X-Git-Url: https://git.saurik.com/apple/xnu.git/blobdiff_plain/316670eb35587141e969394ae8537d66b9211e80..ea3f04195ba4a5034c9c8e9b726d4f7ce96f1832:/bsd/kern/kern_proc.c diff --git a/bsd/kern/kern_proc.c b/bsd/kern/kern_proc.c index 6d696b424..c5ea090ce 100644 --- a/bsd/kern/kern_proc.c +++ b/bsd/kern/kern_proc.c @@ -1,8 +1,8 @@ /* - * Copyright (c) 2000-2007 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 */ @@ -69,7 +69,7 @@ /* HISTORY * 04-Aug-97 Umesh Vaishampayan (umeshv@apple.com) * Added current_proc_EXTERNAL() function for the use of kernel - * lodable modules. + * lodable modules. * * 05-Jun-95 Mac Gillon (mgillon) at NeXT * New version based on 3.3NS and 4.4 @@ -99,11 +99,33 @@ #include #include #include +#include +#include #include #include -#include /* vm_map_switch_protect() */ +#include /* vm_map_switch_protect() */ +#include #include #include +#include +#include +#include +#include +#include +#include +#include /* IOTaskHasEntitlement() */ + +#ifdef CONFIG_32BIT_TELEMETRY +#include +#endif /* CONFIG_32BIT_TELEMETRY */ + +#if CONFIG_CSR +#include +#endif + +#if CONFIG_MEMORYSTATUS +#include +#endif #if CONFIG_MACF #include @@ -111,17 +133,21 @@ #include +#ifdef CONFIG_32BIT_TELEMETRY +#define MAX_32BIT_EXEC_SIG_SIZE 160 +#endif /* CONFIG_32BIT_TELEMETRY */ + /* * Structure associated with user cacheing. */ struct uidinfo { LIST_ENTRY(uidinfo) ui_hash; - uid_t ui_uid; - long ui_proccnt; + uid_t ui_uid; + long ui_proccnt; }; -#define UIHASH(uid) (&uihashtbl[(uid) & uihash]) -LIST_HEAD(uihashhead, uidinfo) *uihashtbl; -u_long uihash; /* size of hash table - 1 */ +#define UIHASH(uid) (&uihashtbl[(uid) & uihash]) +LIST_HEAD(uihashhead, uidinfo) * uihashtbl; +u_long uihash; /* size of hash table - 1 */ /* * Other process lists @@ -137,38 +163,38 @@ struct proclist allproc; struct proclist zombproc; extern struct tty cons; -#if CONFIG_LCTX -/* - * Login Context - */ -static pid_t lastlcid = 1; -static int alllctx_cnt; - -#define LCID_MAX 8192 /* Does this really need to be large? */ -static int maxlcid = LCID_MAX; - -LIST_HEAD(lctxlist, lctx); -static struct lctxlist alllctx; +extern int cs_debug; -lck_mtx_t alllctx_lock; -lck_grp_t * lctx_lck_grp; -lck_grp_attr_t * lctx_lck_grp_attr; -lck_attr_t * lctx_lck_attr; - -static void lctxinit(void); -#endif - -int cs_debug; /* declared further down in this file */ +#if DEVELOPMENT || DEBUG +int syscallfilter_disable = 0; +#endif // DEVELOPMENT || DEBUG #if DEBUG #define __PROC_INTERNAL_DEBUG 1 #endif +#if CONFIG_COREDUMP /* Name to give to core files */ -__private_extern__ char corefilename[MAXPATHLEN+1] = {"/cores/core.%P"}; +#if defined(XNU_TARGET_OS_BRIDGE) +__XNU_PRIVATE_EXTERN char corefilename[MAXPATHLEN + 1] = {"/private/var/internal/%N.core"}; +#elif CONFIG_EMBEDDED +__XNU_PRIVATE_EXTERN char corefilename[MAXPATHLEN + 1] = {"/private/var/cores/%N.core"}; +#else +__XNU_PRIVATE_EXTERN char corefilename[MAXPATHLEN + 1] = {"/cores/core.%P"}; +#endif +#endif + +#if PROC_REF_DEBUG +#include +#endif -static void orphanpg(struct pgrp *pg); -void proc_name_kdp(task_t t, char * buf, int size); -char *proc_name_address(void *p); +typedef uint64_t unaligned_u64 __attribute__((aligned(1))); + +static void orphanpg(struct pgrp * pg); +void proc_name_kdp(task_t t, char * buf, int size); +boolean_t proc_binary_uuid_kdp(task_t task, uuid_t uuid); +int proc_threadname_kdp(void * uth, char * buf, size_t size); +void proc_starttime_kdp(void * p, unaligned_u64 *tv_sec, unaligned_u64 *tv_usec, unaligned_u64 *abstime); +char * proc_name_address(void * p); static void pgrp_add(struct pgrp * pgrp, proc_t parent, proc_t child); static void pgrp_remove(proc_t p); @@ -176,6 +202,7 @@ static void pgrp_replace(proc_t p, struct pgrp *pgrp); static void pgdelete_dropref(struct pgrp *pgrp); extern void pg_rele_dropref(struct pgrp * pgrp); static int csops_internal(pid_t pid, int ops, user_addr_t uaddr, user_size_t usersize, user_addr_t uaddittoken); +static boolean_t proc_parent_is_currentproc(proc_t p); struct fixjob_iterargs { struct pgrp * pg; @@ -185,6 +212,18 @@ struct fixjob_iterargs { int fixjob_callback(proc_t, void *); +uint64_t +get_current_unique_pid(void) +{ + proc_t p = current_proc(); + + if (p) { + return p->p_uniqueid; + } else { + return 0; + } +} + /* * Initialize global process hashing structures. */ @@ -197,8 +236,8 @@ procinit(void) pgrphashtbl = hashinit(maxproc / 4, M_PROC, &pgrphash); sesshashtbl = hashinit(maxproc / 4, M_PROC, &sesshash); uihashtbl = hashinit(maxproc / 16, M_PROC, &uihash); -#if CONFIG_LCTX - lctxinit(); +#if CONFIG_PERSONAS + personas_bootstrap(); #endif } @@ -218,9 +257,11 @@ chgproccnt(uid_t uid, int diff) again: proc_list_lock(); uipp = UIHASH(uid); - for (uip = uipp->lh_first; uip != 0; uip = uip->ui_hash.le_next) - if (uip->ui_uid == uid) + for (uip = uipp->lh_first; uip != 0; uip = uip->ui_hash.le_next) { + if (uip->ui_uid == uid) { break; + } + } if (uip) { uip->ui_proccnt += diff; if (uip->ui_proccnt > 0) { @@ -228,8 +269,9 @@ again: proc_list_unlock(); goto out; } - if (uip->ui_proccnt < 0) + if (uip->ui_proccnt < 0) { panic("chgproccnt: procs < 0"); + } LIST_REMOVE(uip, ui_hash); retval = 0; proc_list_unlock(); @@ -256,13 +298,15 @@ again: } proc_list_unlock(); MALLOC_ZONE(newuip, struct uidinfo *, sizeof(*uip), M_PROC, M_WAITOK); - if (newuip == NULL) + if (newuip == NULL) { panic("chgproccnt: M_PROC zone depleted"); + } goto again; out: - if (newuip != NULL) + if (newuip != NULL) { FREE_ZONE(newuip, sizeof(*uip), M_PROC); - return(retval); + } + return retval; } /* @@ -274,13 +318,15 @@ inferior(proc_t p) int retval = 0; proc_list_lock(); - for (; p != current_proc(); p = p->p_pptr) - if (p->p_pid == 0) + for (; p != current_proc(); p = p->p_pptr) { + if (p->p_pid == 0) { goto out; + } + } retval = 1; out: proc_list_unlock(); - return(retval); + return retval; } /* @@ -294,21 +340,23 @@ isinferior(proc_t p, proc_t t) proc_t start = p; /* if p==t they are not inferior */ - if (p == t) - return(0); + if (p == t) { + return 0; + } proc_list_lock(); for (; p != t; p = p->p_pptr) { nchecked++; /* Detect here if we're in a cycle */ - if ((p->p_pid == 0) || (p->p_pptr == start) || (nchecked >= nprocs)) + if ((p->p_pid == 0) || (p->p_pptr == start) || (nchecked >= nprocs)) { goto out; + } } retval = 1; out: proc_list_unlock(); - return(retval); + return retval; } int @@ -317,22 +365,25 @@ proc_isinferior(int pid1, int pid2) proc_t p = PROC_NULL; proc_t t = PROC_NULL; int retval = 0; - - if (((p = proc_find(pid1)) != (proc_t)0 ) && ((t = proc_find(pid2)) != (proc_t)0)) + + if (((p = proc_find(pid1)) != (proc_t)0) && ((t = proc_find(pid2)) != (proc_t)0)) { retval = isinferior(p, t); + } - if (p != PROC_NULL) + if (p != PROC_NULL) { proc_rele(p); - if (t != PROC_NULL) + } + if (t != PROC_NULL) { proc_rele(t); + } - return(retval); + return retval; } proc_t proc_find(int pid) { - return(proc_findinternal(pid, 0)); + return proc_findinternal(pid, 0); } proc_t @@ -345,14 +396,15 @@ proc_findinternal(int pid, int locked) } p = pfind_locked(pid); - if ((p == PROC_NULL) || (p != proc_ref_locked(p))) + if ((p == PROC_NULL) || (p != proc_ref_locked(p))) { p = PROC_NULL; + } if (locked == 0) { proc_list_unlock(); } - return(p); + return p; } proc_t @@ -363,23 +415,98 @@ proc_findthread(thread_t thread) proc_list_lock(); uth = get_bsdthread_info(thread); - if (uth && (uth->uu_flag & UT_VFORK)) + if (uth && (uth->uu_flag & UT_VFORK)) { p = uth->uu_proc; - else + } else { p = (proc_t)(get_bsdthreadtask_info(thread)); + } p = proc_ref_locked(p); proc_list_unlock(); - return(p); + return p; +} + +void +uthread_reset_proc_refcount(void *uthread) +{ + uthread_t uth; + + uth = (uthread_t) uthread; + uth->uu_proc_refcount = 0; + +#if PROC_REF_DEBUG + if (proc_ref_tracking_disabled) { + return; + } + + uth->uu_pindex = 0; +#endif +} + +#if PROC_REF_DEBUG +int +uthread_get_proc_refcount(void *uthread) +{ + uthread_t uth; + + if (proc_ref_tracking_disabled) { + return 0; + } + + uth = (uthread_t) uthread; + + return uth->uu_proc_refcount; +} +#endif + +static void +record_procref(proc_t p __unused, int count) +{ + uthread_t uth; + + uth = current_uthread(); + uth->uu_proc_refcount += count; + +#if PROC_REF_DEBUG + if (proc_ref_tracking_disabled) { + return; + } + + if (uth->uu_pindex < NUM_PROC_REFS_TO_TRACK) { + backtrace((uintptr_t *) &uth->uu_proc_pcs[uth->uu_pindex], + PROC_REF_STACK_DEPTH, NULL); + + uth->uu_proc_ps[uth->uu_pindex] = p; + uth->uu_pindex++; + } +#endif +} + +static boolean_t +uthread_needs_to_wait_in_proc_refwait(void) +{ + uthread_t uth = current_uthread(); + + /* + * Allow threads holding no proc refs to wait + * in proc_refwait, allowing threads holding + * proc refs to wait in proc_refwait causes + * deadlocks and makes proc_find non-reentrant. + */ + if (uth->uu_proc_refcount == 0) { + return TRUE; + } + + return FALSE; } -int +int proc_rele(proc_t p) { proc_list_lock(); proc_rele_locked(p); proc_list_unlock(); - return(0); + return 0; } proc_t @@ -390,10 +517,11 @@ proc_self(void) p = current_proc(); proc_list_lock(); - if (p != proc_ref_locked(p)) + if (p != proc_ref_locked(p)) { p = PROC_NULL; + } proc_list_unlock(); - return(p); + return p; } @@ -401,67 +529,98 @@ proc_t proc_ref_locked(proc_t p) { proc_t p1 = p; - - /* if process still in creation return failure */ - if ((p == PROC_NULL) || ((p->p_listflag & P_LIST_INCREATE) != 0)) - return (PROC_NULL); - /* do not return process marked for termination */ - if ((p->p_stat != SZOMB) && ((p->p_listflag & P_LIST_EXITED) == 0) && ((p->p_listflag & (P_LIST_DRAINWAIT | P_LIST_DRAIN | P_LIST_DEAD)) == 0)) + int pid = proc_pid(p); + +retry: + /* + * if process still in creation or proc got recycled + * during msleep then return failure. + */ + if ((p == PROC_NULL) || (p1 != p) || ((p->p_listflag & P_LIST_INCREATE) != 0)) { + return PROC_NULL; + } + + /* + * Do not return process marked for termination + * or proc_refdrain called without ref wait. + * Wait for proc_refdrain_with_refwait to complete if + * process in refdrain and refwait flag is set, unless + * the current thread is holding to a proc_ref + * for any proc. + */ + if ((p->p_stat != SZOMB) && + ((p->p_listflag & P_LIST_EXITED) == 0) && + ((p->p_listflag & P_LIST_DEAD) == 0) && + (((p->p_listflag & (P_LIST_DRAIN | P_LIST_DRAINWAIT)) == 0) || + ((p->p_listflag & P_LIST_REFWAIT) != 0))) { + if ((p->p_listflag & P_LIST_REFWAIT) != 0 && uthread_needs_to_wait_in_proc_refwait()) { + msleep(&p->p_listflag, proc_list_mlock, 0, "proc_refwait", 0); + /* + * the proc might have been recycled since we dropped + * the proc list lock, get the proc again. + */ + p = pfind_locked(pid); + goto retry; + } p->p_refcount++; - else + record_procref(p, 1); + } else { p1 = PROC_NULL; + } - return(p1); + return p1; } void proc_rele_locked(proc_t p) { - if (p->p_refcount > 0) { p->p_refcount--; + record_procref(p, -1); if ((p->p_refcount == 0) && ((p->p_listflag & P_LIST_DRAINWAIT) == P_LIST_DRAINWAIT)) { p->p_listflag &= ~P_LIST_DRAINWAIT; wakeup(&p->p_refcount); } - } else + } else { panic("proc_rele_locked -ve ref\n"); - + } } proc_t proc_find_zombref(int pid) { - proc_t p1 = PROC_NULL; - proc_t p = PROC_NULL; + proc_t p; proc_list_lock(); +again: p = pfind_locked(pid); - /* if process still in creation return NULL */ - if ((p == PROC_NULL) || ((p->p_listflag & P_LIST_INCREATE) != 0)) { + /* should we bail? */ + if ((p == PROC_NULL) /* not found */ + || ((p->p_listflag & P_LIST_INCREATE) != 0) /* not created yet */ + || ((p->p_listflag & P_LIST_EXITED) == 0)) { /* not started exit */ proc_list_unlock(); - return (p1); + return PROC_NULL; } - /* if process has not started exit or is being reaped, return NULL */ - if (((p->p_listflag & P_LIST_EXITED) != 0) && ((p->p_listflag & P_LIST_WAITING) == 0)) { - p->p_listflag |= P_LIST_WAITING; - p1 = p; - } else - p1 = PROC_NULL; + /* If someone else is controlling the (unreaped) zombie - wait */ + if ((p->p_listflag & P_LIST_WAITING) != 0) { + (void)msleep(&p->p_stat, proc_list_mlock, PWAIT, "waitcoll", 0); + goto again; + } + p->p_listflag |= P_LIST_WAITING; proc_list_unlock(); - return(p1); + return p; } void proc_drop_zombref(proc_t p) { proc_list_lock(); - if ((p->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) { + if ((p->p_listflag & P_LIST_WAITING) == P_LIST_WAITING) { p->p_listflag &= ~P_LIST_WAITING; wakeup(&p->p_stat); } @@ -472,23 +631,62 @@ proc_drop_zombref(proc_t p) void proc_refdrain(proc_t p) { + proc_refdrain_with_refwait(p, FALSE); +} +proc_t +proc_refdrain_with_refwait(proc_t p, boolean_t get_ref_and_allow_wait) +{ + boolean_t initexec = FALSE; proc_list_lock(); p->p_listflag |= P_LIST_DRAIN; - while (p->p_refcount) { + if (get_ref_and_allow_wait) { + /* + * All the calls to proc_ref_locked will wait + * for the flag to get cleared before returning a ref, + * unless the current thread is holding to a proc ref + * for any proc. + */ + p->p_listflag |= P_LIST_REFWAIT; + if (p == initproc) { + initexec = TRUE; + } + } + + /* Do not wait in ref drain for launchd exec */ + while (p->p_refcount && !initexec) { p->p_listflag |= P_LIST_DRAINWAIT; - msleep(&p->p_refcount, proc_list_mlock, 0, "proc_refdrain", 0) ; + msleep(&p->p_refcount, proc_list_mlock, 0, "proc_refdrain", 0); } + p->p_listflag &= ~P_LIST_DRAIN; - p->p_listflag |= P_LIST_DEAD; + if (!get_ref_and_allow_wait) { + p->p_listflag |= P_LIST_DEAD; + } else { + /* Return a ref to the caller */ + p->p_refcount++; + record_procref(p, 1); + } proc_list_unlock(); + if (get_ref_and_allow_wait) { + return p; + } + return NULL; +} +void +proc_refwake(proc_t p) +{ + proc_list_lock(); + p->p_listflag &= ~P_LIST_REFWAIT; + wakeup(&p->p_listflag); + proc_list_unlock(); } -proc_t +proc_t proc_parentholdref(proc_t p) { proc_t parent = PROC_NULL; @@ -503,7 +701,7 @@ loop: parent = PROC_NULL; goto out; } - + if ((pp->p_listflag & (P_LIST_CHILDDRSTART | P_LIST_CHILDDRAINED)) == P_LIST_CHILDDRSTART) { pp->p_listflag |= P_LIST_CHILDDRWAIT; msleep(&pp->p_childrencnt, proc_list_mlock, 0, "proc_parent", 0); @@ -520,16 +718,17 @@ loop: parent = pp; goto out; } - + out: proc_list_unlock(); - return(parent); + return parent; } -int +int proc_parentdropref(proc_t p, int listlocked) { - if (listlocked == 0) + if (listlocked == 0) { proc_list_lock(); + } if (p->p_parentref > 0) { p->p_parentref--; @@ -537,26 +736,29 @@ proc_parentdropref(proc_t p, int listlocked) p->p_listflag &= ~P_LIST_PARENTREFWAIT; wakeup(&p->p_parentref); } - } else + } else { panic("proc_parentdropref -ve ref\n"); - if (listlocked == 0) + } + if (listlocked == 0) { proc_list_unlock(); + } - return(0); + return 0; } void proc_childdrainstart(proc_t p) { #if __PROC_INTERNAL_DEBUG - if ((p->p_listflag & P_LIST_CHILDDRSTART) == P_LIST_CHILDDRSTART) + if ((p->p_listflag & P_LIST_CHILDDRSTART) == P_LIST_CHILDDRSTART) { panic("proc_childdrainstart: childdrain already started\n"); + } #endif p->p_listflag |= P_LIST_CHILDDRSTART; /* wait for all that hold parentrefs to drop */ while (p->p_parentref > 0) { p->p_listflag |= P_LIST_PARENTREFWAIT; - msleep(&p->p_parentref, proc_list_mlock, 0, "proc_childdrainstart", 0) ; + msleep(&p->p_parentref, proc_list_mlock, 0, "proc_childdrainstart", 0); } } @@ -565,12 +767,13 @@ void proc_childdrainend(proc_t p) { #if __PROC_INTERNAL_DEBUG - if (p->p_childrencnt > 0) + if (p->p_childrencnt > 0) { panic("exiting: children stil hanging around\n"); + } #endif p->p_listflag |= P_LIST_CHILDDRAINED; - if ((p->p_listflag & (P_LIST_CHILDLKWAIT |P_LIST_CHILDDRWAIT)) != 0) { - p->p_listflag &= ~(P_LIST_CHILDLKWAIT |P_LIST_CHILDDRWAIT); + if ((p->p_listflag & (P_LIST_CHILDLKWAIT | P_LIST_CHILDDRWAIT)) != 0) { + p->p_listflag &= ~(P_LIST_CHILDLKWAIT | P_LIST_CHILDDRWAIT); wakeup(&p->p_childrencnt); } } @@ -579,44 +782,133 @@ void proc_checkdeadrefs(__unused proc_t p) { #if __PROC_INTERNAL_DEBUG - if ((p->p_listflag & P_LIST_INHASH) != 0) + if ((p->p_listflag & P_LIST_INHASH) != 0) { panic("proc being freed and still in hash %p: %u\n", p, p->p_listflag); - if (p->p_childrencnt != 0) + } + if (p->p_childrencnt != 0) { panic("proc being freed and pending children cnt %p:%d\n", p, p->p_childrencnt); - if (p->p_refcount != 0) + } + if (p->p_refcount != 0) { panic("proc being freed and pending refcount %p:%d\n", p, p->p_refcount); - if (p->p_parentref != 0) + } + if (p->p_parentref != 0) { panic("proc being freed and pending parentrefs %p:%d\n", p, p->p_parentref); + } #endif } int proc_pid(proc_t p) { - return(p->p_pid); + if (p != NULL) { + return p->p_pid; + } + return -1; } -int +int proc_ppid(proc_t p) { - return(p->p_ppid); + if (p != NULL) { + return p->p_ppid; + } + return -1; +} + +int +proc_original_ppid(proc_t p) +{ + if (p != NULL) { + return p->p_original_ppid; + } + return -1; } -int +int proc_selfpid(void) { - proc_t p = current_proc(); - return(p->p_pid); + return current_proc()->p_pid; } -int +int proc_selfppid(void) { - proc_t p = current_proc(); - return(p->p_ppid); + return current_proc()->p_ppid; +} + +uint64_t +proc_selfcsflags(void) +{ + return (uint64_t)current_proc()->p_csflags; +} + +int +proc_csflags(proc_t p, uint64_t *flags) +{ + if (p && flags) { + *flags = (uint64_t)p->p_csflags; + return 0; + } + return EINVAL; +} + +uint32_t +proc_platform(proc_t p) +{ + if (p != NULL) { + return p->p_platform; + } + return (uint32_t)-1; +} + +uint32_t +proc_sdk(proc_t p) +{ + if (p != NULL) { + return p->p_sdk; + } + return (uint32_t)-1; +} + +#if CONFIG_DTRACE +static proc_t +dtrace_current_proc_vforking(void) +{ + thread_t th = current_thread(); + struct uthread *ut = get_bsdthread_info(th); + + if (ut && + ((ut->uu_flag & (UT_VFORK | UT_VFORKING)) == (UT_VFORK | UT_VFORKING))) { + /* + * Handle the narrow window where we're in the vfork syscall, + * but we're not quite ready to claim (in particular, to DTrace) + * that we're running as the child. + */ + return get_bsdtask_info(get_threadtask(th)); + } + return current_proc(); +} + +int +dtrace_proc_selfpid(void) +{ + return dtrace_current_proc_vforking()->p_pid; +} + +int +dtrace_proc_selfppid(void) +{ + return dtrace_current_proc_vforking()->p_ppid; +} + +uid_t +dtrace_proc_selfruid(void) +{ + return dtrace_current_proc_vforking()->p_ruid; } +#endif /* CONFIG_DTRACE */ -proc_t +proc_t proc_parent(proc_t p) { proc_t parent; @@ -626,21 +918,40 @@ proc_parent(proc_t p) loop: pp = p->p_pptr; parent = proc_ref_locked(pp); - if ((parent == PROC_NULL) && (pp != PROC_NULL) && (pp->p_stat != SZOMB) && ((pp->p_listflag & P_LIST_EXITED) != 0) && ((pp->p_listflag & P_LIST_CHILDDRAINED)== 0)){ + if ((parent == PROC_NULL) && (pp != PROC_NULL) && (pp->p_stat != SZOMB) && ((pp->p_listflag & P_LIST_EXITED) != 0) && ((pp->p_listflag & P_LIST_CHILDDRAINED) == 0)) { pp->p_listflag |= P_LIST_CHILDLKWAIT; msleep(&pp->p_childrencnt, proc_list_mlock, 0, "proc_parent", 0); goto loop; } proc_list_unlock(); - return(parent); + return parent; } +static boolean_t +proc_parent_is_currentproc(proc_t p) +{ + boolean_t ret = FALSE; + + proc_list_lock(); + if (p->p_pptr == current_proc()) { + ret = TRUE; + } + + proc_list_unlock(); + return ret; +} void proc_name(int pid, char * buf, int size) { proc_t p; + if (size <= 0) { + return; + } + + bzero(buf, size); + if ((p = proc_find(pid)) != PROC_NULL) { strlcpy(buf, &p->p_comm[0], size); proc_rele(p); @@ -651,9 +962,71 @@ void proc_name_kdp(task_t t, char * buf, int size) { proc_t p = get_bsdtask_info(t); + if (p == PROC_NULL) { + return; + } - if (p != PROC_NULL) - strlcpy(buf, &p->p_comm[0], size); + if ((size_t)size > sizeof(p->p_comm)) { + strlcpy(buf, &p->p_name[0], MIN((int)sizeof(p->p_name), size)); + } else { + strlcpy(buf, &p->p_comm[0], MIN((int)sizeof(p->p_comm), size)); + } +} + +boolean_t +proc_binary_uuid_kdp(task_t task, uuid_t uuid) +{ + proc_t p = get_bsdtask_info(task); + if (p == PROC_NULL) { + return FALSE; + } + + proc_getexecutableuuid(p, uuid, sizeof(uuid_t)); + + return TRUE; +} + +int +proc_threadname_kdp(void * uth, char * buf, size_t size) +{ + if (size < MAXTHREADNAMESIZE) { + /* this is really just a protective measure for the future in + * case the thread name size in stackshot gets out of sync with + * the BSD max thread name size. Note that bsd_getthreadname + * doesn't take input buffer size into account. */ + return -1; + } + + if (uth != NULL) { + bsd_getthreadname(uth, buf); + } + return 0; +} + + +/* note that this function is generally going to be called from stackshot, + * and the arguments will be coming from a struct which is declared packed + * thus the input arguments will in general be unaligned. We have to handle + * that here. */ +void +proc_starttime_kdp(void *p, unaligned_u64 *tv_sec, unaligned_u64 *tv_usec, unaligned_u64 *abstime) +{ + proc_t pp = (proc_t)p; + if (pp != PROC_NULL) { + if (tv_sec != NULL) { + *tv_sec = pp->p_start.tv_sec; + } + if (tv_usec != NULL) { + *tv_usec = pp->p_start.tv_usec; + } + if (abstime != NULL) { + if (pp->p_stats != NULL) { + *abstime = pp->p_stats->ps_start; + } else { + *abstime = 0; + } + } + } } char * @@ -662,12 +1035,21 @@ proc_name_address(void *p) return &((proc_t)p)->p_comm[0]; } +char * +proc_best_name(proc_t p) +{ + if (p->p_name[0] != 0) { + return &p->p_name[0]; + } + return &p->p_comm[0]; +} + void proc_selfname(char * buf, int size) { proc_t p; - if ((p = current_proc())!= (proc_t)0) { + if ((p = current_proc()) != (proc_t)0) { strlcpy(buf, &p->p_comm[0], size); } } @@ -678,23 +1060,23 @@ proc_signal(int pid, int signum) proc_t p; if ((p = proc_find(pid)) != PROC_NULL) { - psignal(p, signum); - proc_rele(p); - } + psignal(p, signum); + proc_rele(p); + } } int proc_issignal(int pid, sigset_t mask) { proc_t p; - int error=0; + int error = 0; if ((p = proc_find(pid)) != PROC_NULL) { error = proc_pendingsignals(p, mask); proc_rele(p); - } + } - return(error); + return error; } int @@ -702,10 +1084,10 @@ proc_noremotehang(proc_t p) { int retval = 0; - if (p) + if (p) { retval = p->p_flag & P_NOREMOTEHANG; - return(retval? 1: 0); - + } + return retval? 1: 0; } int @@ -713,31 +1095,32 @@ proc_exiting(proc_t p) { int retval = 0; - if (p) + if (p) { retval = p->p_lflag & P_LEXIT; - return(retval? 1: 0); + } + return retval? 1: 0; } int -proc_forcequota(proc_t p) +proc_in_teardown(proc_t p) { int retval = 0; - if (p) - retval = p->p_flag & P_FORCEQUOTA; - return(retval? 1: 0); - + if (p) { + retval = p->p_lflag & P_LPEXIT; + } + return retval? 1: 0; } int -proc_tbe(proc_t p) +proc_forcequota(proc_t p) { int retval = 0; - if (p) - retval = p->p_flag & P_TBE; - return(retval? 1: 0); - + if (p) { + retval = p->p_flag & P_FORCEQUOTA; + } + return retval? 1: 0; } int @@ -749,7 +1132,7 @@ proc_suser(proc_t p) my_cred = kauth_cred_proc_ref(p); error = suser(my_cred, &p->p_acflag); kauth_cred_unref(&my_cred); - return(error); + return error; } task_t @@ -758,7 +1141,7 @@ proc_task(proc_t proc) return (task_t)proc->task; } -/* +/* * Obtain the first thread in a process * * XXX This is a bad thing to do; it exists predominantly to support the @@ -767,20 +1150,21 @@ proc_task(proc_t proc) * XXX needs an audit of the context (proxy vs. not) to clean up. */ thread_t -proc_thread(proc_t proc) -{ - uthread_t uth = TAILQ_FIRST(&proc->p_uthlist); +proc_thread(proc_t proc) +{ + uthread_t uth = TAILQ_FIRST(&proc->p_uthlist); - if (uth != NULL) - return(uth->uu_context.vc_thread); + if (uth != NULL) { + return uth->uu_context.vc_thread; + } - return(NULL); -} + return NULL; +} kauth_cred_t proc_ucred(proc_t p) { - return(p->p_ucred); + return p->p_ucred; } struct uthread * @@ -788,33 +1172,80 @@ current_uthread() { thread_t th = current_thread(); - return((struct uthread *)get_bsdthread_info(th)); + return (struct uthread *)get_bsdthread_info(th); } int proc_is64bit(proc_t p) { - return(IS_64BIT_PROCESS(p)); + return IS_64BIT_PROCESS(p); +} + +int +proc_is64bit_data(proc_t p) +{ + assert(p->task); + return (int)task_get_64bit_data(p->task); } int proc_pidversion(proc_t p) { - return(p->p_idversion); + return p->p_idversion; +} + +uint32_t +proc_persona_id(proc_t p) +{ + return (uint32_t)persona_id_from_proc(p); +} + +uint32_t +proc_getuid(proc_t p) +{ + return p->p_uid; +} + +uint32_t +proc_getgid(proc_t p) +{ + return p->p_gid; } uint64_t proc_uniqueid(proc_t p) { - return(p->p_uniqueid); + return p->p_uniqueid; } uint64_t -proc_selfuniqueid(void) +proc_puniqueid(proc_t p) { - proc_t p = current_proc(); - return(p->p_uniqueid); + return p->p_puniqueid; +} + +void +proc_coalitionids(__unused proc_t p, __unused uint64_t ids[COALITION_NUM_TYPES]) +{ +#if CONFIG_COALITIONS + task_coalition_ids(p->task, ids); +#else + memset(ids, 0, sizeof(uint64_t[COALITION_NUM_TYPES])); +#endif + return; +} + +uint64_t +proc_was_throttled(proc_t p) +{ + return p->was_throttled; +} + +uint64_t +proc_did_throttle(proc_t p) +{ + return p->did_throttle; } int @@ -823,6 +1254,12 @@ proc_getcdhash(proc_t p, unsigned char *cdhash) return vn_getcdhash(p->p_textvp, p->p_textoff, cdhash); } +int +proc_exitstatus(proc_t p) +{ + return p->p_xstat & 0xffff; +} + void proc_getexecutableuuid(proc_t p, unsigned char *uuidbuf, unsigned long size) { @@ -831,26 +1268,144 @@ proc_getexecutableuuid(proc_t p, unsigned char *uuidbuf, unsigned long size) } } +/* Return vnode for executable with an iocount. Must be released with vnode_put() */ +vnode_t +proc_getexecutablevnode(proc_t p) +{ + vnode_t tvp = p->p_textvp; + + if (tvp != NULLVP) { + if (vnode_getwithref(tvp) == 0) { + return tvp; + } + } + + return NULLVP; +} + +int +proc_gettty(proc_t p, vnode_t *vp) +{ + if (!p || !vp) { + return EINVAL; + } + + struct session *procsp = proc_session(p); + int err = EINVAL; + + if (procsp != SESSION_NULL) { + session_lock(procsp); + vnode_t ttyvp = procsp->s_ttyvp; + int ttyvid = procsp->s_ttyvid; + session_unlock(procsp); + + if (ttyvp) { + if (vnode_getwithvid(ttyvp, ttyvid) == 0) { + *vp = procsp->s_ttyvp; + err = 0; + } + } else { + err = ENOENT; + } + + session_rele(procsp); + } + + return err; +} + +int +proc_gettty_dev(proc_t p, dev_t *dev) +{ + struct session *procsp = proc_session(p); + boolean_t has_tty = FALSE; + + if (procsp != SESSION_NULL) { + session_lock(procsp); + + struct tty * tp = SESSION_TP(procsp); + if (tp != TTY_NULL) { + *dev = tp->t_dev; + has_tty = TRUE; + } + + session_unlock(procsp); + session_rele(procsp); + } + + if (has_tty) { + return 0; + } else { + return EINVAL; + } +} + +int +proc_selfexecutableargs(uint8_t *buf, size_t *buflen) +{ + proc_t p = current_proc(); + + // buflen must always be provided + if (buflen == NULL) { + return EINVAL; + } + + // If a buf is provided, there must be at least enough room to fit argc + if (buf && *buflen < sizeof(p->p_argc)) { + return EINVAL; + } + + if (!p->user_stack) { + return EINVAL; + } + + if (buf == NULL) { + *buflen = p->p_argslen + sizeof(p->p_argc); + return 0; + } + + // Copy in argc to the first 4 bytes + memcpy(buf, &p->p_argc, sizeof(p->p_argc)); + + if (*buflen > sizeof(p->p_argc) && p->p_argslen > 0) { + // See memory layout comment in kern_exec.c:exec_copyout_strings() + // We want to copy starting from `p_argslen` bytes away from top of stack + return copyin(p->user_stack - p->p_argslen, + buf + sizeof(p->p_argc), + MIN(p->p_argslen, *buflen - sizeof(p->p_argc))); + } else { + return 0; + } +} + +off_t +proc_getexecutableoffset(proc_t p) +{ + return p->p_textoff; +} void bsd_set_dependency_capable(task_t task) { - proc_t p = get_bsdtask_info(task); + proc_t p = get_bsdtask_info(task); - if (p) { - OSBitOrAtomic(P_DEPENDENCY_CAPABLE, &p->p_flag); - } + if (p) { + OSBitOrAtomic(P_DEPENDENCY_CAPABLE, &p->p_flag); + } } +#ifndef __arm__ int IS_64BIT_PROCESS(proc_t p) { - if (p && (p->p_flag & P_LP64)) - return(1); - else - return(0); + if (p && (p->p_flag & P_LP64)) { + return 1; + } else { + return 0; + } } +#endif /* * Locate a process by number @@ -863,21 +1418,23 @@ pfind_locked(pid_t pid) proc_t q; #endif - if (!pid) - return (kernproc); + if (!pid) { + return kernproc; + } for (p = PIDHASH(pid)->lh_first; p != 0; p = p->p_hash.le_next) { if (p->p_pid == pid) { #if DEBUG for (q = p->p_hash.le_next; q != 0; q = q->p_hash.le_next) { - if ((p !