1.7.1 init386()

init386() is defined in sys/i386/i386/machdep.c and performs low-level initialization, specific to the i386 chip. The switch to protected mode was performed by the loader. The loader has created the very first task, in which the kernel continues to operate. Before running straight away to the code, I will enumerate the tasks the processor must complete to initialize protected mode execution:

• Initialize the kernel tunable parameters, passed from the bootstrapping program.

• Prepare the GDT.

• Prepare the IDT.

• Initialize the system console.

• Initialize the DDB, if it is compiled into kernel.

• Initialize the TSS.

• Prepare the LDT.

• Set up proc0's pcb.

What init386() first does is initialize the tunable parameters passed from bootstrap. This is done by setting the environment pointer (envp) and calling init_param1(). The envp pointer has been passed from loader in the bootinfo structure:

sys/i386/i386/machdep.c:
        kern_envp = (caddr_t)bootinfo.bi_envp + KERNBASE;

    /* Init basic tunables, hz etc */
    init_param1();

init_param1() is defined in sys/kern/subr_param.c. That file has a number of sysctls, and two functions, init_param1() and init_param2(), that are called from init386():

sys/kern/subr_param.c
    hz = HZ;
    TUNABLE_INT_FETCH("kern.hz", &hz);

TUNABLE_<typename>_FETCH is used to fetch the value from the environment:

/usr/src/sys/sys/kernel.h
#define TUNABLE_INT_FETCH(path, var)    getenv_int((path), (var))

Sysctl kern.hz is the system clock tick. Along with this, the following sysctls are set by init_param1(): kern.maxswzone, kern.maxbcache, kern.maxtsiz, kern.dfldsiz, kern.maxdsiz, kern.dflssiz, kern.maxssiz, kern.sgrowsiz.

Then init386() prepares the Global Descriptors Table (GDT). Every task on an x86 is running in its own virtual address space, and this space is addressed by a segment:offset pair. Say, for instance, the current instruction to be executed by the processor lies at CS:EIP, then the linear virtual address for that instruction would be “the virtual address of code segment CS” + EIP. For convenience, segments begin at virtual address 0 and end at a 4Gb boundary. Therefore, the instruction's linear virtual address for this example would just be the value of EIP. Segment registers such as CS, DS etc are the selectors, i.e. indexes, into GDT (to be more precise, an index is not a selector itself, but the INDEX field of a selector). FreeBSD's GDT holds descriptors for 15 selectors per CPU:

sys/i386/i386/machdep.c:
union descriptor gdt[NGDT * MAXCPU];    /* global descriptor table */

sys/i386/include/segments.h:
/*
 * Entries in the Global Descriptor Table (GDT)
 */
#define GNULL_SEL   0   /* Null Descriptor */
#define GCODE_SEL   1   /* Kernel Code Descriptor */
#define GDATA_SEL   2   /* Kernel Data Descriptor */
#define GPRIV_SEL   3   /* SMP Per-Processor Private Data */
#define GPROC0_SEL  4   /* Task state process slot zero and up */
#define GLDT_SEL    5   /* LDT - eventually one per process */
#define GUSERLDT_SEL    6   /* User LDT */
#define GTGATE_SEL  7   /* Process task switch gate */
#define GBIOSLOWMEM_SEL 8   /* BIOS low memory access (must be entry 8) */
#define GPANIC_SEL  9   /* Task state to consider panic from */
#define GBIOSCODE32_SEL 10  /* BIOS interface (32bit Code) */
#define GBIOSCODE16_SEL 11  /* BIOS interface (16bit Code) */
#define GBIOSDATA_SEL   12  /* BIOS interface (Data) */
#define GBIOSUTIL_SEL   13  /* BIOS interface (Utility) */
#define GBIOSARGS_SEL   14  /* BIOS interface (Arguments) */

Note that those #defines are not selectors themselves, but just a field INDEX of a selector, so they are exactly the indices of the GDT. for example, an actual selector for the kernel code (GCODE_SEL) has the value 0x08.

The next step is to initialize the Interrupt Descriptor Table (IDT). This table is to be referenced by the processor when a software or hardware interrupt occurs. For example, to make a system call, user application issues the INT 0x80 instruction. This is a software interrupt, so the processor's hardware looks up a record with index 0x80 in the IDT. This record points to the routine that handles this interrupt, in this particular case, this will be the kernel's syscall gate. The IDT may have a maximum of 256 (0x100) records. The kernel allocates NIDT records for the IDT, where NIDT is the maximum (256):

sys/i386/i386/machdep.c:
static struct gate_descriptor idt0[NIDT];
struct gate_descriptor *idt = &idt0[0]; /* interrupt descriptor table */

For each interrupt, an appropriate handler is set. The syscall gate for INT 0x80 is set as well:

sys/i386/i386/machdep.c:
    setidt(0x80, &IDTVEC(int0x80_syscall),
            SDT_SYS386TGT, SEL_UPL, GSEL(GCODE_SEL, SEL_KPL));

So when a userland application issues the INT 0x80 instruction, control will transfer to the function _Xint0x80_syscall, which is in the kernel code segment and will be executed with supervisor privileges.

