X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/0c530ab8987f0ae6a1a3d9284f40182b88852816..c331a0bec715536613c8dd5f34a4e115d5b15824:/bsd/vm/vm_unix.c?ds=sidebyside diff --git a/bsd/vm/vm_unix.c b/bsd/vm/vm_unix.c index b1bcfddd5..369c91350 100644 --- a/bsd/vm/vm_unix.c +++ b/bsd/vm/vm_unix.c @@ -1,23 +1,29 @@ /* - * Copyright (c) 2000-2004 Apple Computer, Inc. All rights reserved. + * Copyright (c) 2000-2007 Apple Inc. All rights reserved. * - * @APPLE_LICENSE_HEADER_START@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_START@ * - * The contents of this file constitute Original Code as defined in and - * are subject to the Apple Public Source License Version 1.1 (the - * "License"). You may not use this file except in compliance with the - * License. Please obtain a copy of the License at - * http://www.apple.com/publicsource and read it before using this file. + * 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 + * compliance with the License. The rights granted to you under the License + * may not be used to create, or enable the creation or redistribution of, + * 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. * - * This Original Code and all software distributed under the License are - * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER + * 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, * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT. Please see the - * License for the specific language governing rights and limitations - * under the License. + * 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_LICENSE_HEADER_END@ + * @APPLE_OSREFERENCE_LICENSE_HEADER_END@ */ /* * Mach Operating System @@ -25,11 +31,13 @@ * All rights reserved. The CMU software License Agreement specifies * the terms and conditions for use and redistribution. */ - /* + * NOTICE: This file was modified by SPARTA, Inc. in 2006 to introduce + * support for mandatory and extensible security protections. This notice + * is included in support of clause 2.2 (b) of the Apple Public License, + * Version 2.0. */ - #include #include @@ -37,11 +45,14 @@ #include #include #include +#include +#include +#include +#include #include #include #include #include -#include #include #include @@ -64,26 +75,87 @@ #include #include -#include +#include #include #include #include #include +#include #include -#include -#include +#include +#include #include +/* + * Sysctl's related to data/stack execution. See osfmk/vm/vm_map.c + */ + +#ifndef SECURE_KERNEL +extern int allow_stack_exec, allow_data_exec; + +SYSCTL_INT(_vm, OID_AUTO, allow_stack_exec, CTLFLAG_RW, &allow_stack_exec, 0, ""); +SYSCTL_INT(_vm, OID_AUTO, allow_data_exec, CTLFLAG_RW, &allow_data_exec, 0, ""); +#endif /* !SECURE_KERNEL */ + +static const char *prot_values[] = { + "none", + "read-only", + "write-only", + "read-write", + "execute-only", + "read-execute", + "write-execute", + "read-write-execute" +}; + void -log_nx_failure(addr64_t vaddr, vm_prot_t prot) +log_stack_execution_failure(addr64_t vaddr, vm_prot_t prot) { - printf("NX failure: %s - vaddr=%qx, prot=%x\n", current_proc()->p_comm, vaddr, prot); + printf("Data/Stack execution not permitted: %s[pid %d] at virtual address 0x%qx, protections were %s\n", + current_proc()->p_comm, current_proc()->p_pid, vaddr, prot_values[prot & VM_PROT_ALL]); } +int shared_region_unnest_logging = 1; + +SYSCTL_INT(_vm, OID_AUTO, shared_region_unnest_logging, CTLFLAG_RW, + &shared_region_unnest_logging, 0, ""); + +int vm_shared_region_unnest_log_interval = 10; +int shared_region_unnest_log_count_threshold = 5; + +/* These log rate throttling state variables aren't thread safe, but + * are sufficient unto the task. + */ +static int64_t last_unnest_log_time = 0; +static int shared_region_unnest_log_count = 0; + +void log_unnest_badness(vm_map_t m, vm_map_offset_t s, vm_map_offset_t e) { + struct timeval tv; + const char *pcommstr; + + if (shared_region_unnest_logging == 0) + return; + + if (shared_region_unnest_logging == 1) { + microtime(&tv); + if ((tv.tv_sec - last_unnest_log_time) < vm_shared_region_unnest_log_interval) { + if (shared_region_unnest_log_count++ > shared_region_unnest_log_count_threshold) + return; + } + else { + last_unnest_log_time = tv.tv_sec; + shared_region_unnest_log_count = 0; + } + } + + pcommstr = current_proc()->p_comm; + + printf("%s (map: %p) triggered DYLD shared region unnest for map: %p, region 0x%qx->0x%qx. While not abnormal for debuggers, this increases system memory footprint until the target exits.\n", current_proc()->p_comm, get_task_map(current_proc()->task), m, (uint64_t)s, (uint64_t)e); +} int useracc( @@ -214,7 +286,7 @@ suword( long fuword(user_addr_t addr) { - long word; + long word = 0; if (copyin(addr, (void *) &word, sizeof(int))) return(-1); @@ -233,7 +305,7 @@ suiword( long fuiword(user_addr_t addr) { - long word; + long word = 0; if (copyin(addr, (void *) &word, sizeof(int))) return(-1); @@ -276,7 +348,7 @@ suulong(user_addr_t addr, uint64_t uword) if (IS_64BIT_PROCESS(current_proc())) { return(copyout((void *)&uword, addr, sizeof(uword)) == 0 ? 0 : -1); } else { - return(suiword(addr, (u_long)uword)); + return(suiword(addr, (uint32_t)uword)); } } @@ -295,28 +367,42 @@ fuulong(user_addr_t addr) } int -swapon(__unused struct proc *procp, __unused struct swapon_args *uap, __unused int *retval) +swapon(__unused proc_t procp, __unused struct swapon_args *uap, __unused int *retval) { return(ENOTSUP); } - +/* + * pid_for_task + * + * Find the BSD process ID for the Mach task associated with the given Mach port + * name + * + * Parameters: args User argument descriptor (see below) + * + * Indirect parameters: args->t Mach port name + * args->pid Process ID (returned value; see below) + * + * Returns: KERL_SUCCESS Success + * KERN_FAILURE Not success + * + * Implicit returns: args->pid Process ID + * + */ kern_return_t pid_for_task( struct pid_for_task_args *args) { mach_port_name_t t = args->t; user_addr_t pid_addr = args->pid; - struct proc * p; + proc_t p; task_t t1; int pid = -1; kern_return_t err = KERN_SUCCESS; - boolean_t funnel_state; AUDIT_MACH_SYSCALL_ENTER(AUE_PIDFORTASK); AUDIT_ARG(mach_port1, t); - funnel_state = thread_funnel_set(kernel_flock, TRUE); t1 = port_name_to_task(t); if (t1 == TASK_NULL) { @@ -335,11 +421,95 @@ pid_for_task( pftout: AUDIT_ARG(pid, pid); (void) copyout((char *) &pid, pid_addr, sizeof(int)); - thread_funnel_set(kernel_flock, funnel_state); AUDIT_MACH_SYSCALL_EXIT(err); return(err); } +/* + * + * tfp_policy = KERN_TFP_POLICY_DENY; Deny Mode: None allowed except for self + * tfp_policy = KERN_TFP_POLICY_DEFAULT; default mode: all posix checks and upcall via task port for authentication + * + */ +static int tfp_policy = KERN_TFP_POLICY_DEFAULT; + +/* + * Routine: task_for_pid_posix_check + * Purpose: + * Verify that the current process should be allowed to + * get the target process's task port. This is only + * permitted if: + * - The current process is root + * OR all of the following are true: + * - The target process's real, effective, and saved uids + * are the same as the current proc's euid, + * - The target process's group set is a subset of the + * calling process's group set, and + * - The target process hasn't switched credentials. + * + * Returns: TRUE: permitted + * FALSE: denied + */ +static int +task_for_pid_posix_check(proc_t target) +{ + kauth_cred_t targetcred, mycred; + uid_t myuid; + int allowed; + + /* No task_for_pid on bad targets */ + if (target == PROC_NULL || target->p_stat == SZOMB) { + return FALSE; + } + + mycred = kauth_cred_get(); + myuid = kauth_cred_getuid(mycred); + + /* If we're running as root, the check passes */ + if (kauth_cred_issuser(mycred)) + return TRUE; + + /* We're allowed to get our own task port */ + if (target == current_proc()) + return TRUE; + + /* + * Under DENY, only root can get another proc's task port, + * so no more checks are needed. + */ + if (tfp_policy == KERN_TFP_POLICY_DENY) { + return FALSE; + } + + targetcred = kauth_cred_proc_ref(target); + allowed = TRUE; + + /* Do target's ruid, euid, and saved uid match my euid? */ + if ((kauth_cred_getuid(targetcred) != myuid) || + (targetcred->cr_ruid != myuid) || + (targetcred->cr_svuid != myuid)) { + allowed = FALSE; + goto out; + } + + /* Are target's groups a subset of my groups? */ + if (kauth_cred_gid_subset(targetcred, mycred, &allowed) || + allowed == 0) { + allowed = FALSE; + goto out; + } + + /* Has target switched credentials? */ + if (target->p_flag & P_SUGID) { + allowed = FALSE; + goto out; + } + +out: + kauth_cred_unref(&targetcred); + return allowed; +} + /* * Routine: task_for_pid * Purpose: @@ -349,21 +519,10 @@ pftout: * Only permitted to privileged processes, or processes * with the same user ID. * - * XXX This should be a BSD system call, not a Mach trap!!! - */ -/* - * - * tfp_policy = KERN_TFP_POLICY_DENY; Deny Mode: None allowed except for self - * tfp_policy = KERN_TFP_POLICY_PERMISSIVE; Permissive Mode: all permissive; related ones allowed or privileged - * tfp_policy = KERN_TFP_POLICY_RESTRICTED; Restricted Mode: self access allowed; setgid (to tfp_group) are allowed for other tasks + * Note: if pid == 0, an error is return no matter who is calling. * + * XXX This should be a BSD system call, not a Mach trap!!! */ -static int tfp_policy = KERN_TFP_POLICY_RESTRICTED; -/* the groutp is inited to kmem group and is modifiable by sysctl */ -static int tfp_group_inited = 0; /* policy groups are loaded ... */ -static gid_t tfp_group_ronly = 0; /* procview group */ -static gid_t tfp_group_rw = 0; /* procmod group */ - kern_return_t task_for_pid( struct task_for_pid_args *args) @@ -371,152 +530,89 @@ task_for_pid( mach_port_name_t target_tport = args->target_tport; int pid = args->pid; user_addr_t task_addr = args->t; - struct uthread *uthread; - struct proc *p; - struct proc *p1; - task_t t1; - mach_port_name_t tret; + proc_t p = PROC_NULL; + task_t t1 = TASK_NULL; + mach_port_name_t tret = MACH_PORT_NULL; + ipc_port_t tfpport; void * sright; int error = 0; - int is_member = 0; - boolean_t funnel_state; - boolean_t ispermitted = FALSE; -#if DIAGNOSTIC - char procname[MAXCOMLEN+1]; -#endif /* DIAGNOSTIC */ AUDIT_MACH_SYSCALL_ENTER(AUE_TASKFORPID); AUDIT_ARG(pid, pid); AUDIT_ARG(mach_port1, target_tport); + /* Always check if pid == 0 */ + if (pid == 0) { + (void ) copyout((char *)&t1, task_addr, sizeof(mach_port_name_t)); + AUDIT_MACH_SYSCALL_EXIT(KERN_FAILURE); + return(KERN_FAILURE); + } + t1 = port_name_to_task(target_tport); if (t1 == TASK_NULL) { - (void ) copyout((char *)&t1, task_addr, sizeof(mach_port_name_t)); + (void) copyout((char *)&t1, task_addr, sizeof(mach_port_name_t)); AUDIT_MACH_SYSCALL_EXIT(KERN_FAILURE); return(KERN_FAILURE); } - funnel_state = thread_funnel_set(kernel_flock, TRUE); - p1 = current_proc(); + p = proc_find(pid); +#if CONFIG_AUDIT + if (p != PROC_NULL) + AUDIT_ARG(process, p); +#endif - /* - * Delayed binding of thread credential to process credential, if we - * are not running with an explicitly set thread credential. - */ - uthread = get_bsdthread_info(current_thread()); - if (uthread->uu_ucred != p1->p_ucred && - (uthread->uu_flag & UT_SETUID) == 0) { - kauth_cred_t old = uthread->uu_ucred; - proc_lock(p1); - kauth_cred_ref(p1->p_ucred); - uthread->uu_ucred = p1->p_ucred; - proc_unlock(p1); - if (IS_VALID_CRED(old)) - kauth_cred_unref(&old); + if (!