=q) && (q->p_pid == pid)) + if ((p != q) && (q->p_pid == pid)) { panic("two procs with same pid %p:%p:%d:%d\n", p, q, p->p_pid, q->p_pid); + } } #endif - return (p); + return p; } } - return (NULL); + return NULL; } /* @@ -891,13 +1448,15 @@ pzfind(pid_t pid) proc_list_lock(); - for (p = zombproc.lh_first; p != 0; p = p->p_list.le_next) - if (p->p_pid == pid) + for (p = zombproc.lh_first; p != 0; p = p->p_list.le_next) { + if (p->p_pid == pid) { break; + } + } proc_list_unlock(); - return (p); + return p; } /* @@ -911,12 +1470,13 @@ pgfind(pid_t pgid) proc_list_lock(); pgrp = pgfind_internal(pgid); - if ((pgrp == NULL) || ((pgrp->pg_listflags & PGRP_FLAG_TERMINATE) != 0)) + if ((pgrp == NULL) || ((pgrp->pg_listflags & PGRP_FLAG_TERMINATE) != 0)) { pgrp = PGRP_NULL; - else + } else { pgrp->pg_refcount++; + } proc_list_unlock(); - return(pgrp); + return pgrp; } @@ -926,17 +1486,20 @@ pgfind_internal(pid_t pgid) { struct pgrp *pgrp; - for (pgrp = PGRPHASH(pgid)->lh_first; pgrp != 0; pgrp = pgrp->pg_hash.le_next) - if (pgrp->pg_id == pgid) - return (pgrp); - return (NULL); + for (pgrp = PGRPHASH(pgid)->lh_first; pgrp != 0; pgrp = pgrp->pg_hash.le_next) { + if (pgrp->pg_id == pgid) { + return pgrp; + } + } + return NULL; } void pg_rele(struct pgrp * pgrp) { - if(pgrp == PGRP_NULL) + if (pgrp == PGRP_NULL) { return; + } pg_rele_dropref(pgrp); } @@ -959,10 +1522,12 @@ session_find_internal(pid_t sessid) { struct session *sess; - for (sess = SESSHASH(sessid)->lh_first; sess != 0; sess = sess->s_hash.le_next) - if (sess->s_sid == sessid) - return (sess); - return (NULL); + for (sess = SESSHASH(sessid)->lh_first; sess != 0; sess = sess->s_hash.le_next) { + if (sess->s_sid == sessid) { + return sess; + } + } + return NULL; } @@ -991,12 +1556,20 @@ pinsertchild(proc_t parent, proc_t child) TAILQ_INIT(&child->p_evlist); child->p_pptr = parent; child->p_ppid = parent->p_pid; + child->p_original_ppid = parent->p_pid; + child->p_puniqueid = parent->p_uniqueid; + child->p_xhighbits = 0; pg = proc_pgrp(parent); pgrp_add(pg, parent, child); pg_rele(pg); proc_list_lock(); + +#if CONFIG_MEMORYSTATUS + memorystatus_add(child, TRUE); +#endif + parent->p_childrencnt++; LIST_INSERT_HEAD(&parent->p_children, child, p_sibling); @@ -1005,7 +1578,6 @@ pinsertchild(proc_t parent, proc_t child) child->p_listflag &= ~P_LIST_INCREATE; proc_list_unlock(); - } /* @@ -1026,10 +1598,12 @@ enterpgrp(proc_t p, pid_t pgid, int mksess) procsp = proc_session(p); #if DIAGNOSTIC - if (pgrp != NULL && mksess) /* firewalls */ + if (pgrp != NULL && mksess) { /* firewalls */ panic("enterpgrp: setsid into non-empty pgrp"); - if (SESS_LEADER(p, procsp)) + } + if (SESS_LEADER(p, procsp)) { panic("enterpgrp: session leader attempted setpgrp"); + } #endif if (pgrp == PGRP_NULL) { pid_t savepid = p->p_pid; @@ -1038,22 +1612,27 @@ enterpgrp(proc_t p, pid_t pgid, int mksess) * new process group */ #if DIAGNOSTIC - if (p->p_pid != pgid) + if (p->p_pid != pgid) { panic("enterpgrp: new pgrp and pid != pgid"); + } #endif MALLOC_ZONE(pgrp, struct pgrp *, sizeof(struct pgrp), M_PGRP, M_WAITOK); - if (pgrp == NULL) + if (pgrp == NULL) { panic("enterpgrp: M_PGRP zone depleted"); + } if ((np = proc_find(savepid)) == NULL || np != p) { - if (np != PROC_NULL) + if (np != PROC_NULL) { proc_rele(np); - if (mypgrp != PGRP_NULL) + } + if (mypgrp != PGRP_NULL) { pg_rele(mypgrp); - if (procsp != SESSION_NULL) + } + if (procsp != SESSION_NULL) { session_rele(procsp); + } FREE_ZONE(pgrp, sizeof(struct pgrp), M_PGRP); - return (ESRCH); + return ESRCH; } proc_rele(np); if (mksess) { @@ -1063,9 +1642,10 @@ enterpgrp(proc_t p, pid_t pgid, int mksess) * new session */ MALLOC_ZONE(sess, struct session *, - sizeof(struct session), M_SESSION, M_WAITOK); - if (sess == NULL) + sizeof(struct session), M_SESSION, M_WAITOK); + if (sess == NULL) { panic("enterpgrp: M_SESSION zone depleted"); + } sess->s_leader = p; sess->s_sid = p->p_pid; sess->s_count = 1; @@ -1074,11 +1654,9 @@ enterpgrp(proc_t p, pid_t pgid, int mksess) sess->s_flags = 0; sess->s_listflags = 0; sess->s_ttypgrpid = NO_PID; -#if CONFIG_FINE_LOCK_GROUPS + lck_mtx_init(&sess->s_mlock, proc_mlock_grp, proc_lck_attr); -#else - lck_mtx_init(&sess->s_mlock, proc_lck_grp, proc_lck_attr); -#endif + bcopy(procsp->s_login, sess->s_login, sizeof(sess->s_login)); OSBitAndAtomic(~((uint32_t)P_CONTROLT), &p->p_flag); @@ -1087,24 +1665,24 @@ enterpgrp(proc_t p, pid_t pgid, int mksess) proc_list_unlock(); pgrp->pg_session = sess; #if DIAGNOSTIC - if (p != current_proc()) + if (p != current_proc()) { panic("enterpgrp: mksession and p != curproc"); + } #endif } else { proc_list_lock(); pgrp->pg_session = procsp; - - if ((pgrp->pg_session->s_listflags & (S_LIST_TERM | S_LIST_DEAD)) != 0) - panic("enterpgrp: providing ref to terminating session "); + + if ((pgrp->pg_session->s_listflags & (S_LIST_TERM | S_LIST_DEAD)) != 0) { + panic("enterpgrp: providing ref to terminating session "); + } pgrp->pg_session->s_count++; proc_list_unlock(); } pgrp->pg_id = pgid; -#if CONFIG_FINE_LOCK_GROUPS + lck_mtx_init(&pgrp->pg_mlock, proc_mlock_grp, proc_lck_attr); -#else - lck_mtx_init(&pgrp->pg_mlock, proc_lck_grp, proc_lck_attr); -#endif + LIST_INIT(&pgrp->pg_members); pgrp->pg_membercnt = 0; pgrp->pg_jobc = 0; @@ -1115,15 +1693,18 @@ enterpgrp(proc_t p, pid_t pgid, int mksess) proc_list_unlock(); } else if (pgrp == mypgrp) { pg_rele(pgrp); - if (mypgrp != NULL) + if (mypgrp != NULL) { pg_rele(mypgrp); - if (procsp != SESSION_NULL) + } + if (procsp != SESSION_NULL) { session_rele(procsp); - return (0); + } + return 0; } - if (procsp != SESSION_NULL) + if (procsp != SESSION_NULL) { session_rele(procsp); + } /* * Adjust eligibility of affected pgrps to participate in job control. * Increment eligibility counts before decrementing, otherwise we @@ -1132,12 +1713,13 @@ enterpgrp(proc_t p, pid_t pgid, int mksess) fixjobc(p, pgrp, 1); fixjobc(p, mypgrp, 0); - if(mypgrp != PGRP_NULL) + if (mypgrp != PGRP_NULL) { pg_rele(mypgrp); + } pgrp_replace(p, pgrp); pg_rele(pgrp); - return(0); + return 0; } /* @@ -1146,9 +1728,8 @@ enterpgrp(proc_t p, pid_t pgid, int mksess) int leavepgrp(proc_t p) { - pgrp_remove(p); - return (0); + return 0; } /* @@ -1176,7 +1757,7 @@ pgdelete_dropref(struct pgrp *pgrp) } pgrp->pg_listflags |= PGRP_FLAG_TERMINATE; - + if (pgrp->pg_refcount > 0) { proc_list_unlock(); return; @@ -1186,7 +1767,7 @@ pgdelete_dropref(struct pgrp *pgrp) LIST_REMOVE(pgrp, pg_hash); proc_list_unlock(); - + ttyp = SESSION_TP(pgrp->pg_session); if (ttyp != TTY_NULL) { if (ttyp->t_pgrp == pgrp) { @@ -1203,39 +1784,37 @@ pgdelete_dropref(struct pgrp *pgrp) proc_list_lock(); sessp = pgrp->pg_session; - if ((sessp->s_listflags & (S_LIST_TERM | S_LIST_DEAD)) != 0) - panic("pg_deleteref: manipulating refs of already terminating session"); + if ((sessp->s_listflags & (S_LIST_TERM | S_LIST_DEAD)) != 0) { + panic("pg_deleteref: manipulating refs of already terminating session"); + } if (--sessp->s_count == 0) { - if ((sessp->s_listflags & (S_LIST_TERM | S_LIST_DEAD)) != 0) + if ((sessp->s_listflags & (S_LIST_TERM | S_LIST_DEAD)) != 0) { panic("pg_deleteref: terminating already terminated session"); + } sessp->s_listflags |= S_LIST_TERM; ttyp = SESSION_TP(sessp); LIST_REMOVE(sessp, s_hash); proc_list_unlock(); if (ttyp != TTY_NULL) { tty_lock(ttyp); - if (ttyp->t_session == sessp) + if (ttyp->t_session == sessp) { ttyp->t_session = NULL; + } tty_unlock(ttyp); } proc_list_lock(); sessp->s_listflags |= S_LIST_DEAD; - if (sessp->s_count != 0) - panic("pg_deleteref: freeing session in use"); + if (sessp->s_count != 0) { + panic("pg_deleteref: freeing session in use"); + } proc_list_unlock(); -#if CONFIG_FINE_LOCK_GROUPS lck_mtx_destroy(&sessp->s_mlock, proc_mlock_grp); -#else - lck_mtx_destroy(&sessp->s_mlock, proc_lck_grp); -#endif + FREE_ZONE(sessp, sizeof(struct session), M_SESSION); - } else + } else { proc_list_unlock(); -#if CONFIG_FINE_LOCK_GROUPS + } lck_mtx_destroy(&pgrp->pg_mlock, proc_mlock_grp); -#else - lck_mtx_destroy(&pgrp->pg_mlock, proc_lck_grp); -#endif FREE_ZONE(pgrp, sizeof(*pgrp), M_PGRP); } @@ -1266,7 +1845,7 @@ fixjob_callback(proc_t p, void * arg) hispg = proc_pgrp(p); hissess = proc_session(p); - if ((hispg != pg) && + if ((hispg != pg) && (hissess == mysession)) { pgrp_lock(hispg); if (entering) { @@ -1275,15 +1854,18 @@ fixjob_callback(proc_t p, void * arg) } else if (--hispg->pg_jobc == 0) { pgrp_unlock(hispg); orphanpg(hispg); - } else + } else { pgrp_unlock(hispg); + } } - if (hissess != SESSION_NULL) + if (hissess != SESSION_NULL) { session_rele(hissess); - if (hispg != PGRP_NULL) + } + if (hispg != PGRP_NULL) { pg_rele(hispg); + } - return(PROC_RETURNED); + return PROC_RETURNED; } void @@ -1294,12 +1876,26 @@ fixjobc(proc_t p, struct pgrp *pgrp, int entering) struct session *mysession = pgrp->pg_session; proc_t parent; struct fixjob_iterargs fjarg; + boolean_t proc_parent_self; + + /* + * Check if p's parent is current proc, if yes then no need to take + * a ref; calling proc_parent with current proc as parent may + * deadlock if current proc is exiting. + */ + proc_parent_self = proc_parent_is_currentproc(p); + if (proc_parent_self) { + parent = current_proc(); + } else { + parent = proc_parent(p); + } - parent = proc_parent(p); if (parent != PROC_NULL) { - hispgrp = proc_pgrp(parent); + hispgrp = proc_pgrp(parent); hissess = proc_session(parent); - proc_rele(parent); + if (!proc_parent_self) { + proc_rele(parent); + } } @@ -1313,17 +1909,20 @@ fixjobc(proc_t p, struct pgrp *pgrp, int entering) if (entering) { pgrp->pg_jobc++; pgrp_unlock(pgrp); - }else if (--pgrp->pg_jobc == 0) { + } else if (--pgrp->pg_jobc == 0) { pgrp_unlock(pgrp); orphanpg(pgrp); - } else + } else { pgrp_unlock(pgrp); + } } - if (hissess != SESSION_NULL) + if (hissess != SESSION_NULL) { session_rele(hissess); - if (hispgrp != PGRP_NULL) + } + if (hispgrp != PGRP_NULL) { pg_rele(hispgrp); + } /* * Check this process' children to see whether they qualify @@ -1336,66 +1935,152 @@ fixjobc(proc_t p, struct pgrp *pgrp, int entering) proc_childrenwalk(p, fixjob_callback, &fjarg); } -/* - * A process group has become orphaned; - * if there are any stopped processes in the group, - * hang-up all process in that group. +/* + * The pidlist_* routines support the functions in this file that + * walk lists of processes applying filters and callouts to the + * elements of the list. + * + * A prior implementation used a single linear array, which can be + * tricky to allocate on large systems. This implementation creates + * an SLIST of modestly sized arrays of PIDS_PER_ENTRY elements. + * + * The array should be sized large enough to keep the overhead of + * walking the list low, but small enough that blocking allocations of + * pidlist_entry_t structures always succeed. */ + +#define PIDS_PER_ENTRY 1021 + +typedef struct pidlist_entry { + SLIST_ENTRY(pidlist_entry) pe_link; + u_int pe_nused; + pid_t pe_pid[PIDS_PER_ENTRY]; +} pidlist_entry_t; + +typedef struct { + SLIST_HEAD(, pidlist_entry) pl_head; + struct pidlist_entry *pl_active; + u_int pl_nalloc; +} pidlist_t; + +static __inline__ pidlist_t * +pidlist_init(pidlist_t *pl) +{ + SLIST_INIT(&pl->pl_head); + pl->pl_active = NULL; + pl->pl_nalloc = 0; + return pl; +} + +static u_int +pidlist_alloc(pidlist_t *pl, u_int needed) +{ + while (pl->pl_nalloc < needed) { + pidlist_entry_t *pe = kalloc(sizeof(*pe)); + if (NULL == pe) { + panic("no space for pidlist entry"); + } + pe->pe_nused = 0; + SLIST_INSERT_HEAD(&pl->pl_head, pe, pe_link); + pl->pl_nalloc += (sizeof(pe->pe_pid) / sizeof(pe->pe_pid[0])); + } + return pl->pl_nalloc; +} + static void -orphanpg(struct pgrp * pgrp) +pidlist_free(pidlist_t *pl) { - proc_t p; - pid_t * pid_list; - int count, pidcount, i, alloc_count; + pidlist_entry_t *pe; + while (NULL != (pe = SLIST_FIRST(&pl->pl_head))) { + SLIST_FIRST(&pl->pl_head) = SLIST_NEXT(pe, pe_link); + kfree(pe, sizeof(*pe)); + } + pl->pl_nalloc = 0; +} - if (pgrp == PGRP_NULL) - return; - count = 0; - pgrp_lock(pgrp); - for (p = pgrp->pg_members.lh_first; p != 0; p = p->p_pglist.le_next) { - if (p->p_stat == SSTOP) { - for (p = pgrp->pg_members.lh_first; p != 0; - p = p->p_pglist.le_next) - count++; - break; /* ??? stops after