Console and DDB are then initialized:

sys/i386/i386/machdep.c:
    cninit();
/* skipped */
#ifdef DDB
    kdb_init();
    if (boothowto & RB_KDB)
        Debugger("Boot flags requested debugger");
#endif

The Task State Segment is another x86 protected mode structure, the TSS is used by the hardware to store task information when a task switch occurs.

The Local Descriptors Table is used to reference userland code and data. Several selectors are defined to point to the LDT, they are the system call gates and the user code and data selectors:

/usr/include/machine/segments.h
#define LSYS5CALLS_SEL  0   /* forced by intel BCS */
#define LSYS5SIGR_SEL   1
#define L43BSDCALLS_SEL 2   /* notyet */
#define LUCODE_SEL  3
#define LSOL26CALLS_SEL 4   /* Solaris >= 2.6 system call gate */
#define LUDATA_SEL  5
/* separate stack, es,fs,gs sels ? */
/* #define  LPOSIXCALLS_SEL 5*/ /* notyet */
#define LBSDICALLS_SEL  16  /* BSDI system call gate */
#define NLDT        (LBSDICALLS_SEL + 1)

Next, proc0's Process Control Block (struct pcb) structure is initialized. proc0 is a struct proc structure that describes a kernel process. It is always present while the kernel is running, therefore it is declared as global:

sys/kern/kern_init.c:
    struct  proc proc0;

The structure struct pcb is a part of a proc structure. It is defined in /usr/include/machine/pcb.h and has a process's information specific to the i386 architecture, such as registers values.

1.7.2 mi_startup()

This function performs a bubble sort of all the system initialization objects and then calls the entry of each object one by one:

sys/kern/init_main.c:
    for (sipp = sysinit; *sipp; sipp++) {

        /* ... skipped ... */

        /* Call function */
        (*((*sipp)->func))((*sipp)->udata);
        /* ... skipped ... */
    }

Although the sysinit framework is described in the Developers' Handbook, I will discuss the internals of it.

Every system initialization object (sysinit object) is created by calling a SYSINIT() macro. Let us take as example an announce sysinit object. This object prints the copyright message:

sys/kern/init_main.c:
static void
print_caddr_t(void *data __unused)
{
    printf("%s", (char *)data);
}
SYSINIT(announce, SI_SUB_COPYRIGHT, SI_ORDER_FIRST, print_caddr_t, copyright)

The subsystem ID for this object is SI_SUB_COPYRIGHT (0x0800001), which comes right after the SI_SUB_CONSOLE (0x0800000). So, the copyright message will be printed out first, just after the console initialization.

Let us take a look at what exactly the macro SYSINIT() does. It expands to a C_SYSINIT() macro. The C_SYSINIT() macro then expands to a static struct sysinit structure declaration with another DATA_SET macro call:

/usr/include/sys/kernel.h:
      #define C_SYSINIT(uniquifier, subsystem, order, func, ident) \
      static struct sysinit uniquifier ## _sys_init = { \ subsystem, \
      order, \ func, \ ident \ }; \ DATA_SET(sysinit_set,uniquifier ##
      _sys_init);

#define SYSINIT(uniquifier, subsystem, order, func, ident)  \
    C_SYSINIT(uniquifier, subsystem, order,         \
    (sysinit_cfunc_t)(sysinit_nfunc_t)func, (void *)ident)

The DATA_SET() macro expands to a MAKE_SET(), and that macro is the point where the all sysinit magic is hidden:

/usr/include/linker_set.h
#define MAKE_SET(set, sym)                      \
    static void const * const __set_##set##_sym_##sym = &sym;   \
    __asm(".section .set." #set ",\"aw\"");             \
    __asm(".long " #sym);                       \
    __asm(".previous")
#endif
#define TEXT_SET(set, sym) MAKE_SET(set, sym)
#define DATA_SET(set, sym) MAKE_SET(set, sym)

In our case, the following declaration will occur:

static struct sysinit announce_sys_init = {
    SI_SUB_COPYRIGHT,
    SI_ORDER_FIRST,
    (sysinit_cfunc_t)(sysinit_nfunc_t)  print_caddr_t,
    (void *) copyright
};

static void const *const __set_sysinit_set_sym_announce_sys_init =
    &announce_sys_init;
__asm(".section .set.sysinit_set" ",\"aw\"");
__asm(".long " "announce_sys_init");
__asm(".previous");

The first __asm instruction will create an ELF section within the kernel's executable. This will happen at kernel link time. The section will have the name .set.sysinit_set. The content of this section is one 32-bit value, the address of announce_sys_init structure, and that is what the second __asm is. The third __asm instruction marks the end of a section. If a directive with the same section name occurred before, the content, i.e. the 32-bit value, will be appended to the existing section, so forming an array of 32-bit pointers.

