASK_ACCESS_PORT, 0, 0); + p->p_pid, OS_REASON_CODESIGNING, CODESIGNING_EXIT_REASON_TASK_ACCESS_PORT, 0, 0); signature_failure_reason = os_reason_create(OS_REASON_CODESIGNING, CODESIGNING_EXIT_REASON_TASK_ACCESS_PORT); error = EACCES; } @@ -5612,9 +6360,9 @@ check_for_signature(proc_t p, struct image_params *imgp) /* * taskgated returns KERN_SUCCESS if it has completed its work - * and the exec should continue, KERN_FAILURE if the exec should - * fail, or it may error out with different error code in an - * event of mig failure (e.g. process was signalled during the + * and the exec should continue, KERN_FAILURE if the exec should + * fail, or it may error out with different error code in an + * event of mig failure (e.g. process was signalled during the * rpc call, taskgated died, mig server died etc.). */ @@ -5627,14 +6375,14 @@ check_for_signature(proc_t p, struct image_params *imgp) error = EACCES; KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_CODESIGNING, CODESIGNING_EXIT_REASON_TASKGATED_INVALID_SIG, 0, 0); + p->p_pid, OS_REASON_CODESIGNING, CODESIGNING_EXIT_REASON_TASKGATED_INVALID_SIG, 0, 0); signature_failure_reason = os_reason_create(OS_REASON_CODESIGNING, CODESIGNING_EXIT_REASON_TASKGATED_INVALID_SIG); goto done; default: error = EACCES; KERNEL_DEBUG_CONSTANT(BSDDBG_CODE(DBG_BSD_PROC, BSD_PROC_EXITREASON_CREATE) | DBG_FUNC_NONE, - p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_TASKGATED_OTHER, 0, 0); + p->p_pid, OS_REASON_EXEC, EXEC_EXIT_REASON_TASKGATED_OTHER, 0, 0); signature_failure_reason = os_reason_create(OS_REASON_EXEC, EXEC_EXIT_REASON_TASKGATED_OTHER); unexpected_failure = TRUE; goto done; @@ -5642,14 +6390,39 @@ check_for_signature(proc_t p, struct image_params *imgp) /* Only do this if exec_resettextvp() did not fail */ if (p->p_textvp != NULLVP) { - /* - * If there's a new code directory, mark this process - * as signed. - */ - if (0 == ubc_cs_getcdhash(p->p_textvp, p->p_textoff, hash)) { - proc_lock(p); - p->p_csflags |= CS_VALID; - proc_unlock(p); + csb = ubc_cs_blob_get(p->p_textvp, -1, p->p_textoff); + + if (csb != NULL) { + /* As the enforcement we can do here is very limited, we only allow things that + * are the only reason why this code path still exists: + * Adhoc signed non-platform binaries without special cs_flags and without any + * entitlements (unrestricted ones still pass AMFI). */ + if ( + /* Revalidate the blob if necessary through bumped generation count. */ + (ubc_cs_generation_check(p->p_textvp) == 0 || + ubc_cs_blob_revalidate(p->p_textvp, csb, imgp, 0) == 0) && + /* Only CS_ADHOC, no CS_KILL, CS_HARD etc. */ + (csb->csb_flags & CS_ALLOWED_MACHO) == CS_ADHOC && + /* If it has a CMS blob, it's not adhoc. The CS_ADHOC flag can lie. */ + csblob_find_blob_bytes((const uint8_t *)csb->csb_mem_kaddr, csb->csb_mem_size, + CSSLOT_SIGNATURESLOT, + CSMAGIC_BLOBWRAPPER) == NULL && + /* It could still be in a trust cache (unlikely with CS_ADHOC), or a magic path. */ + csb->csb_platform_binary == 0 && + /* No entitlements, not even unrestricted ones. */ + csb->csb_entitlements_blob == NULL) { + proc_lock(p); + p->p_csflags |= CS_SIGNED | CS_VALID; + proc_unlock(p); + } else { + uint8_t cdhash[CS_CDHASH_LEN]; + char cdhash_string[CS_CDHASH_STRING_SIZE]; + proc_getcdhash(p, cdhash); + cdhash_to_string(cdhash_string, cdhash); + printf("ignoring detached code signature on '%s' with cdhash '%s' " + "because it is invalid, or not a simple adhoc signature.\n", + p->p_name, cdhash_string); + } } } @@ -5665,16 +6438,17 @@ done: cdhash_to_string(cdhash_string, cdhash); os_log(peLog, "CS Platform Exec Logging: Executing platform signed binary " - "'%s' with cdhash %s\n", p->p_name, cdhash_string); + "'%s' with cdhash %s\n", p->p_name, cdhash_string); } } else { - if (!unexpected_failure) + if (!unexpected_failure) { p->p_csflags |= CS_KILLED; + } /* make very sure execution fails */ if (vfexec || spawn) { assert(signature_failure_reason != OS_REASON_NULL); psignal_vfork_with_reason(p, p->task, imgp->ip_new_thread, - SIGKILL, signature_failure_reason); + SIGKILL, signature_failure_reason); signature_failure_reason = OS_REASON_NULL; error = 0; } else { @@ -5684,6 +6458,10 @@ done: } } + if (port != IPC_PORT_NULL) { + ipc_port_release_send(port); + } + /* If we hit this, we likely would have leaked an exit reason */ assert(signature_failure_reason == OS_REASON_NULL); return error; @@ -5699,7 +6477,8 @@ done: * in the process' page tables, we prefault some pages if * possible. Errors are non-fatal. */ -static void exec_prefault_data(proc_t p __unused, struct image_params *imgp, load_result_t *load_result) +static void +exec_prefault_data(proc_t p __unused, struct image_params *imgp, load_result_t *load_result) { int ret; size_t expected_all_image_infos_size; @@ -5708,13 +6487,13 @@ static void exec_prefault_data(proc_t p __unused, struct image_params *imgp, loa * Prefault executable or dyld entry point. */ vm_fault(current_map(), - vm_map_trunc_page(load_result->entry_point, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_EXECUTE, - FALSE, VM_KERN_MEMORY_NONE, - THREAD_UNINT, NULL, 