finding one.. */ +static __inline__ void +pidlist_set_active(pidlist_t *pl) +{ + pl->pl_active = SLIST_FIRST(&pl->pl_head); + assert(pl->pl_active); +} + +static void +pidlist_add_pid(pidlist_t *pl, pid_t pid) +{ + pidlist_entry_t *pe = pl->pl_active; + if (pe->pe_nused >= sizeof(pe->pe_pid) / sizeof(pe->pe_pid[0])) { + if (NULL == (pe = SLIST_NEXT(pe, pe_link))) { + panic("pidlist allocation exhausted"); } + pl->pl_active = pe; } - pgrp_unlock(pgrp); + pe->pe_pid[pe->pe_nused++] = pid; +} - count += 20; - if (count > hard_maxproc) - count = hard_maxproc; - alloc_count = count * sizeof(pid_t); - pid_list = (pid_t *)kalloc(alloc_count); - bzero(pid_list, alloc_count); - - pidcount = 0; - pgrp_lock(pgrp); - for (p = pgrp->pg_members.lh_first; p != 0; - p = p->p_pglist.le_next) { - if (p->p_stat == SSTOP) { - for (p = pgrp->pg_members.lh_first; p != 0; - p = p->p_pglist.le_next) { - pid_list[pidcount] = p->p_pid; - pidcount++; - if (pidcount >= count) - break; +static __inline__ u_int +pidlist_nalloc(const pidlist_t *pl) +{ + return pl->pl_nalloc; +} + +/* + * A process group has become orphaned; if there are any stopped processes in + * the group, hang-up all process in that group. + */ +static void +orphanpg(struct pgrp *pgrp) +{ + pidlist_t pid_list, *pl = pidlist_init(&pid_list); + u_int pid_count_available = 0; + proc_t p; + + /* allocate outside of the pgrp_lock */ + for (;;) { + pgrp_lock(pgrp); + + boolean_t should_iterate = FALSE; + pid_count_available = 0; + + PGMEMBERS_FOREACH(pgrp, p) { + pid_count_available++; + if (p->p_stat == SSTOP) { + should_iterate = TRUE; } - break; /* ??? stops after finding one.. */ } + if (pid_count_available == 0 || !should_iterate) { + pgrp_unlock(pgrp); + goto out; /* no orphaned processes OR nothing stopped */ + } + if (pidlist_nalloc(pl) >= pid_count_available) { + break; + } + pgrp_unlock(pgrp); + + pidlist_alloc(pl, pid_count_available); } - pgrp_unlock(pgrp); - - if (pidcount == 0) - goto out; + pidlist_set_active(pl); + u_int pid_count = 0; + PGMEMBERS_FOREACH(pgrp, p) { + pidlist_add_pid(pl, proc_pid(p)); + if (++pid_count >= pid_count_available) { + break; + } + } + pgrp_unlock(pgrp); - for (i = 0; i< pidcount; i++) { - /* No handling or proc0 */ - if (pid_list[i] == 0) - continue; - p = proc_find(pid_list[i]); - if (p) { + const pidlist_entry_t *pe; + SLIST_FOREACH(pe, &(pl->pl_head), pe_link) { + for (u_int i = 0; i < pe->pe_nused; i++) { + const pid_t pid = pe->pe_pid[i]; + if (0 == pid) { + continue; /* skip kernproc */ + } + p = proc_find(pid); + if (!p) { + continue; + } proc_transwait(p, 0); pt_setrunnable(p); psignal(p, SIGHUP); @@ -1404,26 +2089,29 @@ orphanpg(struct pgrp * pgrp) } } out: - kfree(pid_list, alloc_count); - return; + pidlist_free(pl); } - - -/* XXX should be __private_extern__ */ int -proc_is_classic(proc_t p) +proc_is_classic(proc_t p __unused) { - return (p->p_flag & P_TRANSLATED) ? 1 : 0; + return 0; } /* XXX Why does this function exist? Need to kill it off... */ proc_t current_proc_EXTERNAL(void) { - return (current_proc()); + return current_proc(); +} + +int +proc_is_forcing_hfs_case_sensitivity(proc_t p) +{ + return (p->p_vfs_iopolicy & P_VFS_IOPOLICY_FORCE_HFS_CASE_SENSITIVITY) ? 1 : 0; } +#if CONFIG_COREDUMP /* * proc_core_name(name, uid, pid) * Expand the name described in corefilename, using name, uid, and pid. @@ -1437,44 +2125,48 @@ current_proc_EXTERNAL(void) */ __private_extern__ int proc_core_name(const char *name, uid_t uid, pid_t pid, char *cf_name, - size_t cf_name_len) + size_t cf_name_len) { const char *format, *appendstr; - char id_buf[11]; /* Buffer for pid/uid -- max 4B */ + char id_buf[11]; /* Buffer for pid/uid -- max 4B */ size_t i, l, n; - if (cf_name == NULL) + if (cf_name == NULL) { goto toolong; + } format = corefilename; for (i = 0, n = 0; n < cf_name_len && format[i]; i++) { switch (format[i]) { - case '%': /* Format character */ + case '%': /* Format character */ i++; switch (format[i]) { case '%': appendstr = "%"; break; - case 'N': /* process name */ + case 'N': /* process name */ appendstr = name; break; - case 'P': /* process id */ + case 'P': /* process id */ snprintf(id_buf, sizeof(id_buf), "%u", pid); appendstr = id_buf; break; - case 'U': /* user id */ + case 'U': /* user id */ snprintf(id_buf, sizeof(id_buf), "%u", uid); appendstr = id_buf; break; + case '\0': /* format string ended in % symbol */ + goto endofstring; default: appendstr = ""; - log(LOG_ERR, + log(LOG_ERR, "Unknown format character %c in `%s'\n", format[i], format); } l = strlen(appendstr); - if ((n + l) >= cf_name_len) + if ((n + l) >= cf_name_len) { goto toolong; + } bcopy(appendstr, cf_name + n, l); n += l; break; @@ -1482,254 +2174,66 @@ proc_core_name(const char *name, uid_t uid, pid_t pid, char *cf_name, cf_name[n++] = format[i]; } } - if (format[i] != '\0') + if (format[i] != '\0') { goto toolong; - return (0); + } + return 0; toolong: log(LOG_ERR, "pid %ld (%s), uid (%u): corename is too long\n", (long)pid, name, (uint32_t)uid); - return (1); + return 1; +endofstring: + log(LOG_ERR, "pid %ld (%s), uid (%u): unexpected end of string after %% token\n", + (long)pid, name, (uint32_t)uid); + return 1; } +#endif /* CONFIG_COREDUMP */ -#if CONFIG_LCTX +/* Code Signing related routines */ -static void -lctxinit(void) +int +csops(__unused proc_t p, struct csops_args *uap, __unused int32_t *retval) { - LIST_INIT(&alllctx); - alllctx_cnt = 0; - - /* allocate lctx lock group attribute and group */ - lctx_lck_grp_attr = lck_grp_attr_alloc_init(); - lck_grp_attr_setstat(lctx_lck_grp_attr); - - lctx_lck_grp = lck_grp_alloc_init("lctx", lctx_lck_grp_attr); - /* Allocate lctx lock attribute */ - lctx_lck_attr = lck_attr_alloc_init(); - - lck_mtx_init(&alllctx_lock, lctx_lck_grp, lctx_lck_attr); + return csops_internal(uap->pid, uap->ops, uap->useraddr, + uap->usersize, USER_ADDR_NULL); } -/* - * Locate login context by number. - */ -struct lctx * -lcfind(pid_t lcid) +int +csops_audittoken(__unused proc_t p, struct csops_audittoken_args *uap, __unused int32_t *retval) { - struct lctx *l; - - ALLLCTX_LOCK; - LIST_FOREACH(l, &alllctx, lc_list) { - if (l->lc_id == lcid) { - LCTX_LOCK(l); - break; - } + if (uap->uaudittoken == USER_ADDR_NULL) { + return EINVAL; } - ALLLCTX_UNLOCK; - return (l); -} - -#define LCID_INC \ - do { \ - lastlcid++; \ - if (lastlcid > maxlcid) \ - lastlcid = 1; \ - } while (0) \ - -struct lctx * -lccreate(void) -{ - struct lctx *l; - pid_t newlcid; - - /* Not very efficient but this isn't a common operation. */ - while ((l = lcfind(lastlcid)) != NULL) { - LCTX_UNLOCK(l); - LCID_INC; - } - newlcid = lastlcid; - LCID_INC; - - MALLOC(l, struct lctx *, sizeof(struct lctx), M_LCTX, M_WAITOK|M_ZERO); - l->lc_id = newlcid; - LIST_INIT(&l->lc_members); - lck_mtx_init(&l->lc_mtx, lctx_lck_grp, lctx_lck_attr); -#if CONFIG_MACF - l->lc_label = mac_lctx_label_alloc(); -#endif - ALLLCTX_LOCK; - LIST_INSERT_HEAD(&alllctx, l, lc_list); - alllctx_cnt++; - ALLLCTX_UNLOCK; - - return (l); -} - -/* - * Call with proc protected (either by being invisible - * or by having the all-login-context lock held) and - * the lctx locked. - * - * Will unlock lctx on return. - */ -void -enterlctx (proc_t p, struct lctx *l, __unused int create) -{ - if (l == NULL) - return; - - p->p_lctx = l; - LIST_INSERT_HEAD(&l->lc_members, p, p_lclist); - l->lc_mc++; - -#if CONFIG_MACF - if (create) - mac_lctx_notify_create(p, l); - else - mac_lctx_notify_join(p, l); -#endif - LCTX_UNLOCK(l); - - return; -} - -/* - * Remove process from login context (if any). Called with p protected by - * the alllctx lock. - */ -void -leavelctx (proc_t p) -{ - struct lctx *l; - - if (p->p_lctx == NULL) { - return; - } - - LCTX_LOCK(p->p_lctx); - l = p->p_lctx; - p->p_lctx = NULL; - LIST_REMOVE(p, p_lclist); - l->lc_mc--; -#if CONFIG_MACF - mac_lctx_notify_leave(p, l); -#endif - if (LIST_EMPTY(&l->lc_members)) { - LIST_REMOVE(l, lc_list); - alllctx_cnt--; - LCTX_UNLOCK(l); - lck_mtx_destroy(&l->lc_mtx, lctx_lck_grp); -#if CONFIG_MACF - mac_lctx_label_free(l->lc_label); - l->lc_label = NULL; -#endif - FREE(l, M_LCTX); - } else { - LCTX_UNLOCK(l); - } - return; + return csops_internal(uap->pid, uap->ops, uap->useraddr, + uap->usersize, uap->uaudittoken); } static int -sysctl_kern_lctx SYSCTL_HANDLER_ARGS +csops_copy_token(void *start, size_t length, user_size_t usize, user_addr_t uaddr) { - int *name = (int*) arg1; - u_int namelen = arg2; - struct kinfo_lctx kil; - struct lctx *l; + char fakeheader[8] = { 0 }; int error; - error = 0; - - switch (oidp->oid_number) { - case KERN_LCTX_ALL: - ALLLCTX_LOCK; - /* Request for size. */ - if (!req->oldptr) { - error = SYSCTL_OUT(req, 0, - sizeof(struct kinfo_lctx) * (alllctx_cnt + 1)); - goto out; - } - break; - - case KERN_LCTX_LCID: - /* No space */ - if (req->oldlen < sizeof(struct kinfo_lctx)) - return (ENOMEM); - /* No argument */ - if (namelen != 1) - return (EINVAL); - /* No login context */ - l = lcfind((pid_t)name[0]); - if (l == NULL) - return (ENOENT); - kil.id = l->lc_id; - kil.mc = l->lc_mc; - LCTX_UNLOCK(l); - return (SYSCTL_OUT(req, (caddr_t)&kil, sizeof(kil))); - - default: - return (EINVAL); + if (usize < sizeof(fakeheader)) { + return ERANGE; } - /* Provided buffer is too small. */ - if (req->oldlen < (sizeof(struct kinfo_lctx) * alllctx_cnt)) { - error = ENOMEM; - goto out; - } + /* if no blob, fill in zero header */ + if (NULL == start) { + start = fakeheader; + length = sizeof(fakeheader); + } else if (usize < length) { + /* ... if input too short, copy out length of entitlement */ + uint32_t length32 = htonl((uint32_t)length); + memcpy(&fakeheader[4], &length32, sizeof(length32)); - LIST_FOREACH(l, &alllctx, lc_list) { - LCTX_LOCK(l); - kil.id = l->lc_id; - kil.mc = l->lc_mc; - LCTX_UNLOCK(l); - error = SYSCTL_OUT(req, (caddr_t)&kil, sizeof(kil)); - if (error) - break; + error = copyout(fakeheader, uaddr, sizeof(fakeheader)); + if (error == 0) { + return ERANGE; /* input buffer to short, ERANGE signals that */ + } + return error; } -out: - ALLLCTX_UNLOCK; - - return (error); -} - -SYSCTL_NODE(_kern, KERN_LCTX, lctx, CTLFLAG_RW|CTLFLAG_LOCKED, 0, "Login Context"); - -SYSCTL_PROC(_kern_lctx, KERN_LCTX_ALL, all, CTLFLAG_RD|CTLTYPE_STRUCT | CTLFLAG_LOCKED, - 0, 0, sysctl_kern_lctx, "S,lctx", - "Return entire login context table"); -SYSCTL_NODE(_kern_lctx, KERN_LCTX_LCID, lcid, CTLFLAG_RD | CTLFLAG_LOCKED, - sysctl_kern_lctx, "Login Context Table"); -SYSCTL_INT(_kern_lctx, OID_AUTO, last, CTLFLAG_RD | CTLFLAG_LOCKED, &lastlcid, 0, ""); -SYSCTL_INT(_kern_lctx, OID_AUTO, count, CTLFLAG_RD | CTLFLAG_LOCKED, &alllctx_cnt, 0, ""); -SYSCTL_INT(_kern_lctx, OID_AUTO, max, CTLFLAG_RW | CTLFLAG_LOCKED, &maxlcid, 0, ""); - -#endif /* LCTX */ - -/* Code Signing related routines */ - -int -csops(__unused proc_t p, struct csops_args *uap, __unused int32_t *retval) -{ - return(csops_internal(uap->pid, uap->ops, uap->useraddr, - uap->usersize, USER_ADDR_NULL)); -} - -int -csops_audittoken(__unused proc_t p, struct csops_audittoken_args *uap, __unused int32_t *retval) -{ - if (uap->uaudittoken == USER_ADDR_NULL) - return(EINVAL); - switch (uap->ops) { - case CS_OPS_PIDPATH: - case CS_OPS_ENTITLEMENTS_BLOB: - break; - default: - return(EINVAL); - }; - - return(csops_internal(uap->pid, uap->ops, uap->useraddr, - uap->usersize, uap->uaudittoken)); + return copyout(start, uaddr, length); } static int @@ -1737,56 +2241,53 @@ csops_internal(pid_t pid, int ops, user_addr_t uaddr, user_size_t usersize, user { size_t usize = (size_t)CAST_DOWN(size_t, usersize); proc_t pt; - uint32_t retflags; - int vid, forself; + int forself; int error; vnode_t tvp; off_t toff; - char * buf; unsigned char cdhash[SHA1_RESULTLEN]; audit_token_t token; - unsigned int upid=0, uidversion = 0; - + unsigned int upid = 0, uidversion = 0; + forself = error = 0; - if (pid == 0) + if (pid == 0) { pid = proc_selfpid(); - if (pid == proc_selfpid()) + } + if (pid == proc_selfpid()) { forself = 1; + } - /* Pre flight checks for CS_OPS_PIDPATH */ - if (ops == CS_OPS_PIDPATH) { - /* usize is unsigned.. */ - if (usize > 4 * PATH_MAX) - return(EOVERFLOW); - if (kauth_cred_issuser(kauth_cred_get()) != TRUE) - return(EPERM); - } else { - switch (ops) { - case CS_OPS_STATUS: - case CS_OPS_CDHASH: - case CS_OPS_PIDOFFSET: - case CS_OPS_ENTITLEMENTS_BLOB: - break; /* unrestricted */ - default: - if (forself == 0 && kauth_cred_issuser(kauth_cred_get()) != TRUE) - return(EPERM); - break; + switch (ops) { + case CS_OPS_STATUS: + case CS_OPS_CDHASH: + case CS_OPS_PIDOFFSET: + case CS_OPS_ENTITLEMENTS_BLOB: + case CS_OPS_IDENTITY: + case CS_OPS_BLOB: + case CS_OPS_TEAMID: + case CS_OPS_CLEAR_LV: + break; /* not restricted to root */ + default: + if (forself == 0 && kauth_cred_issuser(kauth_cred_get()) != TRUE) { + return EPERM; } + break; } pt = proc_find(pid); - if (pt == PROC_NULL) - return(ESRCH); + if (pt == PROC_NULL) { + return ESRCH; + } upid = pt->p_pid; uidversion = pt->p_idversion; if (uaudittoken != USER_ADDR_NULL) { - error = copyin(uaudittoken, &token, sizeof(audit_token_t)); - if (error != 0) + if (error != 0) { goto out; + } /* verify the audit token pid/idversion matches with proc */ if ((token.val[5] != upid) || (token.val[7] != uidversion)) { error = ESRCH; @@ -1794,524 +2295,659 @@ csops_internal(pid_t pid, int ops, user_addr_t uaddr, user_size_t usersize, user } } +#if CONFIG_MACF switch (ops) { + case CS_OPS_MARKINVALID: + case CS_OPS_MARKHARD: + case