Running objdump on a kernel binary, you may notice the presence of such small sections:

% objdump -h /kernel
  7 .set.cons_set 00000014  c03164c0  c03164c0  002154c0  2**2
                  CONTENTS, ALLOC, LOAD, DATA
  8 .set.kbddriver_set 00000010  c03164d4  c03164d4  002154d4  2**2
                  CONTENTS, ALLOC, LOAD, DATA
  9 .set.scrndr_set 00000024  c03164e4  c03164e4  002154e4  2**2
                  CONTENTS, ALLOC, LOAD, DATA
 10 .set.scterm_set 0000000c  c0316508  c0316508  00215508  2**2
                  CONTENTS, ALLOC, LOAD, DATA
 11 .set.sysctl_set 0000097c  c0316514  c0316514  00215514  2**2
                  CONTENTS, ALLOC, LOAD, DATA
 12 .set.sysinit_set 00000664  c0316e90  c0316e90  00215e90  2**2
                  CONTENTS, ALLOC, LOAD, DATA

This screen dump shows that the size of .set.sysinit_set section is 0x664 bytes, so 0x664/sizeof(void *) sysinit objects are compiled into the kernel. The other sections such as .set.sysctl_set represent other linker sets.

By defining a variable of type struct linker_set the content of .set.sysinit_set section will be “collected” into that variable:

sys/kern/init_main.c:
      extern struct linker_set sysinit_set; /* XXX */

The struct linker_set is defined as follows:

/usr/include/linker_set.h:
  struct linker_set {
    int ls_length;
    void    *ls_items[1];       /* really ls_length of them, trailing NULL */
};

The first node will be equal to the number of a sysinit objects, and the second node will be a NULL-terminated array of pointers to them.

Returning to the mi_startup() discussion, it is must be clear now, how the sysinit objects are being organized. The mi_startup() function sorts them and calls each. The very last object is the system scheduler:

/usr/include/sys/kernel.h:
enum sysinit_sub_id {
    SI_SUB_DUMMY        = 0x0000000,    /* not executed; for linker*/
    SI_SUB_DONE     = 0x0000001,    /* processed*/
    SI_SUB_CONSOLE      = 0x0800000,    /* console*/
    SI_SUB_COPYRIGHT    = 0x0800001,    /* first use of console*/
...
    SI_SUB_RUN_SCHEDULER    = 0xfffffff /* scheduler: no return*/
};

The system scheduler sysinit object is defined in the file sys/vm/vm_glue.c, and the entry point for that object is scheduler(). That function is actually an infinite loop, and it represents a process with PID 0, the swapper process. The proc0 structure, mentioned before, is used to describe it.

The first user process, called init, is created by the sysinit object init:

sys/kern/init_main.c:
static void
create_init(const void *udata __unused)
{
    int error;
    int s;

    s = splhigh();
    error = fork1(&proc0, RFFDG | RFPROC, &initproc);
    if (error)
        panic("cannot fork init: %d\n", error);
    initproc->p_flag |= P_INMEM | P_SYSTEM;
    cpu_set_fork_handler(initproc, start_init, NULL);
    remrunqueue(initproc);
    splx(s);
}
SYSINIT(init,SI_SUB_CREATE_INIT, SI_ORDER_FIRST, create_init, NULL)

The create_init() allocates a new process by calling fork1(), but does not mark it runnable. When this new process is scheduled for execution by the scheduler, the start_init() will be called. That function is defined in init_main.c. It tries to load and exec the init binary, probing /sbin/init first, then /sbin/oinit, /sbin/init.bak, and finally /stand/sysinstall:

sys/kern/init_main.c:
static char init_path[MAXPATHLEN] =
#ifdef  INIT_PATH
    __XSTRING(INIT_PATH);
#else
    "/sbin/init:/sbin/oinit:/sbin/init.bak:/stand/sysinstall";
#endif

3.3.1 Structures

struct kobj_method

3.3.2 Functions

void kobj_class_compile(kobj_class_t cls);
void kobj_class_compile_static(kobj_class_t cls, kobj_ops_t ops);
void kobj_class_free(kobj_class_t cls);
kobj_t kobj_create(kobj_class_t cls, struct malloc_type *mtype, int mflags);
void kobj_init(kobj_t obj, kobj_class_t cls);
void kobj_delete(kobj_t obj, struct malloc_type *mtype);

3.3.3 Macros

KOBJ_CLASS_FIELDS
KOBJ_FIELDS
DEFINE_CLASS(name, methods, size)
KOBJMETHOD(NAME, FUNC)

3.3.4 Headers

<sys/param.h>
<sys/kobj.h>

3.3.5 Creating an interface template

The first step in using Kobj is to create an Interface. Creating the interface involves creating a template that the script src/sys/kern/makeobjops.pl can use to generate the header and code for the method declarations and method lookup functions.