(task_for_pid_posix_check(p))) { + error = KERN_FAILURE; + goto tfpout; } - p = pfind(pid); - AUDIT_ARG(process, p); - - /* - * XXX p_ucred check can be bogus in multithreaded processes, - * XXX unless the funnel is held. - */ - switch (tfp_policy) { - - case KERN_TFP_POLICY_PERMISSIVE: - /* self or suser or related ones */ - if ((p != (struct proc *) 0) - && (p->p_stat != SZOMB) - && (p1 != (struct proc *) 0) - && ( - (p1 == p) - || !(suser(kauth_cred_get(), 0)) - || ((kauth_cred_getuid(p->p_ucred) == kauth_cred_getuid(kauth_cred_get())) && - ((p->p_ucred->cr_ruid == kauth_cred_get()->cr_ruid)) - && ((p->p_flag & P_SUGID) == 0)) - ) - ) - ispermitted = TRUE; - break; - - case KERN_TFP_POLICY_RESTRICTED: - /* self or suser or setgid and related ones only */ - if ((p != (struct proc *) 0) - && (p1 != (struct proc *) 0) - && (p->p_stat != SZOMB) - && ( - (p1 == p) - || !(suser(kauth_cred_get(), 0)) - || (((tfp_group_inited != 0) && - ( - ((kauth_cred_ismember_gid(kauth_cred_get(), - tfp_group_ronly, &is_member) == 0) && is_member) - ||((kauth_cred_ismember_gid(kauth_cred_get(), - tfp_group_rw, &is_member) == 0) && is_member) - ) - ) - && ((kauth_cred_getuid(p->p_ucred) == kauth_cred_getuid(kauth_cred_get())) && - ((p->p_ucred->cr_ruid == kauth_cred_get()->cr_ruid)) - && ((p->p_flag & P_SUGID) == 0)) - ) - ) - ) - ispermitted = TRUE; + if (p->task != TASK_NULL) { + /* If we aren't root and target's task access port is set... */ + if (!kauth_cred_issuser(kauth_cred_get()) && + p != current_proc() && + (task_get_task_access_port(p->task, &tfpport) == 0) && + (tfpport != IPC_PORT_NULL)) { - break; - - case KERN_TFP_POLICY_DENY: - /* self or suser only */ - default: - /* do not return task port of other task at all */ - if ((p1 != (struct proc *) 0) && (p != (struct proc *) 0) && (p->p_stat != SZOMB) - && ((p1 == p) || !(suser(kauth_cred_get(), 0)))) - ispermitted = TRUE; - else - ispermitted = FALSE; - break; - }; + if (tfpport == IPC_PORT_DEAD) { + error = KERN_PROTECTION_FAILURE; + goto tfpout; + } + /* Call up to the task access server */ + error = check_task_access(tfpport, proc_selfpid(), kauth_getgid(), pid); - if (ispermitted == TRUE) { - if (p->task != TASK_NULL) { - task_reference(p->task); - sright = (void *)convert_task_to_port(p->task); - tret = ipc_port_copyout_send( - sright, - get_task_ipcspace(current_task())); - } else - tret = MACH_PORT_NULL; - AUDIT_ARG(mach_port2, tret); - (void ) copyout((char *)&tret, task_addr, sizeof(mach_port_name_t)); - task_deallocate(t1); - error = KERN_SUCCESS; + if (error != MACH_MSG_SUCCESS) { + if (error == MACH_RCV_INTERRUPTED) + error = KERN_ABORTED; + else + error = KERN_FAILURE; + goto tfpout; + } + } +#if CONFIG_MACF + error = mac_proc_check_get_task(kauth_cred_get(), p); + if (error) { + error = KERN_FAILURE; goto tfpout; - } -#if DIAGNOSTIC - else { - /* - * There is no guarantee that p_comm is null terminated and - * kernel implementation of string functions are complete. So - * ensure stale info is not leaked out, bzero the buffer - */ - bzero(&procname[0], MAXCOMLEN+1); - strncpy(&procname[0], &p1->p_comm[0], MAXCOMLEN); - if (tfp_policy != KERN_TFP_POLICY_PERMISSIVE) - log(LOG_NOTICE, "(%d: %s)tfp: failed on %d:\n", - ((p1 != PROC_NULL)?(p1->p_pid):0), &procname[0], - ((p != PROC_NULL)?(p->p_pid):0)); - } -#endif /* DIAGNOSTIC */ + } +#endif + + /* Grant task port access */ + task_reference(p->task); + sright = (void *) convert_task_to_port(p->task); + tret = ipc_port_copyout_send( + sright, + get_task_ipcspace(current_task())); + } + error = KERN_SUCCESS; - task_deallocate(t1); - tret = MACH_PORT_NULL; - (void) copyout((char *) &tret, task_addr, sizeof(mach_port_name_t)); - error = KERN_FAILURE; tfpout: - thread_funnel_set(kernel_flock, funnel_state); + task_deallocate(t1); + AUDIT_ARG(mach_port2, tret); + (void) copyout((char *) &tret, task_addr, sizeof(mach_port_name_t)); + if (p != PROC_NULL) + proc_rele(p); AUDIT_MACH_SYSCALL_EXIT(error); return(error); } @@ -540,14 +636,12 @@ task_name_for_pid( mach_port_name_t target_tport = args->target_tport; int pid = args->pid; user_addr_t task_addr = args->t; - struct uthread *uthread; - struct proc *p; - struct proc *p1; + proc_t p = PROC_NULL; task_t t1; mach_port_name_t tret; void * sright; - int error = 0; - boolean_t funnel_state; + int error = 0, refheld = 0; + kauth_cred_t target_cred; AUDIT_MACH_SYSCALL_ENTER(AUE_TASKNAMEFORPID); AUDIT_ARG(pid, pid); @@ -555,102 +649,202 @@ task_name_for_pid( t1 = port_name_to_task(target_tport); if (t1 == TASK_NULL) { - (void ) copyout((char *)&t1, task_addr, sizeof(mach_port_name_t)); + (void) copyout((char *)&t1, task_addr, sizeof(mach_port_name_t)); AUDIT_MACH_SYSCALL_EXIT(KERN_FAILURE); return(KERN_FAILURE); } - funnel_state = thread_funnel_set(kernel_flock, TRUE); - - p1 = current_proc(); - - /* - * Delayed binding of thread credential to process credential, if we - * are not running with an explicitly set thread credential. - */ - /* - * XXX p_ucred check can be bogus in multithreaded processes, - * XXX unless the funnel is held. - */ - uthread = get_bsdthread_info(current_thread()); - if (uthread->uu_ucred != p1->p_ucred && - (uthread->uu_flag & UT_SETUID) == 0) { - kauth_cred_t old = uthread->uu_ucred; - proc_lock(p1); - kauth_cred_ref(p1->p_ucred); - uthread->uu_ucred = p1->p_ucred; - proc_unlock(p1); - if (IS_VALID_CRED(old)) - kauth_cred_unref(&old); - } - - p = pfind(pid); - AUDIT_ARG(process, p); - - if ((p != (struct proc *) 0) - && (p->p_stat != SZOMB) - && (p1 != (struct proc *) 0) - && ((p1 == p) - || !(suser(kauth_cred_get(), 0)) - || ((kauth_cred_getuid(p->p_ucred) == kauth_cred_getuid(kauth_cred_get())) && - ((p->p_ucred->cr_ruid == kauth_cred_get()->cr_ruid))))) - { - if (p->task != TASK_NULL) - { - task_reference(p->task); - sright = (void *)convert_task_name_to_port(p->task); - tret = ipc_port_copyout_send( - sright, + p = proc_find(pid); + if (p != PROC_NULL) { + AUDIT_ARG(process, p); + target_cred = kauth_cred_proc_ref(p); + refheld = 1; + + if ((p->p_stat != SZOMB) + && ((current_proc() == p) + || kauth_cred_issuser(kauth_cred_get()) + || ((kauth_cred_getuid(target_cred) == kauth_cred_getuid(kauth_cred_get())) && + ((target_cred->cr_ruid == kauth_cred_get()->cr_ruid))))) { + + if (p->task != TASK_NULL) { + task_reference(p->task); +#if CONFIG_MACF + error = mac_proc_check_get_task_name(kauth_cred_get(), p); + if (error) { + task_deallocate(p->task); + goto noperm; + } +#endif + sright = (void *)convert_task_name_to_port(p->task); + tret = ipc_port_copyout_send(sright, get_task_ipcspace(current_task())); - } else - tret = MACH_PORT_NULL; - AUDIT_ARG(mach_port2, tret); - (void ) copyout((char *)&tret, task_addr, sizeof(mach_port_name_t)); - task_deallocate(t1); - error = KERN_SUCCESS; - goto tnfpout; + } else + tret = MACH_PORT_NULL; + + AUDIT_ARG(mach_port2, tret); + (void) copyout((char *)&tret, task_addr, sizeof(mach_port_name_t)); + task_deallocate(t1); + error = KERN_SUCCESS; + goto tnfpout; + } } - task_deallocate(t1); +#if CONFIG_MACF +noperm: +#endif + task_deallocate(t1); tret = MACH_PORT_NULL; (void) copyout((char *) &tret, task_addr, sizeof(mach_port_name_t)); error = KERN_FAILURE; tnfpout: - thread_funnel_set(kernel_flock, funnel_state); + if (refheld != 0) + kauth_cred_unref(&target_cred); + if (p != PROC_NULL) + proc_rele(p); AUDIT_MACH_SYSCALL_EXIT(error); return(error); } -static int -sysctl_settfp_policy(__unused struct sysctl_oid *oidp, void *arg1, - __unused int arg2, struct sysctl_req *req) +kern_return_t +pid_suspend(struct proc *p __unused, struct pid_suspend_args *args, int *ret) { - int error = 0; - int new_value; + task_t target = NULL; + proc_t targetproc = PROC_NULL; + int pid = args->pid; + int error = 0; - error = SYSCTL_OUT(req, arg1, sizeof(int)); - if (error || req->newptr == USER_ADDR_NULL) - return(error); +#if CONFIG_MACF + error = mac_proc_check_suspend_resume(p, 0); /* 0 for suspend */ + if (error) { + error = KERN_FAILURE; + goto out; + } +#endif - if (!is_suser()) - return(EPERM); + if (pid == 0) { + error = KERN_FAILURE; + goto out; + } - if ((error = SYSCTL_IN(req, &new_value, sizeof(int)))) { + targetproc = proc_find(pid); + if (!task_for_pid_posix_check(targetproc)) { + error = KERN_FAILURE; goto out; } - if ((new_value == KERN_TFP_POLICY_DENY) - || (new_value == KERN_TFP_POLICY_PERMISSIVE) - || (new_value == KERN_TFP_POLICY_RESTRICTED)) - tfp_policy = new_value; - else - error = EINVAL; + + target = targetproc->task; +#ifndef CONFIG_EMBEDDED + if (target != TASK_NULL) { + mach_port_t tfpport; + + /* If we aren't root and target's task access port is set... */ + if (!kauth_cred_issuser(kauth_cred_get()) && + targetproc != current_proc() && + (task_get_task_access_port(target, &tfpport) == 0) && + (tfpport != IPC_PORT_NULL)) { + + if (tfpport == IPC_PORT_DEAD) { + error = KERN_PROTECTION_FAILURE; + goto out; + } + + /* Call up to the task access server */ + error = check_task_access(tfpport, proc_selfpid(), kauth_getgid(), pid); + + if (error != MACH_MSG_SUCCESS) { + if (error == MACH_RCV_INTERRUPTED) + error = KERN_ABORTED; + else + error = KERN_FAILURE; + goto out; + } + } + } +#endif + + task_reference(target); + error = task_suspend(target); + task_deallocate(target); + out: - return(error); + if (targetproc != PROC_NULL) + proc_rele(targetproc); + *ret = error; + return error; +} + +kern_return_t +pid_resume(struct proc *p __unused, struct pid_resume_args *args, int *ret) +{ + task_t target = NULL; + proc_t targetproc = PROC_NULL; + int pid = args->pid; + int error = 0; + +#if CONFIG_MACF + error = mac_proc_check_suspend_resume(p, 1); /* 1 for resume */ + if (error) { + error = KERN_FAILURE; + goto out; + } +#endif + + if (pid == 0) { + error = KERN_FAILURE; + goto out; + } + + targetproc = proc_find(pid); + if (!task_for_pid_posix_check(targetproc)) { + error = KERN_FAILURE; + goto out; + } + + target = targetproc->task; +#ifndef CONFIG_EMBEDDED + if (target != TASK_NULL) { + mach_port_t tfpport; + + /* If we aren't root and target's task access port is set... */ + if (!kauth_cred_issuser(kauth_cred_get()) && + targetproc != current_proc() && + (task_get_task_access_port(target, &tfpport) == 0) && + (tfpport != IPC_PORT_NULL)) { + + if (tfpport == IPC_PORT_DEAD) { + error = KERN_PROTECTION_FAILURE; + goto out; + } + + /* Call up to the task access server */ + error = check_task_access(tfpport, proc_selfpid(), kauth_getgid(), pid); + + if (error != MACH_MSG_SUCCESS) { + if (error == MACH_RCV_INTERRUPTED) + error = KERN_ABORTED; + else + error = KERN_FAILURE; + goto out; + } + } + } +#endif + + task_reference(target); + error = task_resume(target); + task_deallocate(target); + +out: + if (targetproc != PROC_NULL) + proc_rele(targetproc); + *ret = error; + return error; + return 0; } static int -sysctl_settfp_groups(__unused struct sysctl_oid *oidp, void *arg1, +sysctl_settfp_policy(__unused struct sysctl_oid *oidp, void *arg1, __unused int arg2, struct sysctl_req *req) { int error = 0; @@ -663,266 +857,169 @@ sysctl_settfp_groups(__unused struct sysctl_oid *oidp, void *arg1, if (!is_suser()) return(EPERM); - /* - * Once set; cannot be reset till next boot. Launchd will set this - * in its pid 1 init and no one can set after that. - */ - if (tfp_group_inited != 0) - return(EPERM); - if ((error = SYSCTL_IN(req, &new_value, sizeof(int)))) { goto out; } - - if (new_value >= 100) + if ((new_value == KERN_TFP_POLICY_DENY) + || (new_value == KERN_TFP_POLICY_DEFAULT)) + tfp_policy = new_value; + else error = EINVAL; - else { - if (arg1 == &tfp_group_ronly) - tfp_group_ronly = new_value; - else if (arg1 == &tfp_group_rw) - tfp_group_rw = new_value; - else - error = EINVAL; - if ((tfp_group_ronly != 0 ) && (tfp_group_rw != 0 )) - tfp_group_inited = 1; - } - out: return(error); + } -SYSCTL_NODE(_kern, KERN_TFP, tfp, CTLFLAG_RW, 0, "tfp"); -SYSCTL_PROC(_kern_tfp, KERN_TFP_POLICY, policy, CTLTYPE_INT | CTLFLAG_RW, - &tfp_policy, sizeof(uint32_t), &sysctl_settfp_policy ,"I","policy"); -SYSCTL_PROC(_kern_tfp, KERN_TFP_READ_GROUP, read_group, CTLTYPE_INT | CTLFLAG_RW, - &tfp_group_ronly, sizeof(uint32_t), &sysctl_settfp_groups ,"I","read_group"); -SYSCTL_PROC(_kern_tfp, KERN_TFP_RW_GROUP, rw_group, CTLTYPE_INT | CTLFLAG_RW, - &tfp_group_rw, sizeof(uint32_t), &sysctl_settfp_groups ,"I","rw_group"); +#if defined(SECURE_KERNEL) +static int kern_secure_kernel = 1; +#else +static int kern_secure_kernel = 0; +#endif +SYSCTL_INT(_kern, OID_AUTO, secure_kernel, CTLFLAG_RD, &kern_secure_kernel, 0, ""); -SYSCTL_INT(_vm, OID_AUTO, shared_region_trace_level, CTLFLAG_RW, &shared_region_trace_level, 0, ""); +SYSCTL_NODE(_kern, KERN_TFP, tfp, CTLFLAG_RW|CTLFLAG_LOCKED, 0, "tfp"); +SYSCTL_PROC(_kern_tfp, KERN_TFP_POLICY, policy, CTLTYPE_INT | CTLFLAG_RW, + &tfp_policy, sizeof(uint32_t), &sysctl_settfp_policy ,"I","policy"); -/* - * Try and cap the number of mappings the user might be trying to deal with, - * so that we don't end up allocating insane amounts of wired memory in the - * kernel based on bogus user arguments. - * There are 2 shared regions (TEXT and DATA). The size of each submap - * is SHARED_TEXT_REGION_SIZE and we can have at most 1 VM map entry per page, - * so the maximum number of mappings we could ever have to deal with is... - */ -#define SHARED_REGION_MAX_MAPPINGS ((2 *SHARED_TEXT_REGION_SIZE) >> PAGE_SHIFT) +SYSCTL_INT(_vm, OID_AUTO, shared_region_trace_level, CTLFLAG_RW, + &shared_region_trace_level, 0, ""); +SYSCTL_INT(_vm, OID_AUTO, shared_region_version, CTLFLAG_RD, + &shared_region_version, 0, ""); +SYSCTL_INT(_vm, OID_AUTO, shared_region_persistence, CTLFLAG_RW, + &shared_region_persistence, 0, ""); /* - * shared_region_make_private_np: + * shared_region_check_np: * - * This system call is for "dyld" only. - * - * It creates a private copy of the current process's "shared region" for - * split libraries. "dyld" uses this when the shared region is full or - * it needs to load a split library that conflicts with an already loaded one - * that this process doesn't need. "dyld" specifies a set of address ranges - * that it wants to keep in the now-private "shared region". These cover - * the set of split libraries that the process needs so far. The kernel needs - * to deallocate the rest of the shared region, so that it's available for - * more libraries for this process. + * This system call is intended for dyld. + * + * dyld calls this when any process starts to see if the process's shared + * region is already set up and ready to use. + * This call returns the base address of the first mapping in the + * process's shared region's first mapping. + * dyld will then check what's mapped at that address. + * + * If the shared region is empty, dyld will then attempt to map the shared + * cache file in the shared region via the shared_region_map_np() system call. + * + * If something's already mapped in the shared region, dyld will check if it + * matches the shared cache it would like to use for that process. + * If it matches, evrything's ready and the process can proceed and use the + * shared region. + * If it doesn't match, dyld will unmap the shared region and map the shared + * cache into the process's address space via mmap(). + * + * ERROR VALUES + * EINVAL no shared region + * ENOMEM shared region is empty + * EFAULT bad address for "start_address" */ int -shared_region_make_private_np( - struct proc *p, - struct shared_region_make_private_np_args *uap, - __unused int *retvalp) +shared_region_check_np( + __unused struct proc *p, + struct shared_region_check_np_args *uap, + __unused int *retvalp) { - int error; - kern_return_t kr; - boolean_t using_shared_regions; - user_addr_t user_ranges; - unsigned int range_count; - vm_size_t ranges_size; - struct shared_region_range_np *ranges; - shared_region_mapping_t shared_region; - struct shared_region_task_mappings task_mapping_info; - shared_region_mapping_t next; - - ranges = NULL; - - range_count = uap->rangeCount; - user_ranges = uap->ranges; - ranges_size = (vm_size_t) (range_count * sizeof (ranges[0])); - - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_INFO, - ("shared_region: %p [%d(%s)] " - "make_private(rangecount=%d)\n", - current_thread(), p->p_pid, p->p_comm, range_count)); - - /* allocate kernel space for the "ranges" */ - if (range_count != 0) { - if (range_count > SHARED_REGION_MAX_MAPPINGS) { - error = EINVAL; - goto done; - } - if ((mach_vm_size_t) ranges_size != - (mach_vm_size_t) range_count * sizeof (ranges[0])) { - /* 32-bit integer overflow */ - error = EINVAL; - goto done; - } - kr = kmem_alloc(kernel_map, - (vm_offset_t *) &ranges, - ranges_size); + vm_shared_region_t shared_region; + mach_vm_offset_t start_address; + int error; + kern_return_t kr; + + SHARED_REGION_TRACE_DEBUG( + ("shared_region: %p [%d(%s)] -> check_np(0x%llx)\n", + current_thread(), p->p_pid, p->p_comm, + (uint64_t)uap->start_address)); + + /* retrieve the current tasks's shared region */ + shared_region = vm_shared_region_get(current_task()); + if (shared_region != NULL) { + /* retrieve address of its first mapping... */ + kr = vm_shared_region_start_address(shared_region, + &start_address); if (kr != KERN_SUCCESS) { error = ENOMEM; - goto done; - } - - /* copy "ranges" from user-space */ - error = copyin(user_ranges, - ranges, - ranges_size); - if (error) { - goto done; + } else { + /* ... and give it to the caller */ + error = copyout(&start_address, + (user_addr_t) uap->start_address, + sizeof (start_address)); + if (error) { + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] " + "check_np(0x%llx) " + "copyout(0x%llx) error %d\n", + current_thread(), p->p_pid, p->p_comm, + (uint64_t)uap->start_address, (uint64_t)start_address, + error)); + } } - } - - if (p->p_flag & P_NOSHLIB) { - /* no split library has been mapped for this process so far */ - using_shared_regions = FALSE; + vm_shared_region_deallocate(shared_region); } else { - /* this process has already mapped some split libraries */ - using_shared_regions = TRUE; - } - - /* - * Get a private copy of the current shared region. - * Do not chain it to the system-wide shared region, as we'll want - * to map other split libraries in place of the old ones. We want - * to completely detach from the system-wide shared region and go our - * own way after this point, not sharing anything with other processes. - */ - error = clone_system_shared_regions(using_shared_regions, - FALSE, /* chain_regions */ - ENV_DEFAULT_ROOT); - if (error) { - goto done; - } - - /* get info on the newly allocated shared region */ - vm_get_shared_region(current_task(), &shared_region); - task_mapping_info.self = (vm_offset_t) shared_region; - shared_region_mapping_info(shared_region, - &(task_mapping_info.text_region), - &(task_mapping_info.text_size), - &(task_mapping_info.data_region), - &(task_mapping_info.data_size), - &(task_mapping_info.region_mappings), - &(task_mapping_info.client_base), - &(task_mapping_info.alternate_base), - &(task_mapping_info.alternate_next), - &(task_mapping_info.fs_base), - &(task_mapping_info.system), - &(task_mapping_info.flags), - &next); - - /* - * We now have our private copy of the shared region, as it was before - * the call to