CS_OPS_MARKKILL: + case CS_OPS_MARKRESTRICT: + case CS_OPS_SET_STATUS: + case CS_OPS_CLEARINSTALLER: + case CS_OPS_CLEARPLATFORM: + case CS_OPS_CLEAR_LV: + if ((error = mac_proc_check_set_cs_info(current_proc(), pt, ops))) { + goto out; + } + break; + default: + if ((error = mac_proc_check_get_cs_info(current_proc(), pt, ops))) { + goto out; + } + } +#endif - case CS_OPS_STATUS: - retflags = pt->p_csflags; - if (uaddr != USER_ADDR_NULL) - error = copyout(&retflags, uaddr, sizeof(uint32_t)); - break; - - case CS_OPS_MARKINVALID: - proc_lock(pt); - if ((pt->p_csflags & CS_VALID) == CS_VALID) { /* is currently valid */ - pt->p_csflags &= ~CS_VALID; /* set invalid */ - if ((pt->p_csflags & CS_KILL) == CS_KILL) { - pt->p_csflags |= CS_KILLED; - proc_unlock(pt); - if (cs_debug) { - printf("CODE SIGNING: marked invalid by pid %d: " - "p=%d[%s] honoring CS_KILL, final status 0x%x\n", - proc_selfpid(), pt->p_pid, pt->p_comm, pt->p_csflags); - } - psignal(pt, SIGKILL); - } else - proc_unlock(pt); - } else - proc_unlock(pt); - - break; + switch (ops) { + case CS_OPS_STATUS: { + uint32_t retflags; - case CS_OPS_MARKHARD: - proc_lock(pt); - pt->p_csflags |= CS_HARD; - if ((pt->p_csflags & CS_VALID) == 0) { - /* @@@ allow? reject? kill? @@@ */ - proc_unlock(pt); - error = EINVAL; - goto out; - } else - proc_unlock(pt); - break; + proc_lock(pt); + retflags = pt->p_csflags; + if (cs_process_enforcement(pt)) { + retflags |= CS_ENFORCEMENT; + } + if (csproc_get_platform_binary(pt)) { + retflags |= CS_PLATFORM_BINARY; + } + if (csproc_get_platform_path(pt)) { + retflags |= CS_PLATFORM_PATH; + } + //Don't return CS_REQUIRE_LV if we turned it on with CS_FORCED_LV but still report CS_FORCED_LV + if ((pt->p_csflags & CS_FORCED_LV) == CS_FORCED_LV) { + retflags &= (~CS_REQUIRE_LV); + } + proc_unlock(pt); - case CS_OPS_MARKKILL: - proc_lock(pt); - pt->p_csflags |= CS_KILL; - if ((pt->p_csflags & CS_VALID) == 0) { + if (uaddr != USER_ADDR_NULL) { + error = copyout(&retflags, uaddr, sizeof(uint32_t)); + } + break; + } + case CS_OPS_MARKINVALID: + proc_lock(pt); + if ((pt->p_csflags & CS_VALID) == CS_VALID) { /* is currently valid */ + pt->p_csflags &= ~CS_VALID; /* set invalid */ + if ((pt->p_csflags & CS_KILL) == CS_KILL) { + pt->p_csflags |= CS_KILLED; proc_unlock(pt); + if (cs_debug) { + printf("CODE SIGNING: marked invalid by pid %d: " + "p=%d[%s] honoring CS_KILL, final status 0x%x\n", + proc_selfpid(), pt->p_pid, pt->p_comm, pt->p_csflags); + } psignal(pt, SIGKILL); - } else + } else { proc_unlock(pt); - break; + } + } else { + proc_unlock(pt); + } - case CS_OPS_PIDPATH: - tvp = pt->p_textvp; - vid = vnode_vid(tvp); + break; - if (tvp == NULLVP) { - proc_rele(pt); - return(EINVAL); - } + case CS_OPS_MARKHARD: + proc_lock(pt); + pt->p_csflags |= CS_HARD; + if ((pt->p_csflags & CS_VALID) == 0) { + /* @@@ allow? reject? kill? @@@ */ + proc_unlock(pt); + error = EINVAL; + goto out; + } else { + proc_unlock(pt); + } + break; - buf = (char *)kalloc(usize); - if (buf == NULL) { - proc_rele(pt); - return(ENOMEM); - } - bzero(buf, usize); - - error = vnode_getwithvid(tvp, vid); - if (error == 0) { - int len; - len = usize; - error = vn_getpath(tvp, buf, &len); - vnode_put(tvp); - if (error == 0) { - error = copyout(buf, uaddr, usize); - } - kfree(buf, usize); - } + case CS_OPS_MARKKILL: + proc_lock(pt); + pt->p_csflags |= CS_KILL; + if ((pt->p_csflags & CS_VALID) == 0) { + proc_unlock(pt); + psignal(pt, SIGKILL); + } else { + proc_unlock(pt); + } + break; - proc_rele(pt); + case CS_OPS_PIDOFFSET: + toff = pt->p_textoff; + proc_rele(pt); + error = copyout(&toff, uaddr, sizeof(toff)); + return error; - return(error); + case CS_OPS_CDHASH: - case CS_OPS_PIDOFFSET: - toff = pt->p_textoff; + /* pt already holds a reference on its p_textvp */ + tvp = pt->p_textvp; + toff = pt->p_textoff; + + if (tvp == NULLVP || usize != SHA1_RESULTLEN) { proc_rele(pt); - error = copyout(&toff, uaddr, sizeof(toff)); - return(error); + return EINVAL; + } - case CS_OPS_CDHASH: + error = vn_getcdhash(tvp, toff, cdhash); + proc_rele(pt); - /* pt already holds a reference on its p_textvp */ - tvp = pt->p_textvp; - toff = pt->p_textoff; + if (error == 0) { + error = copyout(cdhash, uaddr, sizeof(cdhash)); + } - if (tvp == NULLVP || usize != SHA1_RESULTLEN) { - proc_rele(pt); - return EINVAL; - } + return error; - error = vn_getcdhash(tvp, toff, cdhash); - proc_rele(pt); + case CS_OPS_ENTITLEMENTS_BLOB: { + void *start; + size_t length; - if (error == 0) { - error = copyout(cdhash, uaddr, sizeof (cdhash)); - } + proc_lock(pt); - return error; + if ((pt->p_csflags & (CS_VALID | CS_DEBUGGED)) == 0) { + proc_unlock(pt); + error = EINVAL; + break; + } - case CS_OPS_ENTITLEMENTS_BLOB: { - char fakeheader[8] = { 0 }; - void *start; - size_t length; + error = cs_entitlements_blob_get(pt, &start, &length); + proc_unlock(pt); + if (error) { + break; + } - if ((pt->p_csflags & CS_VALID) == 0) { - error = EINVAL; - break; - } - if (usize < sizeof(fakeheader)) { - error = ERANGE; - break; - } - if (0 != (error = cs_entitlements_blob_get(pt, - &start, &length))) - break; - /* if no entitlement, fill in zero header */ - if (NULL == start) { - start = fakeheader; - length = sizeof(fakeheader); - } else if (usize < length) { - /* ... if input too short, copy out length of entitlement */ - uint32_t length32 = htonl((uint32_t)length); - memcpy(&fakeheader[4], &length32, sizeof(length32)); - - error = copyout(fakeheader, uaddr, sizeof(fakeheader)); - if (error == 0) - error = ERANGE; /* input buffer to short, ERANGE signals that */ - break; - } - error = copyout(start, uaddr, length); + error = csops_copy_token(start, length, usize, uaddr); + break; + } + case CS_OPS_MARKRESTRICT: + proc_lock(pt); + pt->p_csflags |= CS_RESTRICT; + proc_unlock(pt); + break; + + case CS_OPS_SET_STATUS: { + uint32_t flags; + + if (usize < sizeof(flags)) { + error = ERANGE; break; } - case CS_OPS_MARKRESTRICT: + error = copyin(uaddr, &flags, sizeof(flags)); + if (error) { + break; + } + + /* only allow setting a subset of all code sign flags */ + flags &= + CS_HARD | CS_EXEC_SET_HARD | + CS_KILL | CS_EXEC_SET_KILL | + CS_RESTRICT | + CS_REQUIRE_LV | + CS_ENFORCEMENT | CS_EXEC_SET_ENFORCEMENT; + + proc_lock(pt); + if (pt->p_csflags & CS_VALID) { + pt->p_csflags |= flags; + } else { + error = EINVAL; + } + proc_unlock(pt); + + break; + } + case CS_OPS_CLEAR_LV: { + /* + * This option is used to remove library validation from + * a running process. This is used in plugin architectures + * when a program needs to load untrusted libraries. This + * allows the process to maintain library validation as + * long as possible, then drop it only when required. + * Once a process has loaded the untrusted library, + * relying on library validation in the future will + * not be effective. An alternative is to re-exec + * your application without library validation, or + * fork an untrusted child. + */ +#ifdef CONFIG_EMBEDDED + // On embedded platforms, we don't support dropping LV + error = ENOTSUP; +#else + /* + * if we have the flag set, and the caller wants + * to remove it, and they're entitled to, then + * we remove it from the csflags + * + * NOTE: We are fine to poke into the task because + * we get a ref to pt when we do the proc_find + * at the beginning of this function. + * + * We also only allow altering ourselves. + */ + if (forself == 1 && IOTaskHasEntitlement(pt->task, CLEAR_LV_ENTITLEMENT)) { proc_lock(pt); - pt->p_csflags |= CS_RESTRICT; + pt->p_csflags &= (~(CS_REQUIRE_LV | CS_FORCED_LV)); + proc_unlock(pt); + error = 0; + } else { + error = EPERM; + } +#endif + break; + } + case CS_OPS_BLOB: { + void *start; + size_t length; + + proc_lock(pt); + if ((pt->p_csflags & (CS_VALID | CS_DEBUGGED)) == 0) { proc_unlock(pt); + error = EINVAL; break; + } - default: + error = cs_blob_get(pt, &start, &length); + proc_unlock(pt); + if (error) { + break; + } + + error = csops_copy_token(start, length, usize, uaddr); + break; + } + case CS_OPS_IDENTITY: + case CS_OPS_TEAMID: { + const char *identity; + uint8_t fakeheader[8]; + uint32_t idlen; + size_t length; + + /* + * Make identity have a blob header to make it + * easier on userland to guess the identity + * length. + */ + if (usize < sizeof(fakeheader)) { + error = ERANGE; + break; + } + memset(fakeheader, 0, sizeof(fakeheader)); + + proc_lock(pt); + if ((pt->p_csflags & (CS_VALID | CS_DEBUGGED)) == 0) { + proc_unlock(pt); error = EINVAL; break; + } + + identity = ops == CS_OPS_TEAMID ? csproc_get_teamid(pt) : cs_identity_get(pt); + proc_unlock(pt); + if (identity == NULL) { + error = ENOENT; + break; + } + + length = strlen(identity) + 1; /* include NUL */ + idlen = htonl(length + sizeof(fakeheader)); + memcpy(&fakeheader[4], &idlen, sizeof(idlen)); + + error = copyout(fakeheader, uaddr, sizeof(fakeheader)); + if (error) { + break; + } + + if (usize < sizeof(fakeheader) + length) { + error = ERANGE; + } else if (usize > sizeof(fakeheader)) { + error = copyout(identity, uaddr + sizeof(fakeheader), length); + } + + break; + } + + case CS_OPS_CLEARINSTALLER: + proc_lock(pt); + pt->p_csflags &= ~(CS_INSTALLER | CS_DATAVAULT_CONTROLLER | CS_EXEC_INHERIT_SIP); + proc_unlock(pt); + break; + + case CS_OPS_CLEARPLATFORM: +#if DEVELOPMENT || DEBUG + if (cs_process_global_enforcement()) { + error = ENOTSUP; + break; + } + +#if CONFIG_CSR + if (csr_check(CSR_ALLOW_APPLE_INTERNAL) != 0) { + error = ENOTSUP; + break; + } +#endif + + proc_lock(pt); + pt->p_csflags &= ~(CS_PLATFORM_BINARY | CS_PLATFORM_PATH); + csproc_clear_platform_binary(pt); + proc_unlock(pt); + break; +#else + error = ENOTSUP; + break; +#endif /* !DEVELOPMENT || DEBUG */ + + default: + error = EINVAL; + break; } out: proc_rele(pt); - return(error); + return error; } -int -proc_iterate(flags, callout, arg, filterfn, filterarg) - int flags; - int (*callout)(proc_t, void *); - void * arg; - int (*filterfn)(proc_t, void *); - void * filterarg; +void +proc_iterate( + unsigned int flags, + proc_iterate_fn_t callout, + void *arg, + proc_iterate_fn_t filterfn, + void *filterarg) { - proc_t p; - pid_t * pid_list; - int count, pidcount, alloc_count, i, retval; + pidlist_t pid_list, *pl = pidlist_init(&pid_list); + u_int pid_count_available = 0; - count = nprocs+ 10; - if (count > hard_maxproc) - count = hard_maxproc; - alloc_count = count * sizeof(pid_t); - pid_list = (pid_t *)kalloc(alloc_count); - bzero(pid_list, alloc_count); + assert(callout != NULL); + /* allocate outside of the proc_list_lock */ + for (;;) { + proc_list_lock(); + pid_count_available = nprocs + 1; /* kernel_task not counted in nprocs */ + assert(pid_count_available > 0); + if (pidlist_nalloc(pl) > pid_count_available) { + break; + } + proc_list_unlock(); - proc_list_lock(); + pidlist_alloc(pl, pid_count_available); + } + pidlist_set_active(pl); + /* filter pids into the pid_list */ - pidcount = 0; + u_int pid_count = 0; if (flags & PROC_ALLPROCLIST) { - for (p = allproc.lh_first; (p != 0); p = p->p_list.le_next) { - if (p->p_stat == SIDL) + proc_t p; + ALLPROC_FOREACH(p) { + /* ignore processes that are being forked */ + if (p->p_stat == SIDL) { continue; - if ( (filterfn == 0 ) || (filterfn(p, filterarg) != 0)) { - pid_list[pidcount] = p->p_pid; - pidcount++; - if (pidcount >= count) - break; } - } - } - if ((pidcount < count ) && (flags & PROC_ZOMBPROCLIST)) { - for (p = zombproc.lh_first; p != 0; p = p->p_list.le_next) { - if ( (filterfn == 0 ) || (filterfn(p, filterarg) != 0)) { - pid_list[pidcount] = p->p_pid; - pidcount++; - if (pidcount >= count) - break; + if ((filterfn != NULL) && (filterfn(p, filterarg) == 0)) { + continue; + } + pidlist_add_pid(pl, proc_pid(p)); + if (++pid_count >= pid_count_available) { + break; } } } - - - proc_list_unlock(); - - for (i = 0; i< pidcount; i++) { - p = proc_find(pid_list[i]); - if (p) { - if ((flags & PROC_NOWAITTRANS) == 0) - proc_transwait(p, 0); - retval = callout(p, arg); - - switch (retval) { - case PROC_RETURNED: - case PROC_RETURNED_DONE: - proc_rele(p); - if (retval == PROC_RETURNED_DONE) { - goto out; - } - break; - - case PROC_CLAIMED_DONE: - goto out; - case PROC_CLAIMED: - default: - break; + if ((pid_count < pid_count_available) && + (flags & PROC_ZOMBPROCLIST)) { + proc_t p; + ZOMBPROC_FOREACH(p) { + if ((filterfn != NULL) && (filterfn(p, filterarg) == 0)) { + continue; } - } else if (flags & PROC_ZOMBPROCLIST) { - p = proc_find_zombref(pid_list[i]); - if (p != PROC_NULL) { - retval = callout(p, arg); - - switch (retval) { - case PROC_RETURNED: - case PROC_RETURNED_DONE: - proc_drop_zombref(p); - if (retval == PROC_RETURNED_DONE) { - goto out; - } - break; - - case PROC_CLAIMED_DONE: - goto out; - case PROC_CLAIMED: - default: - break; - } + pidlist_add_pid(pl, proc_pid(p)); + if (++pid_count >= pid_count_available) { + break; } } } -out: - kfree(pid_list, alloc_count); - return(0); - -} + proc_list_unlock(); + /* call callout on processes in the pid_list */ -#if 0 -/* This is for iteration in case of trivial non blocking callouts */ -int -proc_scanall(flags, callout, arg) - int flags; - int (*callout)(proc_t, void *); - void * arg; -{ - proc_t p; - int retval; + const pidlist_entry_t *pe; + SLIST_FOREACH(pe, &(pl->pl_head), pe_link) { + for (u_int i = 0; i < pe->pe_nused; i++) { + const pid_t pid = pe->pe_pid[i]; + proc_t p = proc_find(pid); + if (p) { + if ((flags & PROC_NOWAITTRANS) == 0) { + proc_transwait(p, 0); + } + const int callout_ret = callout(p, arg); + switch (callout_ret) { + case PROC_RETURNED_DONE: + proc_rele(p); + /* FALLTHROUGH */ + case PROC_CLAIMED_DONE: + goto out; - proc_list_lock(); + case PROC_RETURNED: + proc_rele(p); + /* FALLTHROUGH */ + case PROC_CLAIMED: + break; + default: + panic("%s: callout =%d for pid %d", + __func__, callout_ret, pid); + break; + } + } else if (flags & PROC_ZOMBPROCLIST) { + p = proc_find_zombref(pid); + if (!p) { + continue; + } + const int callout_ret = callout(p, arg); + switch (callout_ret) { + case PROC_RETURNED_DONE: + proc_drop_zombref(p); + /* FALLTHROUGH */ + case PROC_CLAIMED_DONE: + goto out; - if (flags & PROC_ALLPROCLIST) { - for (p = allproc.lh_first; (p != 0); p = p->p_list.le_next) { - retval = callout(p, arg); - if (retval == PROC_RETURNED_DONE) - goto out; - } - } - if (flags & PROC_ZOMBPROCLIST) { - for (p = zombproc.lh_first; p != 0; p = p->p_list.le_next) { - retval = callout(p, arg); - if (retval == PROC_RETURNED_DONE) - goto out; + case PROC_RETURNED: + proc_drop_zombref(p); + /* FALLTHROUGH */ + case PROC_CLAIMED: + break; + default: + panic("%s: callout =%d for zombie %d", + __func__, callout_ret, pid); + break; + } + } } } out: - - proc_list_unlock(); - - return(0); + pidlist_free(pl); } -#endif - -int -proc_rebootscan(callout, arg, filterfn, filterarg) - int (*callout)(proc_t, void *); - void * arg; - int (*filterfn)(proc_t, void *); - void * filterarg; +void +proc_rebootscan( + proc_iterate_fn_t callout, + void *arg, + proc_iterate_fn_t filterfn, + void *filterarg) { proc_t p; - int lockheld = 0, retval; + + assert(callout != NULL); proc_shutdown_exitcount = 0; -ps_allprocscan: +restart_foreach: proc_list_lock(); - lockheld = 1; - - for (p = allproc.lh_first; (p != 0); p = p->p_list.le_next) { - if ( (filterfn == 0 ) || (filterfn(p, filterarg) != 0)) { - p = proc_ref_locked(p); + ALLPROC_FOREACH(p) { + if ((filterfn != NULL) && filterfn(p, filterarg) == 0) { + continue; + } + p = proc_ref_locked(p); + if (!p) { + continue; + } - proc_list_unlock(); - lockheld = 0; + proc_list_unlock(); - if (p) { - proc_transwait(p, 0); - retval = callout(p, arg); - proc_rele(p); - - switch (retval) { - case PROC_RETURNED_DONE: - case PROC_CLAIMED_DONE: - goto out; - } - } - goto ps_allprocscan; - } /* filter pass */ - } /* allproc walk thru */ + proc_transwait(p, 0); + (void)callout(p, arg); + proc_rele(p); - if (lockheld == 1) { - proc_list_unlock(); - lockheld = 0; + goto restart_foreach; } -out: - return(0); - + proc_list_unlock(); } - -int -proc_childrenwalk(parent, callout, arg) - struct proc * parent; - int (*callout)(proc_t, void *); - void * arg; +void +proc_childrenwalk( + proc_t parent, + proc_iterate_fn_t callout, + void *arg) { - register struct proc *p; - pid_t * pid_list; - int count, pidcount, alloc_count, i, retval; - - count = nprocs+ 10; - if (count > hard_maxproc) - count = hard_maxproc; - alloc_count = count * sizeof(pid_t); - pid_list = (pid_t *)kalloc(alloc_count); - bzero(pid_list, alloc_count); + pidlist_t pid_list, *pl = pidlist_init(&pid_list); + u_int pid_count_available = 0; + assert(parent != NULL); + assert(callout != NULL); - proc_list_lock(); + for (;;) { + proc_list_lock(); + pid_count_available = parent->p_childrencnt; + if (pid_count_available == 0) { + proc_list_unlock(); + goto out; + } + if (pidlist_nalloc(pl) > pid_count_available) { + break; + } + proc_list_unlock(); + pidlist_alloc(pl, pid_count_available); + } + pidlist_set_active(pl); - pidcount = 0; - for (p = parent->p_children.lh_first; (p != 0); p = p->p_sibling.le_next) { - if (p->p_stat == SIDL) + u_int pid_count = 0; + proc_t p; + PCHILDREN_FOREACH(parent, p) { + if (p->p_stat == SIDL) { continue; - pid_list[pidcount] = p->p_pid; - pidcount++; - if (pidcount >= count) + } + pidlist_add_pid(pl, proc_pid(p)); + if (++pid_count >= pid_count_available) { break; + } } + proc_list_unlock(); + const pidlist_entry_t *pe; + SLIST_FOREACH(pe, &(pl->pl_head), pe_link) { + for (u_int i = 0; i < pe->pe_nused; i++) { + const pid_t pid = pe->pe_pid[i]; + p = proc_find(pid); + if (!p) { + continue; + } + const int callout_ret = callout(p, arg); - for (i = 0; i< pidcount; i++) { - p = proc_find(pid_list[i]); - if (p) { - proc_transwait(p, 0); - retval = callout(p, arg); - - switch (retval) { - case PROC_RETURNED: - case PROC_RETURNED_DONE: - proc_rele(p); - if (retval == PROC_RETURNED_DONE) { - goto out; - } - break; - - case PROC_CLAIMED_DONE: - goto out; - case PROC_CLAIMED: - default: - break; + switch (callout_ret) { + case PROC_RETURNED_DONE: + proc_rele(p); + /* FALLTHROUGH */ + case PROC_CLAIMED_DONE: + goto out; + + case PROC_RETURNED: + proc_rele(p); + /* FALLTHROUGH */ + case PROC_CLAIMED: + break; + default: + panic("%s: callout =%d for pid %d", + __func__, callout_ret, pid); + break; } } } - -out: - kfree(pid_list, alloc_count); - return(0); - +out: + pidlist_free(pl); } -/* - */ -/* PGRP_BLOCKITERATE is not implemented yet */ -int -pgrp_iterate(pgrp, flags, callout, arg, filterfn, filterarg) - struct pgrp *pgrp; - int flags; - int (*callout)(proc_t, void *); - void * arg; - int (*filterfn)(proc_t, void *); - void * filterarg; +void +pgrp_iterate( + struct pgrp *pgrp, + unsigned int flags, + proc_iterate_fn_t callout, + void * arg, + proc_iterate_fn_t filterfn, + void * filterarg) { - proc_t p; - pid_t * pid_list; - int count, pidcount, i, alloc_count; - int retval; - pid_t pgid; - int dropref = flags & PGRP_DROPREF; -#if 0 - int serialize = flags & PGRP_BLOCKITERATE; -#else - int serialize = 0; -#endif + pidlist_t pid_list, *pl = pidlist_init(&pid_list); + u_int pid_count_available = 0; - if (pgrp == 0) - return(0); - count = pgrp->pg_membercnt + 10; - if (count > hard_maxproc) - count = hard_maxproc; - alloc_count = count * sizeof(pid_t); - pid_list = (pid_t *)kalloc(alloc_count); - bzero(pid_list, alloc_count); - - pgrp_lock(pgrp); - if (serialize != 0) { - while ((pgrp->pg_listflags & PGRP_FLAG_ITERABEGIN) == PGRP_FLAG_ITERABEGIN) { - pgrp->pg_listflags |= PGRP_FLAG_ITERWAIT; - msleep(&pgrp->pg_listflags, &pgrp->pg_mlock, 0, "pgrp_iterate", 0); + assert(pgrp != NULL); + assert(callout != NULL); + + for (;;) { + pgrp_lock(pgrp); + pid_count_available = pgrp->pg_membercnt; + if (pid_count_available == 0) { + pgrp_unlock(pgrp); + if (flags & PGRP_DROPREF) { + pg_rele(pgrp); + } + goto out; } - pgrp->pg_listflags |= PGRP_FLAG_ITERABEGIN; - } + if (pidlist_nalloc(pl) > pid_count_available) { + break; + } + pgrp_unlock(pgrp); - pgid = pgrp->pg_id; + pidlist_alloc(pl, pid_count_available); + } + pidlist_set_active(pl); - pidcount = 0; - for (p = pgrp->pg_members.lh_first; p != 0; - p = p->p_pglist.le_next) { - if ( (filterfn == 0 ) || (filterfn(p, filterarg) != 0)) { - pid_list[pidcount] = p->p_pid; - pidcount++; - if (pidcount >= count) - break; + const pid_t pgid = pgrp->pg_id; + u_int pid_count = 0; + proc_t p; + PGMEMBERS_FOREACH(pgrp, p) { + if ((filterfn != NULL) && (filterfn(p, filterarg) == 0)) { + continue;; + } + pidlist_add_pid(pl, proc_pid(p)); + if (++pid_count >= pid_count_available) { + break; } } - pgrp_unlock(pgrp); - if ((serialize == 0) && (dropref != 0)) - pg_rele(pgrp); + if (flags & PGRP_DROPREF) { + pg_rele(pgrp); + } - for (i = 0; i< pidcount; i++) { - /* No handling or proc0 */ - if (pid_list[i] == 0) - continue; - p = proc_find(pid_list[i]); - if (p) { + const pidlist_entry_t *pe; + SLIST_FOREACH(pe, &(pl->pl_head), pe_link) { + for (u_int i = 0; i < pe->pe_nused; i++) { + const pid_t pid = pe->pe_pid[i]; + if (0 == pid) { + continue; /* skip kernproc */ + } + p = proc_find(pid); + if (!p) { + continue; + } if (p->p_pgrpid != pgid) { proc_rele(p); continue; } - proc_transwait(p, 0); - retval = callout(p, arg); - - switch (retval) { - case PROC_RETURNED: - case PROC_RETURNED_DONE: - proc_rele(p); - if (retval == PROC_RETURNED_DONE) { - goto out; - } - break; - - case PROC_CLAIMED_DONE: - goto out; - case PROC_CLAIMED: - default: - break; + const int callout_ret = callout(p, arg); + + switch (callout_ret) { + case PROC_RETURNED: + proc_rele(p); + /* FALLTHROUGH */ + case PROC_CLAIMED: + break; + case PROC_RETURNED_DONE: + proc_rele(p); + /* FALLTHROUGH */ + case PROC_CLAIMED_DONE: + goto out; + + default: + panic("%s: callout =%d for pid %d", + __func__, callout_ret, pid); } } } + out: - if (serialize != 0) { - pgrp_lock(pgrp); - pgrp->pg_listflags &= ~PGRP_FLAG_ITERABEGIN; - if ((pgrp->pg_listflags & PGRP_FLAG_ITERWAIT) == PGRP_FLAG_ITERWAIT) { - pgrp->pg_listflags &= ~PGRP_FLAG_ITERWAIT; - wakeup(&pgrp->pg_listflags); - } - pgrp_unlock(pgrp); - if (dropref != 0) - pg_rele(pgrp); - } - kfree(pid_list, alloc_count); - return(0); + pidlist_free(pl); } static void @@ -2322,31 +2958,32 @@ pgrp_add(struct pgrp * pgrp, struct proc * parent, struct proc * child) child->p_pgrpid = pgrp->pg_id; child->p_listflag |= P_LIST_INPGRP; /* - * When pgrp is being freed , a process can still - * request addition using setpgid from bash when - * login is terminated (login cycler) return ESRCH - * Safe to hold lock due to refcount on pgrp + * When pgrp is being freed , a process can still + * request addition using setpgid from bash when + * login is terminated (login cycler) return ESRCH + * Safe to hold lock due to refcount on pgrp */ if ((pgrp->pg_listflags & (PGRP_FLAG_TERMINATE | PGRP_FLAG_DEAD)) == PGRP_FLAG_TERMINATE) { - pgrp->pg_listflags &= ~PGRP_FLAG_TERMINATE; + pgrp->pg_listflags &= ~PGRP_FLAG_TERMINATE; } - if ((pgrp->pg_listflags & PGRP_FLAG_DEAD) == PGRP_FLAG_DEAD) + if ((pgrp->pg_listflags & PGRP_FLAG_DEAD) == PGRP_FLAG_DEAD) { panic("pgrp_add : pgrp is dead adding process"); + } proc_list_unlock(); pgrp_lock(pgrp); pgrp->pg_membercnt++; - if ( parent != PROC_NULL) { + if (parent != PROC_NULL) { LIST_INSERT_AFTER(parent, child, p_pglist); - }else { + } else { LIST_INSERT_HEAD(&pgrp->pg_members, child, p_pglist); } pgrp_unlock(pgrp); proc_list_lock(); if (((pgrp->pg_listflags & (PGRP_FLAG_TERMINATE | PGRP_FLAG_DEAD)) == PGRP_FLAG_TERMINATE) && (pgrp->pg_membercnt != 0)) { - pgrp->pg_listflags &= ~PGRP_FLAG_TERMINATE; + pgrp->pg_listflags &= ~PGRP_FLAG_TERMINATE; } proc_list_unlock(); } @@ -2360,21 +2997,24 @@ pgrp_remove(struct proc * p) proc_list_lock(); #if __PROC_INTERNAL_DEBUG - if ((p->p_listflag & P_LIST_INPGRP) == 0) + if ((p->p_listflag & P_LIST_INPGRP) == 0) { panic("removing from pglist but no named ref\n"); + } #endif p->p_pgrpid = PGRPID_DEAD; p->p_listflag &= ~P_LIST_INPGRP; p->p_pgrp = NULL; proc_list_unlock(); - if (pg == PGRP_NULL) + if (pg == PGRP_NULL) { panic("pgrp_remove: pg is NULL"); + } pgrp_lock(pg); pg->pg_membercnt--; - if (pg->pg_membercnt < 0) - panic("pgprp: -ve membercnt pgprp:%p p:%p\n",pg, p); + if (pg->pg_membercnt < 0) { + panic("pgprp: -ve membercnt pgprp:%p p:%p\n", pg, p); + } LIST_REMOVE(p, p_pglist); if (pg->pg_members.lh_first == 0) { @@ -2391,11 +3031,11 @@ pgrp_remove(struct proc * p) static void pgrp_replace(struct proc * p, struct pgrp * newpg) { - struct pgrp * oldpg; + struct pgrp * oldpg; - proc_list_lock(); + proc_list_lock(); while ((p->p_listflag & P_LIST_PGRPTRANS) == P_LIST_PGRPTRANS) { p->p_listflag |= P_LIST_PGRPTRWAIT; @@ -2405,67 +3045,70 @@ pgrp_replace(struct proc * p, struct pgrp * newpg) p->p_listflag |= P_LIST_PGRPTRANS; oldpg = p->p_pgrp; - if (oldpg == PGRP_NULL) + if (oldpg == PGRP_NULL) { panic("pgrp_replace: oldpg NULL"); + } oldpg->pg_refcount++; #if __PROC_INTERNAL_DEBUG - if ((p->p_listflag & P_LIST_INPGRP) == 0) - panic("removing from pglist but no named ref\n"); + if ((p->p_listflag & P_LIST_INPGRP) == 0) { + panic("removing from pglist but no named ref\n"); + } #endif - p->p_pgrpid = PGRPID_DEAD; - p->p_listflag &= ~P_LIST_INPGRP; - p->p_pgrp = NULL; - - proc_list_unlock(); - - pgrp_lock(oldpg); - oldpg->pg_membercnt--; - if (oldpg->pg_membercnt < 0) - panic("pgprp: -ve membercnt pgprp:%p p:%p\n",oldpg, p); - LIST_REMOVE(p, p_pglist); - if (oldpg->pg_members.lh_first == 0) { - pgrp_unlock(oldpg); - pgdelete_dropref(oldpg); - } else { - pgrp_unlock(oldpg); - pg_rele(oldpg); - } - - proc_list_lock(); - p->p_pgrp = newpg; - p->p_pgrpid = newpg->pg_id; - p->p_listflag |= P_LIST_INPGRP; - /* - * When pgrp is being freed , a process can still - * request addition using setpgid from bash when - * login is terminated (login cycler) return ESRCH - * Safe to hold lock due to refcount on pgrp - */ - if ((newpg->pg_listflags & (PGRP_FLAG_TERMINATE | PGRP_FLAG_DEAD)) == PGRP_FLAG_TERMINATE) { - newpg->pg_listflags &= ~PGRP_FLAG_TERMINATE; - } - - if ((newpg->pg_listflags & PGRP_FLAG_DEAD) == PGRP_FLAG_DEAD) - panic("pgrp_add : pgrp is dead adding process"); - proc_list_unlock(); - - pgrp_lock(newpg); - newpg->pg_membercnt++; + p->p_pgrpid = PGRPID_DEAD; + p->p_listflag &= ~P_LIST_INPGRP; + p->p_pgrp = NULL; + + proc_list_unlock(); + + pgrp_lock(oldpg); + oldpg->pg_membercnt--; + if (oldpg->pg_membercnt < 0) { + panic("pgprp: -ve membercnt pgprp:%p p:%p\n", oldpg, p); + } + LIST_REMOVE(p, p_pglist); + if (oldpg->pg_members.lh_first == 0) { + pgrp_unlock(oldpg); + pgdelete_dropref(oldpg); + } else { + pgrp_unlock(oldpg); + pg_rele(oldpg); + } + + proc_list_lock(); + p->p_pgrp = newpg; + p->p_pgrpid = newpg->pg_id; + p->p_listflag |= P_LIST_INPGRP; + /* + * When pgrp is being freed , a process can still + * request addition using setpgid from bash when + * login is terminated (login cycler) return ESRCH + * Safe to hold lock due to refcount on pgrp + */ + if ((newpg->pg_listflags & (PGRP_FLAG_TERMINATE | PGRP_FLAG_DEAD)) == PGRP_FLAG_TERMINATE) { + newpg->pg_listflags &= ~PGRP_FLAG_TERMINATE; + } + + if ((newpg->pg_listflags & PGRP_FLAG_DEAD) == PGRP_FLAG_DEAD) { + panic("pgrp_add : pgrp is dead adding process"); + } + proc_list_unlock(); + + pgrp_lock(newpg); + newpg->pg_membercnt++; LIST_INSERT_HEAD(&newpg->pg_members, p, p_pglist); - pgrp_unlock(newpg); + pgrp_unlock(newpg); - proc_list_lock(); - if (((newpg->pg_listflags & (PGRP_FLAG_TERMINATE | PGRP_FLAG_DEAD)) == PGRP_FLAG_TERMINATE) && (newpg->pg_membercnt != 0)) { - newpg->pg_listflags &= ~PGRP_FLAG_TERMINATE; - } + proc_list_lock(); + if (((newpg->pg_listflags & (PGRP_FLAG_TERMINATE | PGRP_FLAG_DEAD)) == PGRP_FLAG_TERMINATE) && (newpg->pg_membercnt != 0)) { + newpg->pg_listflags &= ~PGRP_FLAG_TERMINATE; + } p->p_listflag &= ~P_LIST_PGRPTRANS; if ((p->p_listflag & P_LIST_PGRPTRWAIT) == P_LIST_PGRPTRWAIT) { p->p_listflag &= ~P_LIST_PGRPTRWAIT; wakeup(&p->p_pgrpid); - } - proc_list_unlock(); + proc_list_unlock(); } void @@ -2498,28 +3141,30 @@ proc_pgrp(proc_t p) { struct pgrp * pgrp; - if (p == PROC_NULL) - return(PGRP_NULL); + if (p == PROC_NULL) { + return PGRP_NULL; + } proc_list_lock(); while ((p->p_listflag & P_LIST_PGRPTRANS) == P_LIST_PGRPTRANS) { p->p_listflag |= P_LIST_PGRPTRWAIT; (void)msleep(&p->p_pgrpid, proc_list_mlock, 0, "proc_pgrp", 0); } - + pgrp = p->p_pgrp; assert(pgrp != NULL); if (pgrp != PGRP_NULL) { pgrp->pg_refcount++; - if ((pgrp->pg_listflags & (PGRP_FLAG_TERMINATE | PGRP_FLAG_DEAD)) != 0) + if ((pgrp->pg_listflags & (PGRP_FLAG_TERMINATE | PGRP_FLAG_DEAD)) != 0) { panic("proc_pgrp: ref being povided for dead pgrp"); + } } - + proc_list_unlock(); - - return(pgrp); + + return pgrp; } struct pgrp * @@ -2531,22 +3176,24 @@ tty_pgrp(struct tty * tp) pg = tp->t_pgrp; if (pg != PGRP_NULL) { - if ((pg->pg_listflags & PGRP_FLAG_DEAD) != 0) + if ((pg->pg_listflags & PGRP_FLAG_DEAD) != 0) { panic("tty_pgrp: ref being povided for dead pgrp"); + } pg->pg_refcount++; } proc_list_unlock(); - return(pg); + return pg; } struct session * proc_session(proc_t p) { struct session * sess = SESSION_NULL; - - if (p == PROC_NULL) - return(SESSION_NULL); + + if (p == PROC_NULL) { + return SESSION_NULL; + } proc_list_lock(); @@ -2557,12 +3204,13 @@ proc_session(proc_t p) } if ((p->p_pgrp != PGRP_NULL) && ((sess = p->p_pgrp->pg_session) != SESSION_NULL)) { - if ((sess->s_listflags & (S_LIST_TERM | S_LIST_DEAD)) != 0) + if ((sess->s_listflags & (S_LIST_TERM | S_LIST_DEAD)) != 0) { panic("proc_session:returning sesssion ref on terminating session"); + } sess->s_count++; } proc_list_unlock(); - return(sess); + return sess; } void @@ -2570,33 +3218,34 @@ session_rele(struct session *sess) { proc_list_lock(); if (--sess->s_count == 0) { - if ((sess->s_listflags & (S_LIST_TERM | S_LIST_DEAD)) != 0) + if ((sess->s_listflags & (S_LIST_TERM | S_LIST_DEAD)) != 0) { panic("session_rele: terminating already terminated session"); + } sess->s_listflags |= S_LIST_TERM; LIST_REMOVE(sess, s_hash); sess->s_listflags |= S_LIST_DEAD; - if (sess->s_count != 0) - panic("session_rele: freeing session in use"); + if (sess->s_count != 0) { + panic("session_rele: freeing session in use"); + } proc_list_unlock(); -#if CONFIG_FINE_LOCK_GROUPS lck_mtx_destroy(&sess->s_mlock, proc_mlock_grp); -#else - lck_mtx_destroy(&sess->s_mlock, proc_lck_grp); -#endif FREE_ZONE(sess, sizeof(struct session), M_SESSION); - } else + } else { proc_list_unlock(); + } } int -proc_transstart(proc_t p, int locked) +proc_transstart(proc_t p, int locked, int non_blocking) { - if (locked == 0) + if (locked == 0) { proc_lock(p); + } while ((p->p_lflag & P_LINTRANSIT) == P_LINTRANSIT) { - if ((p->p_lflag & P_LTRANSCOMMIT) == P_LTRANSCOMMIT) { - if (locked == 0) + if (((p->p_lflag & P_LTRANSCOMMIT) == P_LTRANSCOMMIT) || non_blocking) { + if (locked == 0) { proc_unlock(p); + } return EDEADLK; } p->p_lflag |= P_LTRANSWAIT; @@ -2604,62 +3253,70 @@ proc_transstart(proc_t p, int locked) } p->p_lflag |= P_LINTRANSIT; p->p_transholder = current_thread(); - if (locked == 0) + if (locked == 0) { proc_unlock(p); + } return 0; } void proc_transcommit(proc_t p, int locked) { - if (locked == 0) + if (locked == 0) { proc_lock(p); + } - assert ((p->p_lflag & P_LINTRANSIT) == P_LINTRANSIT); - assert (p->p_transholder == current_thread()); + assert((p->p_lflag & P_LINTRANSIT) == P_LINTRANSIT); + assert(p->p_transholder == current_thread()); p->p_lflag |= P_LTRANSCOMMIT; if ((p->p_lflag & P_LTRANSWAIT) == P_LTRANSWAIT) { p->p_lflag &= ~P_LTRANSWAIT; wakeup(&p->p_lflag); } - if (locked == 0) + if (locked == 0) { proc_unlock(p); + } } void proc_transend(proc_t p, int locked) { - if (locked == 0) + if (locked == 0) { proc_lock(p); + } - p->p_lflag &= ~( P_LINTRANSIT | P_LTRANSCOMMIT); + p->p_lflag &= ~(P_LINTRANSIT | P_LTRANSCOMMIT); p->p_transholder = NULL; if ((p->p_lflag & P_LTRANSWAIT) == P_LTRANSWAIT) { p->p_lflag &= ~P_LTRANSWAIT; wakeup(&p->p_lflag); } - if (locked == 0) + if (locked == 0) { proc_unlock(p); + } } int proc_transwait(proc_t p, int locked) { - if (locked == 0) + if (locked == 0) { proc_lock(p); + } while ((p->p_lflag & P_LINTRANSIT) == P_LINTRANSIT) { if ((p->p_lflag & P_LTRANSCOMMIT) == P_LTRANSCOMMIT && current_proc() == p) { - if (locked == 0) + if (locked == 0) { proc_unlock(p); + } return EDEADLK; } p->p_lflag |= P_LTRANSWAIT; msleep(&p->p_lflag, &p->p_mlock, 0, "proc_signstart", NULL); } - if (locked == 0) + if (locked == 0) { proc_unlock(p); + } return 0; } @@ -2693,114 +3350,13 @@ proc_knote_drain(struct proc *p) */ proc_klist_lock(); while ((kn = SLIST_FIRST(&p->p_klist))) { - kn->kn_ptr.p_proc = PROC_NULL; + kn->kn_proc = PROC_NULL; KNOTE_DETACH(&p->p_klist, kn); } proc_klist_unlock(); } -unsigned long cs_procs_killed = 0; -unsigned long cs_procs_invalidated = 0; -int cs_force_kill = 0; -int cs_force_hard = 0; -int cs_debug = 0; -SYSCTL_INT(_vm, OID_AUTO, cs_force_kill, CTLFLAG_RW | CTLFLAG_LOCKED, &cs_force_kill, 0, ""); -SYSCTL_INT(_vm, OID_AUTO, cs_force_hard, CTLFLAG_RW | CTLFLAG_LOCKED, &cs_force_hard, 0, ""); -SYSCTL_INT(_vm, OID_AUTO, cs_debug, CTLFLAG_RW | CTLFLAG_LOCKED, &cs_debug, 0, ""); - -int -cs_allow_invalid(struct proc *p) -{ -#if MACH_ASSERT - lck_mtx_assert(&p->p_mlock, LCK_MTX_ASSERT_NOTOWNED); -#endif -#if CONFIG_MACF && CONFIG_ENFORCE_SIGNED_CODE - /* There needs to be a MAC policy to implement this hook, or else the - * kill bits will be cleared here every time. If we have - * CONFIG_ENFORCE_SIGNED_CODE, we can assume there is a policy - * implementing the hook. - */ - if( 0 != mac_proc_check_run_cs_invalid(p)) { - if(cs_debug) printf("CODE SIGNING: cs_allow_invalid() " - "not allowed: pid %d\n", - p->p_pid); - return 0; - } - if(cs_debug) printf("CODE SIGNING: cs_allow_invalid() " - "allowed: pid %d\n", - p->p_pid); - proc_lock(p); - p->p_csflags &= ~(CS_KILL | CS_HARD | CS_VALID); - proc_unlock(p); - vm_map_switch_protect(get_task_map(p->task), FALSE); -#endif - return (p->p_csflags & (CS_KILL | CS_HARD)) == 0; -} - -int -cs_invalid_page( - addr64_t vaddr) -{ - struct proc *p; - int retval; - - p = current_proc(); - - /* - * XXX revisit locking when proc is no longer protected - * by the kernel funnel... - */ - - /* XXX for testing */ - proc_lock(p); - if (cs_force_kill) - p->p_csflags |= CS_KILL; - if (cs_force_hard) - p->p_csflags |= CS_HARD; - - /* CS_KILL triggers us to send a kill signal. Nothing else. */ - if (p->p_csflags & CS_KILL) { - p->p_csflags |= CS_KILLED; - proc_unlock(p); - if (cs_debug) { - printf("CODE SIGNING: cs_invalid_page(0x%llx): " - "p=%d[%s] honoring CS_KILL, final status 0x%x\n", - vaddr, p->p_pid, p->p_comm, p->p_csflags); - } - cs_procs_killed++; - psignal(p, SIGKILL); - proc_lock(p); - } - - /* CS_HARD means fail the mapping operation so the process stays valid. */ - if (p->p_csflags & CS_HARD) { - proc_unlock(p); - if (cs_debug) { - printf("CODE SIGNING: cs_invalid_page(0x%llx): " - "p=%d[%s] honoring CS_HARD\n", - vaddr, p->p_pid, p->p_comm); - } - retval = 1; - } else { - if (p->p_csflags & CS_VALID) { - p->p_csflags &= ~CS_VALID; - - proc_unlock(p); - cs_procs_invalidated++; - printf("CODE SIGNING: cs_invalid_page(0x%llx): " - "p=%d[%s] clearing CS_VALID\n", - vaddr, p->p_pid, p->p_comm); - } else { - proc_unlock(p); - } - - retval = 0; - } - - return retval; -} - -void +void proc_setregister(proc_t p) { proc_lock(p); @@ -2808,7 +3364,7 @@ proc_setregister(proc_t p) proc_unlock(p); } -void +void proc_resetregister(proc_t p) { proc_lock(p); @@ -2822,6 +3378,20 @@ proc_pgrpid(proc_t p) return p->p_pgrpid; } +pid_t +proc_sessionid(proc_t p) +{ + pid_t sid = -1; + struct session * sessp = proc_session(p); + + if (sessp != SESSION_NULL) { + sid = sessp->s_sid; + session_rele(sessp); + } + + return sid; +} + pid_t proc_selfpgrpid() { @@ -2836,63 +3406,64 @@ proc_getpcontrol(int pid, int * pcontrolp) proc_t p; p = proc_find(pid); - if (p == PROC_NULL) - return(ESRCH); - if (pcontrolp != NULL) + if (p == PROC_NULL) { + return ESRCH; + } + if (pcontrolp != NULL) { *pcontrolp = p->p_pcaction; + } proc_rele(p); - return(0); + return 0; } int -proc_dopcontrol(proc_t p, void *num_found) +proc_dopcontrol(proc_t p) { int pcontrol; + os_reason_t kill_reason; proc_lock(p); pcontrol = PROC_CONTROL_STATE(p); - if (PROC_ACTION_STATE(p) ==0) { - switch(pcontrol) { - case P_PCTHROTTLE: - PROC_SETACTION_STATE(p); - proc_unlock(p); - printf("low swap: throttling pid %d (%s)\n", p->p_pid, p->p_comm); - (*(int *)num_found)++; - break; + if (PROC_ACTION_STATE(p) == 0) { + switch (pcontrol) { + case P_PCTHROTTLE: + PROC_SETACTION_STATE(p); + proc_unlock(p); + printf("low swap: throttling pid %d (%s)\n", p->p_pid, p->p_comm); + break; - case P_PCSUSP: - PROC_SETACTION_STATE(p); - proc_unlock(p); - printf("low swap: suspending pid %d (%s)\n", p->p_pid, p->p_comm); - task_suspend(p->task); - (*(int *)num_found)++; - break; + case P_PCSUSP: + PROC_SETACTION_STATE(p); + proc_unlock(p); + printf("low swap: suspending pid %d (%s)\n", p->p_pid, p->p_comm); + task_suspend(p->task); + break; - case P_PCKILL: - PROC_SETACTION_STATE(p); - proc_unlock(p); - printf("low swap: killing pid %d (%s)\n", p->p_pid, p->p_comm); - psignal(p, SIGKILL); - (*(int *)num_found)++; - break; + case P_PCKILL: + PROC_SETACTION_STATE(p); + proc_unlock(p); + printf("low swap: killing pid %d (%s)\n", p->p_pid, p->p_comm); + kill_reason = os_reason_create(OS_REASON_JETSAM, JETSAM_REASON_LOWSWAP); + psignal_with_reason(p, SIGKILL, kill_reason); + break; - default: - proc_unlock(p); + default: + proc_unlock(p); } - - } else + } else { proc_unlock(p); + } - return(PROC_RETURNED); + return PROC_RETURNED; } /* * Resume a throttled or suspended process. This is an internal interface that's only - * used by the user level code that presents the GUI when we run out of swap space and + * used by the user level code that presents the GUI when we run out of swap space and * hence is restricted to processes with superuser privileges. */ @@ -2905,142 +3476,466 @@ proc_resetpcontrol(int pid) proc_t self = current_proc(); /* if the process has been validated to handle resource control or root is valid one */ - if (((self->p_lflag & P_LVMRSRCOWNER) == 0) && (error = suser(kauth_cred_get(), 0))) + if (((self->p_lflag & P_LVMRSRCOWNER) == 0) && (error = suser(kauth_cred_get(), 0))) { return error; + } p = proc_find(pid); - if (p == PROC_NULL) - return(ESRCH); - + if (p == PROC_NULL) { + return ESRCH; + } + proc_lock(p); pcontrol = PROC_CONTROL_STATE(p); - if(PROC_ACTION_STATE(p) !