Within this template the following keywords are used: #include, INTERFACE, CODE, METHOD, STATICMETHOD, and DEFAULT.

The #include statement and what follows it is copied verbatim to the head of the generated code file.

For example:

#include <sys/foo.h>

The INTERFACE keyword is used to define the interface name. This name is concatenated with each method name as [interface name]_[method name]. Its syntax is INTERFACE [interface name];.

For example:

INTERFACE foo;

The CODE keyword copies its arguments verbatim into the code file. Its syntax is CODE { [whatever] };

For example:

CODE {
    struct foo * foo_alloc_null(struct bar *)
    {
        return NULL;
}
};

The METHOD keyword describes a method. Its syntax is METHOD [return type] [method name] { [object [, arguments]] };

For example:

METHOD int bar {
    struct object *;
    struct foo *;
    struct bar;
};

The DEFAULT keyword may follow the METHOD keyword. It extends the METHOD key word to include the default function for method. The extended syntax is METHOD [return type] [method name] { [object; [other arguments]] }DEFAULT [default function];

For example:

METHOD int bar {
    struct object *;
    struct foo *;
    int bar;
} DEFAULT foo_hack;

The STATICMETHOD keyword is used like the METHOD keyword except the kobj data is not at the head of the object structure so casting to kobj_t would be incorrect. Instead STATICMETHOD relies on the Kobj data being referenced as 'ops'. This is also useful for calling methods directly out of a class's method table.

Other complete examples:

src/sys/kern/bus_if.m
src/sys/kern/device_if.m

3.3.6 Creating a Class

The second step in using Kobj is to create a class. A class consists of a name, a table of methods, and the size of objects if Kobj's object handling facilities are used. To create the class use the macro DEFINE_CLASS(). To create the method table create an array of kobj_method_t terminated by a NULL entry. Each non-NULL entry may be created using the macro KOBJMETHOD().

For example:

DEFINE_CLASS(fooclass, foomethods, sizeof(struct foodata));

kobj_method_t foomethods[] = {
    KOBJMETHOD(bar_doo, foo_doo),
    KOBJMETHOD(bar_foo, foo_foo),
    { NULL, NULL}
};

The class must be “compiled”. Depending on the state of the system at the time that the class is to be initialized a statically allocated cache, “ops table” have to be used. This can be accomplished by declaring a struct kobj_ops and using kobj_class_compile_static(); otherwise, kobj_class_compile() should be used.

3.3.7 Creating an Object

The third step in using Kobj involves how to define the object. Kobj object creation routines assume that Kobj data is at the head of an object. If this in not appropriate you will have to allocate the object yourself and then use kobj_init() on the Kobj portion of it; otherwise, you may use kobj_create() to allocate and initialize the Kobj portion of the object automatically. kobj_init() may also be used to change the class that an object uses.

To integrate Kobj into the object you should use the macro KOBJ_FIELDS.

For example

struct foo_data {
    KOBJ_FIELDS;
    foo_foo;
    foo_bar;
};

3.3.8 Calling Methods

The last step in using Kobj is to simply use the generated functions to use the desired method within the object's class. This is as simple as using the interface name and the method name with a few modifications. The interface name should be concatenated with the method name using a '_' between them, all in upper case.

For example, if the interface name was foo and the method was bar then the call would be:

[return value = ] FOO_BAR(object [, other parameters]);

3.3.9 Cleaning Up

When an object allocated through kobj_create() is no longer needed kobj_delete() may be called on it, and when a class is no longer being used kobj_class_free() may be called on it.

4.1.1 Userland Code

The source for the userland jail is located in /usr/src/usr.sbin/jail, consisting of one file, jail.c. The program takes these arguments: the path of the jail, hostname, IP address, and the command to be executed.

/usr/include/sys/jail.h: 

struct jail {
        u_int32_t       version;
        char            *path;
        char            *hostname;
        u_int32_t       ip_number;
};

/usr/src/usr.sbin/jail/jail.c
char path[PATH_MAX];
...
if (realpath(argv[0], path) == NULL)
    err(1, "realpath: %s", argv[0]);
if (chdir(path) != 0)
    err(1, "chdir: %s", path);
memset(&j, 0, sizeof(j));
j.version = 0;
j.path = path;
j.hostname = argv[1];

/usr/src/usr.sbin/jail/jail.c:
struct in_addr in; 
... 
if (inet_aton(argv[2], &in) == 0)
    errx(1, "Could not make sense of ip-number: %s", argv[2]);
j.ip_number = ntohl(in.s_addr);

/usr/src/usr.sbin/jail/jail.c
i = jail(&j); 
... 
if (execv(argv[3], argv + 3) != 0)
    err(1, "execv: %s", argv[3]);

4.1.2 Kernel Space

We will now be looking at the file /usr/src/sys/kern/kern_jail.c. This is the file where the jail(2) system call, appropriate sysctls, and networking functions are defined.