0); - - if (imgp->ip_flags & IMGPF_IS_64BIT) { + vm_map_trunc_page(load_result->entry_point, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_EXECUTE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); + + if (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) { expected_all_image_infos_size = sizeof(struct user64_dyld_all_image_infos); } else { expected_all_image_infos_size = sizeof(struct user32_dyld_all_image_infos); @@ -5722,11 +6501,11 @@ static void exec_prefault_data(proc_t p __unused, struct image_params *imgp, loa /* Decode dyld anchor structure from */ if (load_result->dynlinker && - load_result->all_image_info_addr && - load_result->all_image_info_size >= expected_all_image_infos_size) { + load_result->all_image_info_addr && + load_result->all_image_info_size >= expected_all_image_infos_size) { union { - struct user64_dyld_all_image_infos infos64; - struct user32_dyld_all_image_infos infos32; + struct user64_dyld_all_image_infos infos64; + struct user32_dyld_all_image_infos infos32; } all_image_infos; /* @@ -5734,33 +6513,32 @@ static void exec_prefault_data(proc_t p __unused, struct image_params *imgp, loa * and recovery path. */ vm_fault(current_map(), - vm_map_trunc_page(load_result->all_image_info_addr, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_WRITE, - FALSE, VM_KERN_MEMORY_NONE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(load_result->all_image_info_addr, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_WRITE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); if ((load_result->all_image_info_addr & PAGE_MASK) + expected_all_image_infos_size > PAGE_SIZE) { /* all_image_infos straddles a page */ vm_fault(current_map(), - vm_map_trunc_page(load_result->all_image_info_addr + expected_all_image_infos_size - 1, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_WRITE, - FALSE, VM_KERN_MEMORY_NONE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(load_result->all_image_info_addr + expected_all_image_infos_size - 1, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_WRITE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); } ret = copyin(load_result->all_image_info_addr, - &all_image_infos, - expected_all_image_infos_size); + &all_image_infos, + expected_all_image_infos_size); if (ret == 0 && all_image_infos.infos32.version >= DYLD_ALL_IMAGE_INFOS_ADDRESS_MINIMUM_VERSION) { - user_addr_t notification_address; user_addr_t dyld_image_address; user_addr_t dyld_version_address; user_addr_t dyld_all_image_infos_address; user_addr_t dyld_slide_amount; - if (imgp->ip_flags & IMGPF_IS_64BIT) { + if (imgp->ip_flags & IMGPF_IS_64BIT_ADDR) { notification_address = all_image_infos.infos64.notification; dyld_image_address = all_image_infos.infos64.dyldImageLoadAddress; dyld_version_address = all_image_infos.infos64.dyldVersion; @@ -5789,38 +6567,38 @@ static void exec_prefault_data(proc_t p __unused, struct image_params *imgp, loa #if 0 kprintf("exec_prefault: 0x%016llx 0x%08x 0x%016llx 0x%016llx 0x%016llx 0x%016llx\n", - (uint64_t)load_result->all_image_info_addr, - all_image_infos.infos32.version, - (uint64_t)notification_address, - (uint64_t)dyld_image_address, - (uint64_t)dyld_version_address, - (uint64_t)dyld_all_image_infos_address); + (uint64_t)load_result->all_image_info_addr, + all_image_infos.infos32.version, + (uint64_t)notification_address, + (uint64_t)dyld_image_address, + (uint64_t)dyld_version_address, + (uint64_t)dyld_all_image_infos_address); #endif vm_fault(current_map(), - vm_map_trunc_page(notification_address + dyld_slide_amount, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_EXECUTE, - FALSE, VM_KERN_MEMORY_NONE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(notification_address + dyld_slide_amount, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_EXECUTE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); vm_fault(current_map(), - vm_map_trunc_page(dyld_image_address + dyld_slide_amount, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_EXECUTE, - FALSE, VM_KERN_MEMORY_NONE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(dyld_image_address + dyld_slide_amount, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_EXECUTE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); vm_fault(current_map(), - vm_map_trunc_page(dyld_version_address + dyld_slide_amount, - vm_map_page_mask(current_map())), - VM_PROT_READ, - FALSE, VM_KERN_MEMORY_NONE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(dyld_version_address + dyld_slide_amount, + vm_map_page_mask(current_map())), + VM_PROT_READ, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); vm_fault(current_map(), - vm_map_trunc_page(dyld_all_image_infos_address + dyld_slide_amount, - vm_map_page_mask(current_map())), - VM_PROT_READ | VM_PROT_WRITE, - FALSE, VM_KERN_MEMORY_NONE, - THREAD_UNINT, NULL, 0); + vm_map_trunc_page(dyld_all_image_infos_address + dyld_slide_amount, + vm_map_page_mask(current_map())), + VM_PROT_READ | VM_PROT_WRITE, + FALSE, VM_KERN_MEMORY_NONE, + THREAD_UNINT, NULL, 0); } } }