=0) { - switch(pcontrol) { - case P_PCTHROTTLE: - PROC_RESETACTION_STATE(p); - proc_unlock(p); - printf("low swap: unthrottling pid %d (%s)\n", p->p_pid, p->p_comm); - break; + if (PROC_ACTION_STATE(p) != 0) { + switch (pcontrol) { + case P_PCTHROTTLE: + PROC_RESETACTION_STATE(p); + proc_unlock(p); + printf("low swap: unthrottling pid %d (%s)\n", p->p_pid, p->p_comm); + break; - case P_PCSUSP: - PROC_RESETACTION_STATE(p); - proc_unlock(p); - printf("low swap: resuming pid %d (%s)\n", p->p_pid, p->p_comm); - task_resume(p->task); - break; + case P_PCSUSP: + PROC_RESETACTION_STATE(p); + proc_unlock(p); + printf("low swap: resuming pid %d (%s)\n", p->p_pid, p->p_comm); + task_resume(p->task); + break; - case P_PCKILL: - /* Huh? */ - PROC_SETACTION_STATE(p); - proc_unlock(p); - printf("low swap: attempt to unkill pid %d (%s) ignored\n", p->p_pid, p->p_comm); - break; + case P_PCKILL: + /* Huh? */ + PROC_SETACTION_STATE(p); + proc_unlock(p); + printf("low swap: attempt to unkill pid %d (%s) ignored\n", p->p_pid, p->p_comm); + break; - default: - proc_unlock(p); + default: + proc_unlock(p); } - - } else + } else { proc_unlock(p); + } proc_rele(p); - return(0); + return 0; } -/* - * Return true if the specified process has an action state specified for it and it isn't - * already in an action state and it's using more physical memory than the specified threshold. - * Note: the memory_threshold argument is specified in bytes and is of type uint64_t. - */ + +struct no_paging_space { + uint64_t pcs_max_size; + uint64_t pcs_uniqueid; + int pcs_pid; + int pcs_proc_count; + uint64_t pcs_total_size; + + uint64_t npcs_max_size; + uint64_t npcs_uniqueid; + int npcs_pid; + int npcs_proc_count; + uint64_t npcs_total_size; + + int apcs_proc_count; + uint64_t apcs_total_size; +}; + static int -proc_pcontrol_filter(proc_t p, void *memory_thresholdp) +proc_pcontrol_filter(proc_t p, void *arg) { - - return PROC_CONTROL_STATE(p) && /* if there's an action state specified... */ - (PROC_ACTION_STATE(p) == 0) && /* and we're not in the action state yet... */ - (get_task_resident_size(p->task) > *((uint64_t *)memory_thresholdp)); /* and this proc is over the mem threshold, */ - /* then return true to take action on this proc */ + struct no_paging_space *nps; + uint64_t compressed; + + nps = (struct no_paging_space *)arg; + + compressed = get_task_compressed(p->task); + + if (PROC_CONTROL_STATE(p)) { + if (PROC_ACTION_STATE(p) == 0) { + if (compressed > nps->pcs_max_size) { + nps->pcs_pid = p->p_pid; + nps->pcs_uniqueid = p->p_uniqueid; + nps->pcs_max_size = compressed; + } + nps->pcs_total_size += compressed; + nps->pcs_proc_count++; + } else { + nps->apcs_total_size += compressed; + nps->apcs_proc_count++; + } + } else { + if (compressed > nps->npcs_max_size) { + nps->npcs_pid = p->p_pid; + nps->npcs_uniqueid = p->p_uniqueid; + nps->npcs_max_size = compressed; + } + nps->npcs_total_size += compressed; + nps->npcs_proc_count++; + } + return 0; } +static int +proc_pcontrol_null(__unused proc_t p, __unused void *arg) +{ + return PROC_RETURNED; +} + /* - * Deal with the out of swap space condition. This routine gets called when - * we want to swap something out but there's no more space left. Since this - * creates a memory deadlock situtation, we need to take action to free up - * some memory resources in order to prevent the system from hanging completely. - * The action we take is based on what the system processes running at user level - * have specified. Processes are marked in one of four categories: ones that - * can be killed immediately, ones that should be suspended, ones that should - * be throttled, and all the rest which are basically none of the above. Which - * processes are marked as being in which category is a user level policy decision; - * we just take action based on those decisions here. + * Deal with the low on compressor pool space condition... this function + * gets called when we are approaching the limits of the compressor pool or + * we are unable to create a new swap file. + * Since this eventually creates a memory deadlock situtation, we need to take action to free up + * memory resources (both compressed and uncompressed) in order to prevent the system from hanging completely. + * There are 2 categories of processes to deal with. Those that have an action + * associated with them by the task itself and those that do not. Actionable + * tasks can have one of three categories specified: ones that + * can be killed immediately, ones that should be suspended, and ones that should + * be throttled. Processes that do not have an action associated with them are normally + * ignored unless they are utilizing such a large percentage of the compressor pool (currently 50%) + * that only by killing them can we hope to put the system back into a usable state. */ -#define STARTING_PERCENTAGE 50 /* memory threshold expressed as a percentage */ - /* of physical memory */ +#define NO_PAGING_SPACE_DEBUG 0 -struct timeval last_no_space_action = {0, 0}; +extern uint64_t vm_compressor_pages_compressed(void); -void -no_paging_space_action(void) +struct timeval last_no_space_action = {.tv_sec = 0, .tv_usec = 0}; + +#if DEVELOPMENT || DEBUG +extern boolean_t kill_on_no_paging_space; +#endif /* DEVELOPMENT || DEBUG */ + +#define MB_SIZE (1024 * 1024ULL) +boolean_t memorystatus_kill_on_VM_compressor_space_shortage(boolean_t); + +extern int32_t max_kill_priority; +extern int memorystatus_get_proccnt_upto_priority(int32_t max_bucket_index); + +int +no_paging_space_action() { + proc_t p; + struct no_paging_space nps; + struct timeval now; + os_reason_t kill_reason; + + /* + * Throttle how often we come through here. Once every 5 seconds should be plenty. + */ + microtime(&now); + + if (now.tv_sec <= last_no_space_action.tv_sec + 5) { + return 0; + } + + /* + * Examine all processes and find the biggest (biggest is based on the number of pages this + * task has in the compressor pool) that has been marked to have some action + * taken when swap space runs out... we also find the biggest that hasn't been marked for + * action. + * + * If the biggest non-actionable task is over the "dangerously big" threashold (currently 50% of + * the total number of pages held by the compressor, we go ahead and kill it since no other task + * can have any real effect on the situation. Otherwise, we go after the actionable process. + */ + bzero(&nps, sizeof(nps)); + + proc_iterate(PROC_ALLPROCLIST, proc_pcontrol_null, (void *)NULL, proc_pcontrol_filter, (void *)&nps); + +#if NO_PAGING_SPACE_DEBUG + printf("low swap: npcs_proc_count = %d, npcs_total_size = %qd, npcs_max_size = %qd\n", + nps.npcs_proc_count, nps.npcs_total_size, nps.npcs_max_size); + printf("low swap: pcs_proc_count = %d, pcs_total_size = %qd, pcs_max_size = %qd\n", + nps.pcs_proc_count, nps.pcs_total_size, nps.pcs_max_size); + printf("low swap: apcs_proc_count = %d, apcs_total_size = %qd\n", + nps.apcs_proc_count, nps.apcs_total_size); +#endif + if (nps.npcs_max_size > (vm_compressor_pages_compressed() * 50) / 100) { + /* + * for now we'll knock out any task that has more then 50% of the pages + * held by the compressor + */ + if ((p = proc_find(nps.npcs_pid)) != PROC_NULL) { + if (nps.npcs_uniqueid == p->p_uniqueid) { + /* + * verify this is still the same process + * in case the proc exited and the pid got reused while + * we were finishing the proc_iterate and getting to this point + */ + last_no_space_action = now; + + printf("low swap: killing largest compressed process with pid %d (%s) and size %llu MB\n", p->p_pid, p->p_comm, (nps.pcs_max_size / MB_SIZE)); + kill_reason = os_reason_create(OS_REASON_JETSAM, JETSAM_REASON_LOWSWAP); + psignal_with_reason(p, SIGKILL, kill_reason); + + proc_rele(p); + + return 0; + } + + proc_rele(p); + } + } - uint64_t memory_threshold; - int num_found; - struct timeval now; + /* + * We have some processes within our jetsam bands of consideration and hence can be killed. + * So we will invoke the memorystatus thread to go ahead and kill something. + */ + if (memorystatus_get_proccnt_upto_priority(max_kill_priority) > 0) { + last_no_space_action = now; + memorystatus_kill_on_VM_compressor_space_shortage(TRUE /* async */); + return 1; + } /* - * Throttle how often we come through here. Once every 20 seconds should be plenty. + * No eligible processes to kill. So let's suspend/kill the largest + * process depending on its policy control specifications. */ - microtime(&now); + if (nps.pcs_max_size > 0) { + if ((p = proc_find(nps.pcs_pid)) != PROC_NULL) { + if (nps.pcs_uniqueid == p->p_uniqueid) { + /* + * verify this is still the same process + * in case the proc exited and the pid got reused while + * we were finishing the proc_iterate and getting to this point + */ + last_no_space_action = now; - if (now.tv_sec <= last_no_space_action.tv_sec + 20) - return; + proc_dopcontrol(p); + + proc_rele(p); + + return 1; + } + proc_rele(p); + } + } last_no_space_action = now; + printf("low swap: unable to find any eligible processes to take action on\n"); + + return 0; +} + +int +proc_trace_log(__unused proc_t p, struct proc_trace_log_args *uap, __unused int *retval) +{ + int ret = 0; + proc_t target_proc = PROC_NULL; + pid_t target_pid = uap->pid; + uint64_t target_uniqueid = uap->uniqueid; + task_t target_task = NULL; + + if (priv_check_cred(kauth_cred_get(), PRIV_PROC_TRACE_INSPECT, 0)) { + ret = EPERM; + goto out; + } + target_proc = proc_find(target_pid); + if (target_proc != PROC_NULL) { + if (target_uniqueid != proc_uniqueid(target_proc)) { + ret = ENOENT; + goto out; + } + + target_task = proc_task(target_proc); + if (task_send_trace_memory(target_task, target_pid, target_uniqueid)) { + ret = EINVAL; + goto out; + } + } else { + ret = ENOENT; + } + +out: + if (target_proc != PROC_NULL) { + proc_rele(target_proc); + } + return ret; +} + +#if VM_SCAN_FOR_SHADOW_CHAIN +extern int vm_map_shadow_max(vm_map_t map); +int proc_shadow_max(void); +int +proc_shadow_max(void) +{ + int retval, max; + proc_t p; + task_t task; + vm_map_t map; + + max = 0; + proc_list_lock(); + for (p = allproc.lh_first; (p != 0); p = p->p_list.le_next) { + if (p->p_stat == SIDL) { + continue; + } + task = p->task; + if (task == NULL) { + continue; + } + map = get_task_map(task); + if (map == NULL) { + continue; + } + retval = vm_map_shadow_max(map); + if (retval > max) { + max = retval; + } + } + proc_list_unlock(); + return max; +} +#endif /* VM_SCAN_FOR_SHADOW_CHAIN */ + +void proc_set_responsible_pid(proc_t target_proc, pid_t responsible_pid); +void +proc_set_responsible_pid(proc_t target_proc, pid_t responsible_pid) +{ + if (target_proc != NULL) { + target_proc->p_responsible_pid = responsible_pid; + } + return; +} + +int +proc_chrooted(proc_t p) +{ + int retval = 0; + + if (p) { + proc_fdlock(p); + retval = (p->p_fd->fd_rdir != NULL) ? 1 : 0; + proc_fdunlock(p); + } + + return retval; +} + +boolean_t +proc_send_synchronous_EXC_RESOURCE(proc_t p) +{ + if (p == PROC_NULL) { + return FALSE; + } + + /* Send sync EXC_RESOURCE if the process is traced */ + if (ISSET(p->p_lflag, P_LTRACED)) { + return TRUE; + } + return FALSE; +} + +size_t +proc_get_syscall_filter_mask_size(int which) +{ + if (which == SYSCALL_MASK_UNIX) { + return nsysent; + } + + return 0; +} + +int +proc_set_syscall_filter_mask(proc_t p, int which, unsigned char *maskptr, size_t masklen) +{ +#if DEVELOPMENT || DEBUG + if (syscallfilter_disable) { + printf("proc_set_syscall_filter_mask: attempt to set policy for pid %d, but disabled by boot-arg\n", proc_pid(p)); + return KERN_SUCCESS; + } +#endif // DEVELOPMENT || DEBUG + + if (which != SYSCALL_MASK_UNIX || + (maskptr != NULL && masklen != nsysent)) { + return EINVAL; + } + + p->syscall_filter_mask = maskptr; + + return KERN_SUCCESS; +} + +#ifdef CONFIG_32BIT_TELEMETRY +void +proc_log_32bit_telemetry(proc_t p) +{ + /* Gather info */ + char signature_buf[MAX_32BIT_EXEC_SIG_SIZE] = { 0 }; + char * signature_cur_end = &signature_buf[0]; + char * signature_buf_end = &signature_buf[MAX_32BIT_EXEC_SIG_SIZE - 1]; + int bytes_printed = 0; + + const char * teamid = NULL; + const char * identity = NULL; + struct cs_blob * csblob = NULL; + + proc_list_lock(); + /* - * Examine all processes and find those that have been marked to have some action - * taken when swap space runs out. Of those processes, select one or more and - * apply the specified action to them. The idea is to only take action against - * a few processes rather than hitting too many at once. If the low swap condition - * persists, this routine will get called again and we'll take action against more - * processes. - * - * Of the processes that have been marked, we choose which ones to take action - * against according to how much physical memory they're presently using. We - * start with the STARTING_THRESHOLD and any processes using more physical memory - * than the percentage threshold will have action taken against it. If there - * are no processes over the threshold, then the threshold is cut in half and we - * look again for processes using more than this threshold. We continue in - * this fashion until we find at least one process to take action against. This - * iterative approach is less than ideally efficient, however we only get here - * when the system is almost in a memory deadlock and is pretty much just - * thrashing if it's doing anything at all. Therefore, the cpu overhead of - * potentially multiple passes here probably isn't revelant. + * Get proc name and parent proc name; if the parent execs, we'll get a + * garbled name. */ + bytes_printed = scnprintf(signature_cur_end, + signature_buf_end - signature_cur_end, + "%s,%s,", p->p_name, + (p->p_pptr ? p->p_pptr->p_name : "")); - memory_threshold = (sane_size * STARTING_PERCENTAGE) / 100; /* resident threshold in bytes */ + if (bytes_printed > 0) { + signature_cur_end += bytes_printed; + } - for (num_found = 0; num_found == 0; memory_threshold = memory_threshold / 2) { - proc_iterate(PROC_ALLPROCLIST, proc_dopcontrol, (void *)&num_found, proc_pcontrol_filter, (void *)&memory_threshold); + proc_list_unlock(); - /* - * If we just looked with memory_threshold == 0, then there's no need to iterate any further since - * we won't find any eligible processes at this point. - */ + /* Get developer info. */ + vnode_t v = proc_getexecutablevnode(p); - if (memory_threshold == 0) { - if (num_found == 0) /* log that we couldn't do anything in this case */ - printf("low swap: unable to find any eligible processes to take action on\n"); + if (v) { + csblob = csvnode_get_blob(v, 0); - break; + if (csblob) { + teamid = csblob_get_teamid(csblob); + identity = csblob_get_identity(csblob); } } + + if (teamid == NULL) { + teamid = ""; + } + + if (identity == NULL) { + identity = ""; + } + + bytes_printed = scnprintf(signature_cur_end, + signature_buf_end - signature_cur_end, + "%s,%s", teamid, identity); + + if (bytes_printed > 0) { + signature_cur_end += bytes_printed; + } + + if (v) { + vnode_put(v); + } + + /* + * We may want to rate limit here, although the SUMMARIZE key should + * help us aggregate events in userspace. + */ + + /* Emit log */ + kern_asl_msg(LOG_DEBUG, "messagetracer", 3, + /* 0 */ "com.apple.message.domain", "com.apple.kernel.32bit_exec", + /* 1 */ "com.apple.message.signature", signature_buf, + /* 2 */ "com.apple.message.summarize", "YES", + NULL); } +#endif /* CONFIG_32BIT_TELEMETRY */