/usr/src/sys/kern/kern_jail.c:

int     jail_set_hostname_allowed = 1;
SYSCTL_INT(_security_jail, OID_AUTO, set_hostname_allowed, CTLFLAG_RW,
    &jail_set_hostname_allowed, 0,
    "Processes in jail can set their hostnames");

int     jail_socket_unixiproute_only = 1;
SYSCTL_INT(_security_jail, OID_AUTO, socket_unixiproute_only, CTLFLAG_RW,
    &jail_socket_unixiproute_only, 0,
    "Processes in jail are limited to creating UNIX/IPv4/route sockets only");

int     jail_sysvipc_allowed = 0;
SYSCTL_INT(_security_jail, OID_AUTO, sysvipc_allowed, CTLFLAG_RW,
    &jail_sysvipc_allowed, 0,
    "Processes in jail can use System V IPC primitives");

static int jail_enforce_statfs = 2;
SYSCTL_INT(_security_jail, OID_AUTO, enforce_statfs, CTLFLAG_RW,
    &jail_enforce_statfs, 0,
    "Processes in jail cannot see all mounted file systems");

int    jail_allow_raw_sockets = 0;
SYSCTL_INT(_security_jail, OID_AUTO, allow_raw_sockets, CTLFLAG_RW,
    &jail_allow_raw_sockets, 0,
    "Prison root can create raw sockets");

int    jail_chflags_allowed = 0;
SYSCTL_INT(_security_jail, OID_AUTO, chflags_allowed, CTLFLAG_RW,
    &jail_chflags_allowed, 0,
    "Processes in jail can alter system file flags");

int     jail_mount_allowed = 0;
SYSCTL_INT(_security_jail, OID_AUTO, mount_allowed, CTLFLAG_RW,
    &jail_mount_allowed, 0,
    "Processes in jail can mount/unmount jail-friendly file systems");

/usr/src/sys/kern/kern_jail.c:
/*
 * struct jail_args {
 *  struct jail *jail;
 * };
 */ 
int 
jail(struct thread *td, struct jail_args *uap)

/usr/src/sys/kern/kern_jail.c: 
error = copyin(uap->jail, &j, sizeof(j));

/usr/include/sys/jail.h:
struct prison {
        LIST_ENTRY(prison) pr_list;                     /* (a) all prisons */
        int              pr_id;                         /* (c) prison id */
        int              pr_ref;                        /* (p) refcount */
        char             pr_path[MAXPATHLEN];           /* (c) chroot path */
        struct vnode    *pr_root;                       /* (c) vnode to rdir */
        char             pr_host[MAXHOSTNAMELEN];       /* (p) jail hostname */
        u_int32_t        pr_ip;                         /* (c) ip addr host */
        void            *pr_linux;                      /* (p) linux abi */
        int              pr_securelevel;                /* (p) securelevel */
        struct task      pr_task;                       /* (d) destroy task */
        struct mtx       pr_mtx;
      void            **pr_slots;                     /* (p) additional data */
};

/usr/src/sys/kern/kern_jail.c:
MALLOC(pr, struct prison *, sizeof(*pr), M_PRISON, M_WAITOK | M_ZERO);
...
error = copyinstr(j.path, &pr->pr_path, sizeof(pr->pr_path), 0);
if (error)
    goto e_killmtx;
...
error = copyinstr(j.hostname, &pr->pr_host, sizeof(pr->pr_host), 0);
if (error)
     goto e_dropvnref;
pr->pr_ip = j.ip_number;

/usr/src/sys/kern/kern_jail.c:
/*
 * struct jail_attach_args {
 *      int jid;
 * };
 */
int
jail_attach(struct thread *td, struct jail_attach_args *uap)

/usr/include/sys/proc.h: 
struct thread {
    ...
    struct proc *td_proc;
    ...
};
struct proc { 
    ...
    struct ucred *p_ucred; 
    ...
};
/usr/include/sys/ucred.h
struct ucred {
    ...
    struct prison *cr_prison;
    ...
};

/usr/src/sys/kern/kern_jail.c
int
jail(struct thread *td, struct jail_args *uap)
{
...
    struct jail_attach_args jaa;
...
    error = jail_attach(td, &jaa);
    if (error)
        goto e_dropprref;
...
}

int
jail_attach(struct thread *td, struct jail_attach_args *uap)
{
    struct proc *p;
    struct ucred *newcred, *oldcred;
    struct prison *pr;
...
    p = td->td_proc;
...
    pr = prison_find(uap->jid);
...
    change_root(pr->pr_root, td);
...
    newcred->cr_prison = pr;
    p->p_ucred = newcred;
...
}

/usr/src/sys/kern/kern_fork.c:
p2->p_ucred = crhold(td->td_ucred);
...
td2->td_ucred = crhold(p2->p_ucred);

4.2.1 SysV IPC

System V IPC is based on messages. Processes can send each other these messages which tell them how to act. The functions which deal with messages are: msgctl(3), msgget(3), msgsnd(3) and msgrcv(3). Earlier, I mentioned that there were certain sysctls you could turn on or off in order to affect the behavior of jail. One of these sysctls was security.jail.sysvipc_allowed. By default, this sysctl is set to 0. If it were set to 1, it would defeat the whole purpose of having a jail; privileged users from the jail would be able to affect processes outside the jailed environment. The difference between a message and a signal is that the message only consists of the signal number.