clone_system_shared_regions(). We now need to clean it - * up and keep only the memory areas described by the "ranges" array. - */ - kr = shared_region_cleanup(range_count, ranges, &task_mapping_info); - switch (kr) { - case KERN_SUCCESS: - error = 0; - break; - default: + /* no shared region ! */ error = EINVAL; - goto done; } -done: - if (ranges != NULL) { - kmem_free(kernel_map, - (vm_offset_t) ranges, - ranges_size); - ranges = NULL; - } - - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_INFO, - ("shared_region: %p [%d(%s)] " - "make_private(rangecount=%d) -> %d " - "shared_region=%p[%x,%x,%x]\n", + SHARED_REGION_TRACE_DEBUG( + ("shared_region: %p [%d(%s)] check_np(0x%llx) <- 0x%llx %d\n", current_thread(), p->p_pid, p->p_comm, - range_count, error, shared_region, - task_mapping_info.fs_base, - task_mapping_info.system, - task_mapping_info.flags)); + (uint64_t)uap->start_address, (uint64_t)start_address, error)); return error; } - /* - * shared_region_map_file_np: - * - * This system call is for "dyld" only. + * shared_region_map_np() * - * "dyld" wants to map parts of a split library in the shared region. - * We get a file descriptor on the split library to be mapped and a set - * of mapping instructions, describing which parts of the file to map in\ - * which areas of the shared segment and with what protection. - * The "shared region" is split in 2 areas: - * 0x90000000 - 0xa0000000 : read-only area (for TEXT and LINKEDIT sections), - * 0xa0000000 - 0xb0000000 : writable area (for DATA sections). + * This system call is intended for dyld. * + * dyld uses this to map a shared cache file into a shared region. + * This is usually done only the first time a shared cache is needed. + * Subsequent processes will just use the populated shared region without + * requiring any further setup. */ int -shared_region_map_file_np( +shared_region_map_np( struct proc *p, - struct shared_region_map_file_np_args *uap, + struct shared_region_map_np_args *uap, __unused int *retvalp) { - int error; - kern_return_t kr; - int fd; - unsigned int mapping_count; - user_addr_t user_mappings; /* 64-bit */ - user_addr_t user_slide_p; /* 64-bit */ - struct shared_file_mapping_np *mappings; - vm_size_t mappings_size; - struct fileproc *fp; - mach_vm_offset_t slide; - struct vnode *vp; - struct vfs_context context; - memory_object_control_t file_control; - memory_object_size_t file_size; - shared_region_mapping_t shared_region; - struct shared_region_task_mappings task_mapping_info; - shared_region_mapping_t next; - shared_region_mapping_t default_shared_region; - boolean_t using_default_region; - unsigned int j; - vm_prot_t max_prot; - mach_vm_offset_t base_offset, end_offset; - mach_vm_offset_t original_base_offset; - boolean_t mappings_in_segment; -#define SFM_MAX_STACK 6 - struct shared_file_mapping_np stack_mappings[SFM_MAX_STACK]; - + int error; + kern_return_t kr; + int fd; + struct fileproc *fp; + struct vnode *vp, *root_vp; + struct vnode_attr va; + off_t fs; + memory_object_size_t file_size; + user_addr_t user_mappings; + struct shared_file_mapping_np *mappings; +#define SFM_MAX_STACK 8 + struct shared_file_mapping_np stack_mappings[SFM_MAX_STACK]; + unsigned int mappings_count; + vm_size_t mappings_size; + memory_object_control_t file_control; + struct vm_shared_region *shared_region; + + SHARED_REGION_TRACE_DEBUG( + ("shared_region: %p [%d(%s)] -> map\n", + current_thread(), p->p_pid, p->p_comm)); + + shared_region = NULL; + mappings_count = 0; mappings_size = 0; mappings = NULL; - mapping_count = 0; fp = NULL; vp = NULL; - /* get file descriptor for split library from arguments */ + /* get file descriptor for shared region cache file */ fd = uap->fd; /* get file structure from file descriptor */ error = fp_lookup(p, fd, &fp, 0); if (error) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_ERROR, - ("shared_region: %p [%d(%s)] map_file: " + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map: " "fd=%d lookup failed (error=%d)\n", current_thread(), p->p_pid, p->p_comm, fd, error)); goto done; @@ -930,9 +1027,8 @@ shared_region_map_file_np( /* make sure we're attempting to map a vnode */ if (fp->f_fglob->fg_type != DTYPE_VNODE) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_ERROR, - ("shared_region: %p [%d(%s)] map_file: " + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map: " "fd=%d not a vnode (type=%d)\n", current_thread(), p->p_pid, p->p_comm, fd, fp->f_fglob->fg_type)); @@ -942,9 +1038,8 @@ shared_region_map_file_np( /* we need at least read permission on the file */ if (! (fp->f_fglob->fg_flag & FREAD)) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_ERROR, - ("shared_region: %p [%d(%s)] map_file: " + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map: " "fd=%d not readable\n", current_thread(), p->p_pid, p->p_comm, fd)); error = EPERM; @@ -952,11 +1047,10 @@ shared_region_map_file_np( } /* get vnode from file structure */ - error = vnode_getwithref((vnode_t)fp->f_fglob->fg_data); + error = vnode_getwithref((vnode_t) fp->f_fglob->fg_data); if (error) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_ERROR, - ("shared_region: %p [%d(%s)] map_file: " + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map: " "fd=%d getwithref failed (error=%d)\n", current_thread(), p->p_pid, p->p_comm, fd, error)); goto done; @@ -965,9 +1059,8 @@ shared_region_map_file_np( /* make sure the vnode is a regular file */ if (vp->v_type != VREG) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_ERROR, - ("shared_region: %p [%d(%s)] map_file(%p:'%s'): " + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " "not a file (type=%d)\n", current_thread(), p->p_pid, p->p_comm, vp, vp->v_name, vp->v_type)); @@ -975,342 +1068,159 @@ shared_region_map_file_np( goto done; } - /* get vnode size */ - { - off_t fs; - - context.vc_proc = p; - context.vc_ucred = kauth_cred_get(); - if ((error = vnode_size(vp, &fs, &context)) != 0) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_ERROR, - ("shared_region: %p [%d(%s)] " - "map_file(%p:'%s'): " - "vnode_size(%p) failed (error=%d)\n", - current_thread(), p->p_pid, p->p_comm, - vp, vp->v_name, vp)); - goto done; - } - file_size = fs; + /* make sure vnode is on the process's root volume */ + root_vp = p->p_fd->fd_rdir; + if (root_vp == NULL) { + root_vp = rootvnode; } - - /* - * Get the list of mappings the caller wants us to establish. - */ - mapping_count = uap->mappingCount; /* the number of mappings */ - mappings_size = (vm_size_t) (mapping_count * sizeof (mappings[0])); - if (mapping_count == 0) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_INFO, - ("shared_region: %p [%d(%s)] map_file(%p:'%s'): " - "no mappings\n", + if (vp->v_mount != root_vp->v_mount) { + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " + "not on process's root volume\n", current_thread(), p->p_pid, p->p_comm, vp, vp->v_name)); - error = 0; /* no mappings: we're done ! */ + error = EPERM; goto done; - } else if (mapping_count <= SFM_MAX_STACK) { - mappings = &stack_mappings[0]; - } else { - if (mapping_count > SHARED_REGION_MAX_MAPPINGS) { - error = EINVAL; - goto done; - } - if ((mach_vm_size_t) mappings_size != - (mach_vm_size_t) mapping_count * sizeof (mappings[0])) { - /* 32-bit integer overflow */ - error = EINVAL; - goto done; - } - kr = kmem_alloc(kernel_map, - (vm_offset_t *) &mappings, - mappings_size); - if (kr != KERN_SUCCESS) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_ERROR, - ("shared_region: %p [%d(%s)] " - "map_file(%p:'%s'): " - "failed to allocate %d mappings (kr=0x%x)\n", - current_thread(), p->p_pid, p->p_comm, - vp, vp->v_name, mapping_count, kr)); - error = ENOMEM; - goto done; - } } - user_mappings = uap->mappings; /* the mappings, in user space */ - error = copyin(user_mappings, - mappings, - mappings_size); - if (error != 0) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_ERROR, - ("shared_region: %p [%d(%s)] map_file(%p:'%s'): " - "failed to copyin %d mappings (error=%d)\n", + /* make sure vnode is owned by "root" */ + VATTR_INIT(&va); + VATTR_WANTED(&va, va_uid); + error = vnode_getattr(vp, &va, vfs_context_current()); + if (error) { + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " + "vnode_getattr(%p) failed (error=%d)\n", current_thread(), p->p_pid, p->p_comm, - vp, vp->v_name, mapping_count, error)); + vp, vp->v_name, vp, error)); goto done; } - - /* - * If the caller provides a "slide" pointer, it means they're OK - * with us moving the mappings around to make them fit. - */ - user_slide_p = uap->slide_p; - - /* - * Make each mapping address relative to the beginning of the - * shared region. Check that all mappings are in the shared region. - * Compute the maximum set of protections required to tell the - * buffer cache how we mapped the file (see call to ubc_map() below). - */ - max_prot = VM_PROT_NONE; - base_offset = -1LL; - end_offset = 0; - mappings_in_segment = TRUE; - for (j = 0; j < mapping_count; j++) { - mach_vm_offset_t segment; - segment = (mappings[j].sfm_address & - GLOBAL_SHARED_SEGMENT_MASK); - if (segment != GLOBAL_SHARED_TEXT_SEGMENT && - segment != GLOBAL_SHARED_DATA_SEGMENT) { - /* this mapping is not in the shared region... */ - if (user_slide_p == NULL) { - /* ... and we can't slide it in: fail */ - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_CONFLICT, - ("shared_region: %p [%d(%s)] " - "map_file(%p:'%s'): " - "mapping %p not in shared segment & " - "no sliding\n", - current_thread(), p->p_pid, p->p_comm, - vp, vp->v_name, - mappings[j].sfm_address)); - error = EINVAL; - goto done; - } - if (j == 0) { - /* expect all mappings to be outside */ - mappings_in_segment = FALSE; - } else if (mappings_in_segment != FALSE) { - /* other mappings were not outside: fail */ - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_CONFLICT, - ("shared_region: %p [%d(%s)] " - "map_file(%p:'%s'): " - "mapping %p not in shared segment & " - "other mappings in shared segment\n", - current_thread(), p->p_pid, p->p_comm, - vp, vp->v_name, - mappings[j].sfm_address)); - error = EINVAL; - goto done; - } - /* we'll try and slide that mapping in the segments */ - } else { - if (j == 0) { - /* expect all mappings to be inside */ - mappings_in_segment = TRUE; - } else if (mappings_in_segment != TRUE) { - /* other mappings were not inside: fail */ - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_CONFLICT, - ("shared_region: %p [%d(%s)] " - "map_file(%p:'%s'): " - "mapping %p in shared segment & " - "others in shared segment\n", - current_thread(), p->p_pid, p->p_comm, - vp, vp->v_name, - mappings[j].sfm_address)); - error = EINVAL; - goto done; - } - /* get a relative offset inside the shared segments */ - mappings[j].sfm_address -= GLOBAL_SHARED_TEXT_SEGMENT; - } - if ((mappings[j].sfm_address & SHARED_TEXT_REGION_MASK) - < base_offset) { - base_offset = (mappings[j].sfm_address & - SHARED_TEXT_REGION_MASK); - } - if ((mappings[j].sfm_address & SHARED_TEXT_REGION_MASK) + - mappings[j].sfm_size > end_offset) { - end_offset = - (mappings[j].sfm_address & - SHARED_TEXT_REGION_MASK) + - mappings[j].sfm_size; - } - max_prot |= mappings[j].sfm_max_prot; - } - /* Make all mappings relative to the base_offset */ - base_offset = vm_map_trunc_page(base_offset); - end_offset = vm_map_round_page(end_offset); - for (j = 0; j < mapping_count; j++) { - mappings[j].sfm_address -= base_offset; + if (va.va_uid != 0) { + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " + "owned by uid=%d instead of 0\n", + current_thread(), p->p_pid, p->p_comm, + vp, vp->v_name, va.va_uid)); + error = EPERM; + goto done; } - original_base_offset = base_offset; - if (mappings_in_segment == FALSE) { - /* - * We're trying to map a library that was not pre-bound to - * be in the shared segments. We want to try and slide it - * back into the shared segments but as far back as possible, - * so that it doesn't clash with pre-bound libraries. Set - * the base_offset to the end of the region, so that it can't - * possibly fit there and will have to be slid. - */ - base_offset = SHARED_TEXT_REGION_SIZE - end_offset; + + /* get vnode size */ + error = vnode_size(vp, &fs, vfs_context_current()); + if (error) { + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " + "vnode_size(%p) failed (error=%d)\n", + current_thread(), p->p_pid, p->p_comm, + vp, vp->v_name, vp, error)); + goto done; } + file_size = fs; /* get the file's memory object handle */ - UBCINFOCHECK("shared_region_map_file_np", vp); file_control = ubc_getobject(vp, UBC_HOLDOBJECT); if (file_control == MEMORY_OBJECT_CONTROL_NULL) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_ERROR, - ("shared_region: %p [%d(%s)] map_file(%p:'%s'): " - "ubc_getobject() failed\n", + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " + "no memory object\n", current_thread(), p->p_pid, p->p_comm, vp, vp->v_name)); error = EINVAL; goto done; } - - /* - * Get info about the current process's shared region. - * This might change if we decide we need to clone the shared region. - */ - vm_get_shared_region(current_task(), &shared_region); - task_mapping_info.self = (vm_offset_t) shared_region; - shared_region_mapping_info(shared_region, - &(task_mapping_info.text_region), - &(task_mapping_info.text_size), - &(task_mapping_info.data_region), - &(task_mapping_info.data_size), - &(task_mapping_info.region_mappings), - &(task_mapping_info.client_base), - &(task_mapping_info.alternate_base), - &(task_mapping_info.alternate_next), - &(task_mapping_info.fs_base), - &(task_mapping_info.system), - &(task_mapping_info.flags), - &next); - - /* - * Are we using the system's current shared region - * for this environment ? - */ - default_shared_region = - lookup_default_shared_region(ENV_DEFAULT_ROOT, - task_mapping_info.system); - if (shared_region == default_shared_region) { - using_default_region = TRUE; + + /* get the list of mappings the caller wants us to establish */ + mappings_count = uap->count; /* number of mappings */ + mappings_size = (vm_size_t) (mappings_count * sizeof (mappings[0])); + if (mappings_count == 0) { + SHARED_REGION_TRACE_INFO( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " + "no mappings\n", + current_thread(), p->p_pid, p->p_comm, + vp, vp->v_name)); + error = 0; /* no mappings: we're done ! */ + goto done; + } else if (mappings_count <= SFM_MAX_STACK) { + mappings = &stack_mappings[0]; } else { - using_default_region = FALSE; + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " + "too many mappings (%d)\n", + current_thread(), p->p_pid, p->p_comm, + vp, vp->v_name, mappings_count)); + error = EINVAL; + goto done; } - shared_region_mapping_dealloc(default_shared_region); - if (vp->v_mount != rootvnode->v_mount && - using_default_region) { - /* - * The split library is not on the root filesystem. We don't - * want to polute the system-wide ("default") shared region - * with it. - * Reject the mapping. The caller (dyld) should "privatize" - * (via shared_region_make_private()) the shared region and - * try to establish the mapping privately for this process. - */ - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_CONFLICT, - ("shared_region: %p [%d(%s)] " - "map_file(%p:'%s'): " - "not on root volume\n", + user_mappings = uap->mappings; /* the mappings, in user space */ + error = copyin(user_mappings, + mappings, + mappings_size); + if (error) { + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " + "copyin(0x%llx, %d) failed (error=%d)\n", current_thread(), p->p_pid, p->p_comm, - vp->v_name)); - error = EXDEV; + vp, vp->v_name, (uint64_t)user_mappings, mappings_count, error)); goto done; } + /* get the process's shared region (setup in vm_map_exec()) */ + shared_region = vm_shared_region_get(current_task()); + if (shared_region == NULL) { + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " + "no shared region\n", + current_thread(), p->p_pid, p->p_comm, + vp, vp->v_name)); + goto done; + } - /* - * Map the split library. - */ - kr = map_shared_file(mapping_count, - mappings, - file_control, - file_size, - &task_mapping_info, - base_offset, - (user_slide_p) ? &slide : NULL); - - if (kr == KERN_SUCCESS) { - /* - * The mapping was successful. Let the buffer cache know - * that we've mapped that file with these protections. This - * prevents the vnode from getting recycled while it's mapped. - */ - (void) ubc_map(vp, max_prot); - error = 0; - } else { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_CONFLICT, - ("shared_region: %p [%d(%s)] " - "map_file(%p:'%s'): " - "map_shared_file failed, kr=0x%x\n", + /* map the file into that shared region's submap */ + kr = vm_shared_region_map_file(shared_region, + mappings_count, + mappings, + file_control, + file_size, + (void *) p->p_fd->fd_rdir); + if (kr != KERN_SUCCESS) { + SHARED_REGION_TRACE_ERROR( + ("shared_region: %p [%d(%s)] map(%p:'%s'): " + "vm_shared_region_map_file() failed kr=0x%x\n", current_thread(), p->p_pid, p->p_comm, vp, vp->v_name, kr)); switch (kr) { case KERN_INVALID_ADDRESS: error = EFAULT; - goto done; + break; case KERN_PROTECTION_FAILURE: error = EPERM; - goto done; + break; case KERN_NO_SPACE: error = ENOMEM; - goto done; + break; case KERN_FAILURE: case KERN_INVALID_ARGUMENT: default: error = EINVAL; - goto done; + break; } + goto done; } - if (p->p_flag & P_NOSHLIB) { - /* signal that this process is now using split libraries */ - p->p_flag &= ~P_NOSHLIB; + error = 0; + + /* update the vnode's access time */ + if (! (vnode_vfsvisflags(vp) & MNT_NOATIME)) { + VATTR_INIT(&va); + nanotime(&va.va_access_time); + VATTR_SET_ACTIVE(&va, va_access_time); + vnode_setattr(vp, &va, vfs_context_current()); } - if (user_slide_p) { - /* - * The caller provided a pointer to a "slide" offset. Let - * them know by how much we slid the mappings. - */ - if (mappings_in_segment == FALSE) { - /* - * We faked the base_offset earlier, so undo that - * and take into account the real base_offset. - */ - slide += SHARED_TEXT_REGION_SIZE - end_offset; - slide -= original_base_offset; - /* - * The mappings were slid into the shared segments - * and "slide" is relative to the beginning of the - * shared segments. Adjust it to be absolute. - */ - slide += GLOBAL_SHARED_TEXT_SEGMENT; - } - error = copyout(&slide, - user_slide_p, - sizeof (slide)); - if (slide != 0) { - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_CONFLICT, - ("shared_region: %p [%d(%s)] " - "map_file(%p:'%s'): " - "slid by 0x%llx\n", - current_thread(), p->p_pid, p->p_comm, - vp, vp->v_name, slide)); - } + if (p->p_flag & P_NOSHLIB) { + /* signal that this process is now using split libraries */ + OSBitAndAtomic(~((uint32_t)P_NOSHLIB), &p->p_flag); } done: @@ -1327,1175 +1237,112 @@ done: fp_drop(p, fd, fp, 0); fp = NULL; } - if (mappings != NULL && - mappings != &stack_mappings[0]) { - kmem_free(kernel_map, - (vm_offset_t) mappings, - mappings_size); - } - mappings = NULL; - - return error; -} - -int -load_shared_file( - __unused struct proc *p, - __unused struct load_shared_file_args *uap, - __unused int *retval) -{ - return ENOSYS; -} - -int -reset_shared_file( - __unused struct proc *p, - __unused struct reset_shared_file_args *uap, - __unused int *retval) -{ - return ENOSYS; -} - -int -new_system_shared_regions( - __unused struct proc *p, - __unused struct new_system_shared_regions_args *uap, - __unused int *retval) -{ - return ENOSYS; -} - - -int -clone_system_shared_regions( - int shared_regions_active, - int chain_regions, - int base_vnode) -{ - shared_region_mapping_t new_shared_region; - shared_region_mapping_t next; - shared_region_mapping_t old_shared_region; - struct shared_region_task_mappings old_info; - struct shared_region_task_mappings new_info; - - vm_get_shared_region(current_task(), &old_shared_region); - old_info.self = (vm_offset_t)old_shared_region; - shared_region_mapping_info(old_shared_region, - &(old_info.text_region), - &(old_info.text_size), - &(old_info.data_region), - &(old_info.data_size), - &(old_info.region_mappings), - &(old_info.client_base), - &(old_info.alternate_base), - &(old_info.alternate_next), - &(old_info.fs_base), - &(old_info.system), - &(old_info.flags), &next); - - if (shared_regions_active || - base_vnode == ENV_DEFAULT_ROOT) { - if (shared_file_create_system_region(&new_shared_region, - old_info.fs_base, - old_info.system)) - return ENOMEM; - } else { - if (old_shared_region && - base_vnode == ENV_DEFAULT_ROOT) { - base_vnode = old_info.fs_base; - } - new_shared_region = - lookup_default_shared_region(base_vnode, - old_info.system); - if (new_shared_region == NULL) { - shared_file_boot_time_init(base_vnode, - old_info.system); - vm_get_shared_region(current_task(), - &new_shared_region); - } else { - vm_set_shared_region(current_task(), new_shared_region); - } - if (old_shared_region) - shared_region_mapping_dealloc(old_shared_region); + if (shared_region != NULL) { + vm_shared_region_deallocate(shared_region); } - new_info.self = (vm_offset_t)new_shared_region; - shared_region_mapping_info(new_shared_region, - &(new_info.text_region), - &(new_info.text_size), - &(new_info.data_region), - &(new_info.data_size), - &(new_info.region_mappings), - &(new_info.client_base), - &(new_info.alternate_base), - &(new_info.alternate_next), - &(new_info.fs_base), - &(new_info.system), - &(new_info.flags), &next); - if(shared_regions_active) { - if(vm_region_clone(old_info.text_region, new_info.text_region)) { - panic("clone_system_shared_regions: shared region mis-alignment 1"); - shared_region_mapping_dealloc(new_shared_region); - return(EINVAL); - } - if (vm_region_clone(old_info.data_region, new_info.data_region)) { - panic("clone_system_shared_regions: shared region mis-alignment 2"); - shared_region_mapping_dealloc(new_shared_region); - return(EINVAL); - } - if (chain_regions) { - /* - * We want a "shadowed" clone, a private superset of the old - * shared region. The info about the old mappings is still - * valid for us. - */ - shared_region_object_chain_attach( - new_shared_region, old_shared_region); - } - } - if (!chain_regions) { - /* - * We want a completely detached clone with no link to - * the old shared region. We'll be removing some mappings - * in our private, cloned, shared region, so the old mappings - * will become irrelevant to us. Since we have a private - * "shared region" now, it isn't going to be shared with - * anyone else and we won't need to maintain mappings info. - */ - shared_region_object_chain_detached(new_shared_region); - } - if (vm_map_region_replace(current_map(), old_info.text_region, - new_info.text_region, old_info.client_base, - old_info.client_base+old_info.text_size)) { - panic("clone_system_shared_regions: shared region mis-alignment 3"); - shared_region_mapping_dealloc(new_shared_region); - return(EINVAL); - } - if(vm_map_region_replace(current_map(), old_info.data_region, - new_info.data_region, - old_info.client_base + old_info.text_size, - old_info.client_base - + old_info.text_size + old_info.data_size)) { - panic("clone_system_shared_regions: shared region mis-alignment 4"); - shared_region_mapping_dealloc(new_shared_region); - return(EINVAL); - } - vm_set_shared_region(current_task(), new_shared_region); - - /* consume the reference which wasn't accounted for in object */ - /* chain attach */ - if (!shared_regions_active || !chain_regions) - shared_region_mapping_dealloc(old_shared_region); - - SHARED_REGION_TRACE( - SHARED_REGION_TRACE_INFO, - ("shared_region: %p task=%p " - "clone(active=%d, base=0x%x,chain=%d) " - "old=%p[%x,%x,%x] new=%p[%x,%x,%x]\n", - current_thread(), current_task(), - shared_regions_active, base_vnode, chain_regions, - old_shared_region, - old_info.fs_base, - old_info.system, - old_info.flags, - new_shared_region, - new_info.fs_base, - new_info.system, - new_info.flags)); - - return(0); - -} - -/* header for the profile name file. The profiled app info is held */ -/* in the data file and pointed to by elements in the name file */ - -struct profile_names_header { - unsigned int number_of_profiles; - unsigned int user_id; - unsigned int version; - off_t element_array; - unsigned int spare1; - unsigned int spare2; - unsigned int spare3; -}; - -struct profile_element { - off_t addr; - vm_size_t size; - unsigned int mod_date; - unsigned int inode; - char name[12]; -}; - -struct global_profile { - struct vnode *names_vp; - struct vnode *data_vp; - vm_offset_t buf_ptr; - unsigned int user; - unsigned int age; - unsigned int busy; -}; -struct global_profile_cache { - int max_ele; - unsigned int age; - struct global_profile profiles[3]; -}; - -/* forward declarations */ -int bsd_open_page_cache_files(unsigned int user, - struct global_profile **profile); -void bsd_close_page_cache_files(struct global_profile *profile); -int bsd_search_page_cache_data_base( - struct vnode *vp, - struct profile_names_header *database, - char *app_name, - unsigned int mod_date, - unsigned int inode, - off_t *profile, - unsigned int *profile_size); - -struct global_profile_cache global_user_profile_cache = - {3, 0, {{NULL, NULL, 0, 0, 0, 0}, - {NULL, NULL, 0, 0, 0, 0}, - {NULL, NULL, 0, 0, 0, 0}} }; - -/* BSD_OPEN_PAGE_CACHE_FILES: */ -/* Caller provides a user id. This id was used in */ -/* prepare_profile_database to create two unique absolute */ -/* file paths to the associated profile files. These files */ -/* are either opened or bsd_open_page_cache_files returns an */ -/* error. The header of the names file is then consulted. */ -/* The header and the vnodes for the names and data files are */ -/* returned. */ + SHARED_REGION_TRACE_DEBUG( + ("shared_region: %p [%d(%s)] <- map\n", + current_thread(), p->p_pid, p->p_comm)); -int -bsd_open_page_cache_files( - unsigned int user, - struct global_profile **profile) -{ - const char *cache_path = "/var/vm/app_profile/"; - struct proc *p; - int error; - vm_size_t resid; - off_t resid_off; - unsigned int lru; - vm_size_t size; - - struct vnode *names_vp; - struct vnode *data_vp; - vm_offset_t names_buf; - vm_offset_t buf_ptr; - - int profile_names_length; - int profile_data_length; - char *profile_data_string; - char *profile_names_string; - char *substring; - - off_t file_size; - struct vfs_context context; - - kern_return_t ret; - - struct nameidata nd_names; - struct nameidata nd_data; - int i; - - - p = current_proc(); - - context.vc_proc = p; - context.vc_ucred = kauth_cred_get(); - -restart: - for(i = 0; ibusy) { - /* - * drop funnel and wait - */ - (void)tsleep((void *) - *profile, - PRIBIO, "app_profile", 0); - goto restart; - } - (*profile)->busy = 1; - (*profile)->age = global_user_profile_cache.age; - - /* - * entries in cache are held with a valid - * usecount... take an iocount which will - * be dropped in "bsd_close_page_cache_files" - * which is called after the read or writes to - * these files are done - */ - if ( (vnode_getwithref((*profile)->data_vp)) ) { - - vnode_rele((*profile)->data_vp); - vnode_rele((*profile)->names_vp); - - (*profile)->data_vp = NULL; - (*profile)->busy = 0; - wakeup(*profile); - - goto restart; - } - if ( (vnode_getwithref((*profile)->names_vp)) ) { - - vnode_put((*profile)->data_vp); - vnode_rele((*profile)->data_vp); - vnode_rele((*profile)->names_vp); - - (*profile)->data_vp = NULL; - (*profile)->busy = 0; - wakeup(*profile); - - goto restart; - } - global_user_profile_cache.age+=1; - return 0; - } - } - - lru = global_user_profile_cache.age; - *profile = NULL; - for(i = 0; iage = global_user_profile_cache.age; - break; - } - /* Otherwise grab the oldest entry */ - if(global_user_profile_cache.profiles[i].age < lru) { - lru = global_user_profile_cache.profiles[i].age; - *profile = &global_user_profile_cache.profiles[i]; - } - } - - /* Did we set it? */ - if (*profile == NULL) { - /* - * No entries are available; this can only happen if all - * of them are currently in the process of being reused; - * if this happens, we sleep on the address of the first - * element, and restart. This is less than ideal, but we - * know it will work because we know that there will be a - * wakeup on any entry currently in the process of being - * reused. - * - * XXX Reccomend a two handed clock and more than 3 total - * XXX cache entries at some point in the future. - */ - /* - * drop funnel and wait - */ - (void)tsleep((void *) - &global_user_profile_cache.profiles[0], - PRIBIO, "app_profile", 0); - goto restart; - } - - /* - * If it's currently busy, we've picked the one at the end of the - * LRU list, but it's currently being actively used. We sleep on - * its address and restart. - */ - if ((*profile)->busy) { - /* - * drop funnel and wait - */ - (void)tsleep((void *) - *profile, - PRIBIO, "app_profile", 0); - goto restart; - } - (*profile)->busy = 1; - (*profile)->user = user; - - /* - * put dummy value in for now to get competing request to wait - * above until we are finished - * - * Save the data_vp before setting it, so we can set it before - * we kmem_free() or vrele(). If we don't do this, then we - * have a potential funnel race condition we have to deal with. - */ - data_vp = (*profile)->data_vp; - (*profile)->data_vp = (struct vnode *)0xFFFFFFFF; - - /* - * Age the cache here in all cases; this guarantees that we won't - * be reusing only one entry over and over, once the system reaches - * steady-state. - */ - global_user_profile_cache.age+=1; - - if(data_vp != NULL) { - kmem_free(kernel_map, - (*profile)->buf_ptr, 4 * PAGE_SIZE); - if ((*profile)->names_vp) { - vnode_rele((*profile)->names_vp); - (*profile)->names_vp = NULL; - } - vnode_rele(data_vp); - } - - /* Try to open the appropriate users profile files */ - /* If neither file is present, try to create them */ - /* If one file is present and the other not, fail. */ - /* If the files do exist, check them for the app_file */ - /* requested and read it in if present */ - - ret = kmem_alloc(kernel_map, - (vm_offset_t *)&profile_data_string, PATH_MAX); - - if(ret) { - (*profile)->data_vp = NULL; - (*profile)->busy = 0; - wakeup(*profile); - return ENOMEM; - } - - /* Split the buffer in half since we know the size of */ - /* our file path and our allocation is adequate for */ - /* both file path names */ - profile_names_string = profile_data_string + (PATH_MAX/2); - - - strcpy(profile_data_string, cache_path); - strcpy(profile_names_string, cache_path); - profile_names_length = profile_data_length - = strlen(profile_data_string); - substring = profile_data_string + profile_data_length; - sprintf(substring, "%x_data", user); - substring = profile_names_string + profile_names_length; - sprintf(substring, "%x_names", user); - - /* We now have the absolute