/usr/src/sys/kern/sysv_msg.c:

• msgget(key, msgflg): msgget returns (and possibly creates) a message descriptor that designates a message queue for use in other functions.

• msgctl(msgid, cmd, buf): Using this function, a process can query the status of a message descriptor.

• msgsnd(msgid, msgp, msgsz, msgflg): msgsnd sends a message to a process.

• msgrcv(msgid, msgp, msgsz, msgtyp, msgflg): a process receives messages using this function

In each of the system calls corresponding to these functions, there is this conditional:

/usr/src/sys/kern/sysv_msg.c:
if (!jail_sysvipc_allowed && jailed(td->td_ucred))
    return (ENOSYS);

Semaphore system calls allow processes to synchronize execution by doing a set of operations atomically on a set of semaphores. Basically semaphores provide another way for processes lock resources. However, process waiting on a semaphore, that is being used, will sleep until the resources are relinquished. The following semaphore system calls are blocked inside a jail: semget(2), semctl(2) and semop(2).

/usr/src/sys/kern/sysv_sem.c:

• semctl(semid, semnum, cmd, ...): semctl does the specified cmd on the semaphore queue indicated by semid.

• semget(key, nsems, flag): semget creates an array of semaphores, corresponding to key.

  key and flag take on the same meaning as they do in msgget.

• semop(semid, array, nops): semop performs a group of operations indicated by array, to the set of semaphores identified by semid.

System V IPC allows for processes to share memory. Processes can communicate directly with each other by sharing parts of their virtual address space and then reading and writing data stored in the shared memory. These system calls are blocked within a jailed environment: shmdt(2), shmat(2), shmctl(2) and shmget(2).

/usr/src/sys/kern/sysv_shm.c:

• shmctl(shmid, cmd, buf): shmctl does various control operations on the shared memory region identified by shmid.

• shmget(key, size, flag): shmget accesses or creates a shared memory region of size bytes.

• shmat(shmid, addr, flag): shmat attaches a shared memory region identified by shmid to the address space of a process.

• shmdt(addr): shmdt detaches the shared memory region previously attached at addr.

4.2.2 Sockets

Jail treats the socket(2) system call and related lower-level socket functions in a special manner. In order to determine whether a certain socket is allowed to be created, it first checks to see if the sysctl security.jail.socket_unixiproute_only is set. If set, sockets are only allowed to be created if the family specified is either PF_LOCAL, PF_INET or PF_ROUTE. Otherwise, it returns an error.

/usr/src/sys/kern/uipc_socket.c:
int
socreate(int dom, struct socket **aso, int type, int proto,
    struct ucred *cred, struct thread *td)
{
    struct protosw *prp;
...
    if (jailed(cred) && jail_socket_unixiproute_only &&
        prp->pr_domain->dom_family != PF_LOCAL &&
        prp->pr_domain->dom_family != PF_INET &&
        prp->pr_domain->dom_family != PF_ROUTE) {
        return (EPROTONOSUPPORT);
    }
...
}

4.2.3 Berkeley Packet Filter

The Berkeley Packet Filter provides a raw interface to data link layers in a protocol independent fashion. BPF is now controlled by the devfs(8) whether it can be used in a jailed environment.

4.2.4 Protocols

There are certain protocols which are very common, such as TCP, UDP, IP and ICMP. IP and ICMP are on the same level: the network layer 2. There are certain precautions which are taken in order to prevent a jailed process from binding a protocol to a certain address only if the nam parameter is set. nam is a pointer to a sockaddr structure, which describes the address on which to bind the service. A more exact definition is that sockaddr "may be used as a template for referring to the identifying tag and length of each address". In the function in_pcbbind_setup(), sin is a pointer to a sockaddr_in structure, which contains the port, address, length and domain family of the socket which is to be bound. Basically, this disallows any processes from jail to be able to specify the address that doesn't belong to the jail in which the calling process exists.

/usr/src/sys/netinet/in_pcb.c: 
int
in_pcbbind_setup(struct inpcb *inp, struct sockaddr *nam, in_addr_t *laddrp,
    u_short *lportp, struct ucred *cred)
{
    ...
    struct sockaddr_in *sin;
    ...
    if (nam) {
        sin = (struct sockaddr_in *)nam;
        ...
        if (sin->sin_addr.s_addr != INADDR_ANY)
            if (prison_ip(cred, 0, &sin->sin_addr.s_addr))
                return(EINVAL);
        ...
        if (lport) {
            ...
            if (prison && prison_ip(cred, 0, &sin->sin_addr.s_addr))
                return (EADDRNOTAVAIL);
            ...
        }
    }
    if (lport == 0) {
        ...
        if (laddr.s_addr != INADDR_ANY)
            if (prison_ip(cred, 0, &laddr.s_addr))
                return (EINVAL);
        ...
    }
...
    if (prison_ip(cred, 0, &laddr.s_addr))
        return (EINVAL);
...
}

You might be wondering what function prison_ip() does. prison_ip() is given three arguments, a pointer to the credential(represented by cred), any flags, and an IP address. It returns 1 if the IP address does NOT belong to the jail or 0 otherwise. As you can see from the code, if it is indeed an IP address not belonging to the jail, the protocol is not allowed to bind to that address.

/usr/src/sys/kern/kern_jail.c:
int
prison_ip(struct ucred *cred, int flag, u_int32_t *ip)
{
    u_int32_t tmp;

    if (!jailed(cred))
        return (0);
    if (flag)
        tmp = *ip;
    else
        tmp = ntohl(*ip);
    if (tmp == INADDR_ANY) {
        if (flag)
            *ip = cred->cr_prison->pr_ip;
        else
            *ip = htonl(cred->cr_prison->pr_ip);
        return (0);
    }
    if (tmp == INADDR_LOOPBACK) {
        if (flag)
            *ip = cred->cr_prison->pr_ip;
        else
            *ip = htonl(cred->cr_prison->pr_ip);
        return (0);
    }
    if (cred->cr_prison->pr_ip != tmp)
        return (1);
    return (0);
}

4.2.5 Filesystem

Even root users within the jail are not allowed to unset or modify any file flags, such as immutable, append-only, and undeleteable flags, if the securelevel is greater than 0.

/usr/src/sys/ufs/ufs/ufs_vnops.c:
static int
ufs_setattr(ap)
    ...
{
    ...
        if (!priv_check_cred(cred, PRIV_VFS_SYSFLAGS, 0)) {
            if (ip->i_flags
                & (SF_NOUNLINK | SF_IMMUTABLE | SF_APPEND)) {
                    error = securelevel_gt(cred, 0);
                    if (error)
                        return (error);
            }
            ...
        }
}
/usr/src/sys/kern/kern_priv.c
int
priv_check_cred(struct ucred *cred, int priv, int flags)
{
    ...
    error = prison_priv_check(cred, priv);
    if (error)
        return (error);
    ...
}
/usr/src/sys/kern/kern_jail.c
int
prison_priv_check(struct ucred *cred, int priv)
{
    ...
    switch (priv) {
    ...
    case PRIV_VFS_SYSFLAGS:
        if (jail_chflags_allowed)
            return (0);
        else
            return (EPERM);
    ...
    }
    ...
}

5.3.1 Interface

<sys/kernel.h>

SYSINIT(uniquifier, subsystem, order, func, ident)
SYSUNINIT(uniquifier, subsystem, order, func, ident)

5.3.2 Startup

The SYSINIT() macro creates the necessary SYSINIT data in SYSINIT's startup data set for SYSINIT to sort and dispatch a function at system startup and module load. SYSINIT() takes a uniquifier that SYSINIT uses to identify the particular function dispatch data, the subsystem order, the subsystem element order, the function to call, and the data to pass the function. All functions must take a constant pointer argument.

#include <sys/kernel.h>

void foo_null(void *unused)
{
        foo_doo();
}
SYSINIT(foo, SI_SUB_FOO, SI_ORDER_FOO, foo_null, NULL);

struct foo foo_voodoo = {
        FOO_VOODOO;
}

void foo_arg(void *vdata)
{
        struct foo *foo = (struct foo *)vdata;
        foo_data(foo);
}
SYSINIT(bar, SI_SUB_FOO, SI_ORDER_FOO, foo_arg, &foo_voodoo);
   

Note that SI_SUB_FOO and SI_ORDER_FOO need to be in the sysinit_sub_id and sysinit_elem_order enum's as mentioned above. Either use existing ones or add your own to the enum's. You can also use math for fine-tuning the order a SYSINIT will run in. This example shows a SYSINIT that needs to be run just barely before the SYSINIT's that handle tuning kernel parameters.

static void
mptable_register(void *dummy __unused)
{

    apic_register_enumerator(&mptable_enumerator);
}

SYSINIT(mptable_register, SI_SUB_TUNABLES - 1, SI_ORDER_FIRST,
    mptable_register, NULL);

5.3.3 Shutdown

The SYSUNINIT() macro behaves similarly to the SYSINIT() macro except that it adds the SYSINIT data to SYSINIT's shutdown data set.