file names */ - - ret = kmem_alloc(kernel_map, - (vm_offset_t *)&names_buf, 4 * PAGE_SIZE); - if(ret) { - kmem_free(kernel_map, - (vm_offset_t)profile_data_string, PATH_MAX); - (*profile)->data_vp = NULL; - (*profile)->busy = 0; - wakeup(*profile); - return ENOMEM; - } - - NDINIT(&nd_names, LOOKUP, FOLLOW | LOCKLEAF, - UIO_SYSSPACE32, CAST_USER_ADDR_T(profile_names_string), &context); - NDINIT(&nd_data, LOOKUP, FOLLOW | LOCKLEAF, - UIO_SYSSPACE32, CAST_USER_ADDR_T(profile_data_string), &context); - - if ( (error = vn_open(&nd_data, FREAD | FWRITE, 0)) ) { -#ifdef notdef - printf("bsd_open_page_cache_files: CacheData file not found %s\n", - profile_data_string); -#endif - kmem_free(kernel_map, - (vm_offset_t)names_buf, 4 * PAGE_SIZE); - kmem_free(kernel_map, - (vm_offset_t)profile_data_string, PATH_MAX); - (*profile)->data_vp = NULL; - (*profile)->busy = 0; - wakeup(*profile); - return error; - } - data_vp = nd_data.ni_vp; - - if ( (error = vn_open(&nd_names, FREAD | FWRITE, 0)) ) { - printf("bsd_open_page_cache_files: NamesData file not found %s\n", - profile_data_string); - kmem_free(kernel_map, - (vm_offset_t)names_buf, 4 * PAGE_SIZE); - kmem_free(kernel_map, - (vm_offset_t)profile_data_string, PATH_MAX); - - vnode_rele(data_vp); - vnode_put(data_vp); - - (*profile)->data_vp = NULL; - (*profile)->busy = 0; - wakeup(*profile); - return error; - } - names_vp = nd_names.ni_vp; - - if ((error = vnode_size(names_vp, &file_size, &context)) != 0) { - printf("bsd_open_page_cache_files: Can't stat name file %s\n", profile_names_string); - kmem_free(kernel_map, - (vm_offset_t)profile_data_string, PATH_MAX); - kmem_free(kernel_map, - (vm_offset_t)names_buf, 4 * PAGE_SIZE); - - vnode_rele(names_vp); - vnode_put(names_vp); - vnode_rele(data_vp); - vnode_put(data_vp); - - (*profile)->data_vp = NULL; - (*profile)->busy = 0; - wakeup(*profile); - return error; - } - - size = file_size; - if(size > 4 * PAGE_SIZE) - size = 4 * PAGE_SIZE; - buf_ptr = names_buf; - resid_off = 0; - - while(size) { - int resid_int; - error = vn_rdwr(UIO_READ, names_vp, (caddr_t)buf_ptr, - size, resid_off, - UIO_SYSSPACE32, IO_NODELOCKED, kauth_cred_get(), - &resid_int, p); - resid = (vm_size_t) resid_int; - if((error) || (size == resid)) { - if(!error) { - error = EINVAL; - } - kmem_free(kernel_map, - (vm_offset_t)profile_data_string, PATH_MAX); - kmem_free(kernel_map, - (vm_offset_t)names_buf, 4 * PAGE_SIZE); - - vnode_rele(names_vp); - vnode_put(names_vp); - vnode_rele(data_vp); - vnode_put(data_vp); - - (*profile)->data_vp = NULL; - (*profile)->busy = 0; - wakeup(*profile); - return error; - } - buf_ptr += size-resid; - resid_off += size-resid; - size = resid; - } - kmem_free(kernel_map, (vm_offset_t)profile_data_string, PATH_MAX); - - (*profile)->names_vp = names_vp; - (*profile)->data_vp = data_vp; - (*profile)->buf_ptr = names_buf; - - /* - * at this point, the both the names_vp and the data_vp have - * both a valid usecount and an iocount held - */ - return 0; - -} - -void -bsd_close_page_cache_files( - struct global_profile *profile) -{ - vnode_put(profile->data_vp); - vnode_put(profile->names_vp); - - profile->busy = 0; - wakeup(profile); + return error; } -int -bsd_read_page_cache_file( - unsigned int user, - int *fid, - int *mod, - char *app_name, - struct vnode *app_vp, - vm_offset_t *buffer, - vm_offset_t *bufsize) -{ - - boolean_t funnel_state; - - struct proc *p; - int error; - unsigned int resid; - - off_t profile; - unsigned int profile_size; - - vm_offset_t names_buf; - struct vnode_attr va; - struct vfs_context context; - - kern_return_t ret; - - struct vnode *names_vp; - struct vnode *data_vp; - - struct global_profile *uid_files; - - funnel_state = thread_funnel_set(kernel_flock, TRUE); - - /* Try to open the appropriate users profile files */ - /* If neither file is present, try to create them */ - /* If one file is present and the other not, fail. */ - /* If the files do exist, check them for the app_file */ - /* requested and read it in if present */ - - - error = bsd_open_page_cache_files(user, &uid_files); - if(error) { - thread_funnel_set(kernel_flock, funnel_state); - return EINVAL; - } - - p = current_proc(); - - names_vp = uid_files->names_vp; - data_vp = uid_files->data_vp; - names_buf = uid_files->buf_ptr; - context.vc_proc = p; - context.vc_ucred = kauth_cred_get(); +/* sysctl overflow room */ - VATTR_INIT(&va); - VATTR_WANTED(&va, va_fileid); - VATTR_WANTED(&va, va_modify_time); - - if ((error = vnode_getattr(app_vp, &va, &context))) { - printf("bsd_read_cache_file: Can't stat app file %s\n", app_name); - bsd_close_page_cache_files(uid_files); - thread_funnel_set(kernel_flock, funnel_state); - return error; - } +/* vm_page_free_target is provided as a makeshift solution for applications that want to + allocate buffer space, possibly purgeable memory, but not cause inactive pages to be + reclaimed. It allows the app to calculate how much memory is free outside the free target. */ +extern unsigned int vm_page_free_target; +SYSCTL_INT(_vm, OID_AUTO, vm_page_free_target, CTLFLAG_RD, + &vm_page_free_target, 0, "Pageout daemon free target"); - *fid = (u_long)va.va_fileid; - *mod = va.va_modify_time.tv_sec; - - if (bsd_search_page_cache_data_base( - names_vp, - (struct profile_names_header *)names_buf, - app_name, - (unsigned int) va.va_modify_time.tv_sec, - (u_long)va.va_fileid, &profile, &profile_size) == 0) { - /* profile is an offset in the profile data base */ - /* It is zero if no profile data was found */ - - if(profile_size == 0) { - *buffer = 0; - *bufsize = 0; - bsd_close_page_cache_files(uid_files); - thread_funnel_set(kernel_flock, funnel_state); - return 0; - } - ret = (vm_offset_t)(kmem_alloc(kernel_map, buffer, profile_size)); - if(ret) { - bsd_close_page_cache_files(uid_files); - thread_funnel_set(kernel_flock, funnel_state); - return ENOMEM; - } - *bufsize = profile_size; - while(profile_size) { - int resid_int; - error = vn_rdwr(UIO_READ, data_vp, - (caddr_t) *buffer, profile_size, - profile, UIO_SYSSPACE32, IO_NODELOCKED, - kauth_cred_get(), &resid_int, p); - resid = (vm_size_t) resid_int; - if((error) || (profile_size == resid)) { - bsd_close_page_cache_files(uid_files); - kmem_free(kernel_map, (vm_offset_t)*buffer, profile_size); - thread_funnel_set(kernel_flock, funnel_state); - return EINVAL; - } - profile += profile_size - resid; - profile_size = resid; - } - bsd_close_page_cache_files(uid_files); - thread_funnel_set(kernel_flock, funnel_state); - return 0; - } else { - bsd_close_page_cache_files(uid_files); - thread_funnel_set(kernel_flock, funnel_state); - return EINVAL; - } - -} +extern unsigned int vm_memory_pressure; +SYSCTL_INT(_vm, OID_AUTO, memory_pressure, CTLFLAG_RD, + &vm_memory_pressure, 0, "Memory pressure indicator"); -int -bsd_search_page_cache_data_base( - struct vnode *vp, - struct profile_names_header *database, - char *app_name, - unsigned int mod_date, - unsigned int inode, - off_t *profile, - unsigned int *profile_size) +static int +vm_ctl_page_free_wanted SYSCTL_HANDLER_ARGS { +#pragma unused(oidp, arg1, arg2) + unsigned int page_free_wanted; - struct proc *p; - - unsigned int i; - struct profile_element *element; - unsigned int ele_total; - unsigned int extended_list = 0; - off_t file_off = 0; - unsigned int size; - off_t resid_off; - unsigned int resid; - vm_offset_t local_buf = 0; - - int error; - kern_return_t ret; - - p = current_proc(); - - if(((vm_offset_t)database->element_array) != - sizeof(struct profile_names_header)) { - return EINVAL; - } - element = (struct profile_element *)( - (vm_offset_t)database->element_array + - (vm_offset_t)database); - - ele_total = database->number_of_profiles; - - *profile = 0; - *profile_size = 0; - while(ele_total) { - /* note: code assumes header + n*ele comes out on a page boundary */ - if(((local_buf == 0) && (sizeof(struct profile_names_header) + - (ele_total * sizeof(struct profile_element))) - > (PAGE_SIZE * 4)) || - ((local_buf != 0) && - (ele_total * sizeof(struct profile_element)) - > (PAGE_SIZE * 4))) { - extended_list = ele_total; - if(element == (struct profile_element *) - ((vm_offset_t)database->element_array + - (vm_offset_t)database)) { - ele_total = ((PAGE_SIZE * 4)/sizeof(struct profile_element)) - 1; - } else { - ele_total = (PAGE_SIZE * 4)/sizeof(struct profile_element); - } - extended_list -= ele_total; - } - for (i=0; i - (PAGE_SIZE * 4)) { - size = PAGE_SIZE * 4; - } else { - size = ele_total * sizeof(struct profile_element); - } - resid_off = 0; - while(size) { - int resid_int; - error = vn_rdwr(UIO_READ, vp, - CAST_DOWN(caddr_t, (local_buf + resid_off)), - size, file_off + resid_off, UIO_SYSSPACE32, - IO_NODELOCKED, kauth_cred_get(), &resid_int, p); - resid = (vm_size_t) resid_int; - if((error) || (size == resid)) { - if(local_buf != 0) { - kmem_free(kernel_map, local_buf, 4 * PAGE_SIZE); - } - return EINVAL; - } - resid_off += size-resid; - size = resid; - } - } - if(local_buf != 0) { - kmem_free(kernel_map, local_buf, 4 * PAGE_SIZE); - } - return 0; + page_free_wanted = mach_vm_ctl_page_free_wanted(); + return SYSCTL_OUT(req, &page_free_wanted, sizeof (page_free_wanted)); } +SYSCTL_PROC(_vm, OID_AUTO, page_free_wanted, + CTLTYPE_INT | CTLFLAG_RD | CTLFLAG_LOCKED, + 0, 0, vm_ctl_page_free_wanted, "I", ""); + +extern unsigned int vm_page_purgeable_count; +SYSCTL_INT(_vm, OID_AUTO, page_purgeable_count, CTLFLAG_RD, + &vm_page_purgeable_count, 0, "Purgeable page count"); + +extern unsigned int vm_page_purgeable_wired_count; +SYSCTL_INT(_vm, OID_AUTO, page_purgeable_wired_count, CTLFLAG_RD, + &vm_page_purgeable_wired_count, 0, "Wired purgeable page count"); + +SYSCTL_INT(_vm, OID_AUTO, page_reusable_count, CTLFLAG_RD, + &vm_page_stats_reusable.reusable_count, 0, "Reusable page count"); +SYSCTL_QUAD(_vm, OID_AUTO, reusable_success, CTLFLAG_RD, + &vm_page_stats_reusable.reusable_pages_success, ""); +SYSCTL_QUAD(_vm, OID_AUTO, reusable_failure, CTLFLAG_RD, + &vm_page_stats_reusable.reusable_pages_failure, ""); +SYSCTL_QUAD(_vm, OID_AUTO, reusable_shared, CTLFLAG_RD, + &vm_page_stats_reusable.reusable_pages_shared, ""); +SYSCTL_QUAD(_vm, OID_AUTO, all_reusable_calls, CTLFLAG_RD, + &vm_page_stats_reusable.all_reusable_calls, ""); +SYSCTL_QUAD(_vm, OID_AUTO, partial_reusable_calls, CTLFLAG_RD, + &vm_page_stats_reusable.partial_reusable_calls, ""); +SYSCTL_QUAD(_vm, OID_AUTO, reuse_success, CTLFLAG_RD, + &vm_page_stats_reusable.reuse_pages_success, ""); +SYSCTL_QUAD(_vm, OID_AUTO, reuse_failure, CTLFLAG_RD, + &vm_page_stats_reusable.reuse_pages_failure, ""); +SYSCTL_QUAD(_vm, OID_AUTO, all_reuse_calls, CTLFLAG_RD, + &vm_page_stats_reusable.all_reuse_calls, ""); +SYSCTL_QUAD(_vm, OID_AUTO, partial_reuse_calls, CTLFLAG_RD, + &vm_page_stats_reusable.partial_reuse_calls, ""); +SYSCTL_QUAD(_vm, OID_AUTO, can_reuse_success, CTLFLAG_RD, + &vm_page_stats_reusable.can_reuse_success, ""); +SYSCTL_QUAD(_vm, OID_AUTO, can_reuse_failure, CTLFLAG_RD, + &vm_page_stats_reusable.can_reuse_failure, ""); -int -bsd_write_page_cache_file( - unsigned int user, - char *file_name, - caddr_t buffer, - vm_size_t size, - int mod, - int fid) -{ - struct proc *p; - int resid; - off_t resid_off; - int error; - boolean_t funnel_state; - off_t file_size; - struct vfs_context context; - off_t profile; - unsigned int profile_size; - - vm_offset_t names_buf; - struct vnode *names_vp; - struct vnode *data_vp; - struct profile_names_header *profile_header; - off_t name_offset; - struct global_profile *uid_files; - - - funnel_state = thread_funnel_set(kernel_flock, TRUE); - - - error = bsd_open_page_cache_files(user, &uid_files); - if(error) { - thread_funnel_set(kernel_flock, funnel_state); - return EINVAL; - } - - p = current_proc(); - - names_vp = uid_files->names_vp; - data_vp = uid_files->data_vp; - names_buf = uid_files->buf_ptr; - - /* Stat data file for size */ - - context.vc_proc = p; - context.vc_ucred = kauth_cred_get(); - - if ((error = vnode_size(data_vp, &file_size, &context)) != 0) { - printf("bsd_write_page_cache_file: Can't stat profile data %s\n", file_name); - bsd_close_page_cache_files(uid_files); - thread_funnel_set(kernel_flock, funnel_state); - return error; - } - - if (bsd_search_page_cache_data_base(names_vp, - (struct profile_names_header *)names_buf, - file_name, (unsigned int) mod, - fid, &profile, &profile_size) == 0) { - /* profile is an offset in the profile data base */ - /* It is zero if no profile data was found */ - - if(profile_size == 0) { - unsigned int header_size; - vm_offset_t buf_ptr; - - /* Our Write case */ - - /* read header for last entry */ - profile_header = - (struct profile_names_header *)names_buf; - name_offset = sizeof(struct profile_names_header) + - (sizeof(struct profile_element) - * profile_header->number_of_profiles); - profile_header->number_of_profiles += 1; - - if(name_offset < PAGE_SIZE * 4) { - struct profile_element *name; - /* write new entry */ - name = (struct profile_element *) - (names_buf + (vm_offset_t)name_offset); - name->addr = file_size; - name->size = size; - name->mod_date = mod; - name->inode = fid; - strncpy (name->name, file_name, 12); - } else { - unsigned int ele_size; - struct profile_element name; - /* write new entry */ - name.addr = file_size; - name.size = size; - name.mod_date = mod; - name.inode = fid; - strncpy (name.name, file_name, 12); - /* write element out separately */ - ele_size = sizeof(struct profile_element); - buf_ptr = (vm_offset_t)&name; - resid_off = name_offset; - - while(ele_size) { - error = vn_rdwr(UIO_WRITE, names_vp, - (caddr_t)buf_ptr, - ele_size, resid_off, - UIO_SYSSPACE32, IO_NODELOCKED, - kauth_cred_get(), &resid, p); - if(error) { - printf("bsd_write_page_cache_file: Can't write name_element %x\n", user); - bsd_close_page_cache_files( - uid_files); - thread_funnel_set( - kernel_flock, - funnel_state); - return error; - } - buf_ptr += (vm_offset_t) - ele_size-resid; - resid_off += ele_size-resid; - ele_size = resid; - } - } - - if(name_offset < PAGE_SIZE * 4) { - header_size = name_offset + - sizeof(struct profile_element); - - } else { - header_size = - sizeof(struct profile_names_header); - } - buf_ptr = (vm_offset_t)profile_header; - resid_off = 0; - - /* write names file header */ - while(header_size) { - error = vn_rdwr(UIO_WRITE, names_vp, - (caddr_t)buf_ptr, - header_size, resid_off, - UIO_SYSSPACE32, IO_NODELOCKED, - kauth_cred_get(), &resid, p); - if(error) { - printf("bsd_write_page_cache_file: Can't write header %x\n", user); - bsd_close_page_cache_files( - uid_files); - thread_funnel_set( - kernel_flock, funnel_state); - return error; - } - buf_ptr += (vm_offset_t)header_size-resid; - resid_off += header_size-resid; - header_size = resid; - } - /* write profile to data file */ - resid_off = file_size; - while(size) { - error = vn_rdwr(UIO_WRITE, data_vp, - (caddr_t)buffer, size, resid_off, - UIO_SYSSPACE32, IO_NODELOCKED, - kauth_cred_get(), &resid, p); - if(error) { - printf("bsd_write_page_cache_file: Can't write header %x\n", user); - bsd_close_page_cache_files( - uid_files); - thread_funnel_set( - kernel_flock, funnel_state); - return error; - } - buffer += size-resid; - resid_off += size-resid; - size = resid; - } - bsd_close_page_cache_files(uid_files); - thread_funnel_set(kernel_flock, funnel_state); - return 0; - } - /* Someone else wrote a twin profile before us */ - bsd_close_page_cache_files(uid_files); - thread_funnel_set(kernel_flock, funnel_state); - return 0; - } else { - bsd_close_page_cache_files(uid_files); - thread_funnel_set(kernel_flock, funnel_state); - return EINVAL; - } - -} int -prepare_profile_database(int user) +vm_pressure_monitor( + __unused struct proc *p, + struct vm_pressure_monitor_args *uap, + int *retval) { - const char *cache_path = "/var/vm/app_profile/"; - struct proc *p; - int error; - int resid; - off_t resid_off; - vm_size_t size; - - struct vnode *names_vp; - struct vnode *data_vp; - vm_offset_t names_buf; - vm_offset_t buf_ptr; - - int profile_names_length; - int profile_data_length; - char *profile_data_string; - char *profile_names_string; - char *substring; - - struct vnode_attr va; - struct vfs_context context; - - struct profile_names_header *profile_header; - kern_return_t ret; - - struct nameidata nd_names; - struct nameidata nd_data; - - p = current_proc(); - - context.vc_proc = p; - context.vc_ucred = kauth_cred_get(); + kern_return_t kr; + uint32_t pages_reclaimed; + uint32_t pages_wanted; - ret = kmem_alloc(kernel_map, - (vm_offset_t *)&profile_data_string, PATH_MAX); + kr = mach_vm_pressure_monitor( + (boolean_t) uap->wait_for_pressure, + uap->nsecs_monitored, + (uap->pages_reclaimed) ? &pages_reclaimed : NULL, + &pages_wanted); - if(ret) { - return ENOMEM; - } - - /* Split the buffer in half since we know the size of */ - /* our file path and our allocation is adequate for */ - /* both file path names */ - profile_names_string = profile_data_string + (PATH_MAX/2); - - - strcpy(profile_data_string, cache_path); - strcpy(profile_names_string, cache_path); - profile_names_length = profile_data_length - = strlen(profile_data_string); - substring = profile_data_string + profile_data_length; - sprintf(substring, "%x_data", user); - substring = profile_names_string + profile_names_length; - sprintf(substring, "%x_names", user); - - /* We now have the absolute file names */ - - ret = kmem_alloc(kernel_map, - (vm_offset_t *)&names_buf, 4 * PAGE_SIZE); - if(ret) { - kmem_free(kernel_map, - (vm_offset_t)profile_data_string, PATH_MAX); - return ENOMEM; - } - - NDINIT(&nd_names, LOOKUP, FOLLOW, - UIO_SYSSPACE32, CAST_USER_ADDR_T(profile_names_string), &context); - NDINIT(&nd_data, LOOKUP, FOLLOW, - UIO_SYSSPACE32, CAST_USER_ADDR_T(profile_data_string), &context); - - if ( (error = vn_open(&nd_data, - O_CREAT | O_EXCL | FWRITE, S_IRUSR|S_IWUSR)) ) { - kmem_free(kernel_map, - (vm_offset_t)names_buf, 4 * PAGE_SIZE); - kmem_free(kernel_map, - (vm_offset_t)profile_data_string, PATH_MAX); - - return 0; + switch (kr) { + case KERN_SUCCESS: + break; + case KERN_ABORTED: + return EINTR; + default: + return EINVAL; } - data_vp = nd_data.ni_vp; - - if ( (error = vn_open(&nd_names, - O_CREAT | O_EXCL | FWRITE, S_IRUSR|S_IWUSR)) ) { - printf("prepare_profile_database: Can't create CacheNames %s\n", - profile_data_string); - kmem_free(kernel_map, - (vm_offset_t)names_buf, 4 * PAGE_SIZE); - kmem_free(kernel_map, - (vm_offset_t)profile_data_string, PATH_MAX); - vnode_rele(data_vp); - vnode_put(data_vp); - - return error; - } - names_vp = nd_names.ni_vp; - - /* Write Header for new names file */ - - profile_header = (struct profile_names_header *)names_buf; - - profile_header->number_of_profiles = 0; - profile_header->user_id = user; - profile_header->version = 1; - profile_header->element_array = - sizeof(struct profile_names_header); - profile_header->spare1 = 0; - profile_header->spare2 = 0; - profile_header->spare3 = 0; - - size = sizeof(struct profile_names_header); - buf_ptr = (vm_offset_t)profile_header; - resid_off = 0; - - while(size) { - error = vn_rdwr(UIO_WRITE, names_vp, - (caddr_t)buf_ptr, size, resid_off, - UIO_SYSSPACE32, IO_NODELOCKED, - kauth_cred_get(), &resid, p); - if(error) { - printf("prepare_profile_database: Can't write header %s\n", profile_names_string); - kmem_free(kernel_map, - (vm_offset_t)names_buf, 4 * PAGE_SIZE); - kmem_free(kernel_map, - (vm_offset_t)profile_data_string, - PATH_MAX); - - vnode_rele(names_vp); - vnode_put(names_vp); - vnode_rele(data_vp); - vnode_put(data_vp); - - return error; + if (uap->pages_reclaimed) { + if (copyout((void *)&pages_reclaimed, + uap->pages_reclaimed, + sizeof (pages_reclaimed)) != 0) { + return EFAULT; } - buf_ptr += size-resid; - resid_off += size-resid; - size = resid; } - VATTR_INIT(&va); - VATTR_SET(&va, va_uid, user); - error = vnode_setattr(names_vp, &va, &context); - if(error) { - printf("prepare_profile_database: " - "Can't set user %s\n", profile_names_string); - } - vnode_rele(names_vp); - vnode_put(names_vp); - - VATTR_INIT(&va); - VATTR_SET(&va, va_uid, user); - error = vnode_setattr(data_vp, &va, &context); - if(error) { - printf("prepare_profile_database: " - "Can't set user %s\n", profile_data_string); - } - vnode_rele(data_vp); - vnode_put(data_vp); - - kmem_free(kernel_map, - (vm_offset_t)profile_data_string, PATH_MAX); - kmem_free(kernel_map, - (vm_offset_t)names_buf, 4 * PAGE_SIZE); + *retval = (int) pages_wanted; return 0; - }