#include <sys/kernel.h>

void foo_cleanup(void *unused)
{
        foo_kill();
}
SYSUNINIT(foobar, SI_SUB_FOO, SI_ORDER_FOO, foo_cleanup, NULL);

struct foo_stack foo_stack = {
        FOO_STACK_VOODOO;
}

void foo_flush(void *vdata)
{
}
SYSUNINIT(barfoo, SI_SUB_FOO, SI_ORDER_FOO, foo_flush, &foo_stack);
   

6.5.1 Kernel Elements

The MAC Framework contains a number of kernel elements:

• Framework management interfaces

• Concurrency and synchronization primitives.

• Policy registration

• Extensible security label for kernel objects

• Policy entry point composition operators

• Label management primitives

• Entry point API invoked by kernel services

• Entry point API to policy modules

• Entry points implementations (policy life cycle, object life cycle/label management, access control checks).

• Policy-agnostic label-management system calls

• mac_syscall() multiplex system call

• Various security policies implemented as MAC policy modules

6.5.2 Framework Management Interfaces

The TrustedBSD MAC Framework may be directly managed using sysctl's, loader tunables, and system calls.

In most cases, sysctl's and loader tunables of the same name modify the same parameters, and control behavior such as enforcement of protections relating to various kernel subsystems. In addition, if MAC debugging support is compiled into the kernel, several counters will be maintained tracking label allocation. It is generally advisable that per-subsystem enforcement controls not be used to control policy behavior in production environments, as they broadly impact the operation of all active policies. Instead, per-policy controls should be preferred, as they provide greater granularity and greater operational consistency for policy modules.

Loading and unloading of policy modules is performed using the system module management system calls and other system interfaces, including boot loader variables; policy modules will have the opportunity to influence load and unload events, including preventing undesired unloading of the policy.

6.5.3 Policy List Concurrency and Synchronization

As the set of active policies may change at run-time, and the invocation of entry points is non-atomic, synchronization is required to prevent loading or unloading of policies while an entry point invocation is in progress, freezing the set of active policies for the duration. This is accomplished by means of a framework busy count: whenever an entry point is entered, the busy count is incremented; whenever it is exited, the busy count is decremented. While the busy count is elevated, policy list changes are not permitted, and threads attempting to modify the policy list will sleep until the list is not busy. The busy count is protected by a mutex, and a condition variable is used to wake up sleepers waiting on policy list modifications. One side effect of this synchronization model is that recursion into the MAC Framework from within a policy module is permitted, although not generally used.

Various optimizations are used to reduce the overhead of the busy count, including avoiding the full cost of incrementing and decrementing if the list is empty or contains only static entries (policies that are loaded before the system starts, and cannot be unloaded). A compile-time option is also provided which prevents any change in the set of loaded policies at run-time, which eliminates the mutex locking costs associated with supporting dynamically loaded and unloaded policies as synchronization is no longer required.

As the MAC Framework is not permitted to block in some entry points, a normal sleep lock cannot be used; as a result, it is possible for the load or unload attempt to block for a substantial period of time waiting for the framework to become idle.

6.5.4 Label Synchronization

As kernel objects of interest may generally be accessed from more than one thread at a time, and simultaneous entry of more than one thread into the MAC Framework is permitted, security attribute storage maintained by the MAC Framework is carefully synchronized. In general, existing kernel synchronization on kernel object data is used to protect MAC Framework security labels on the object: for example, MAC labels on sockets are protected using the existing socket mutex. Likewise, semantics for concurrent access are generally identical to those of the container objects: for credentials, copy-on-write semantics are maintained for label contents as with the remainder of the credential structure. The MAC Framework asserts necessary locks on objects when invoked with an object reference. Policy authors must be aware of these synchronization semantics, as they will sometimes limit the types of accesses permitted on labels: for example, when a read-only reference to a credential is passed to a policy via an entry point, only read operations are permitted on the label state attached to the credential.

6.5.5 Policy Synchronization and Concurrency

Policy modules must be written to assume that many kernel threads may simultaneously enter one more policy entry points due to the parallel and preemptive nature of the FreeBSD kernel. If the policy module makes use of mutable state, this may require the use of synchronization primitives within the policy to prevent inconsistent views on that state resulting in incorrect operation of the policy. Policies will generally be able to make use of existing FreeBSD synchronization primitives for this purpose, including mutexes, sleep locks, condition variables, and counting semaphores. However, policies should be written to employ these primitives carefully, respecting existing kernel lock orders, and recognizing that some entry points are not permitted to sleep, limiting the use of primitives in those entry points to mutexes and wakeup operations.

When policy modules call out to other kernel subsystems, they will generally need to release any in-policy locks in order to avoid violating the kernel lock order or risking lock recursion. This will maintain policy locks as leaf locks in the global lock order, helping to avoid deadlock.

6.5.6 Policy Registration

The MAC Framework maintains two lists of active policies: a static list, and a dynamic list. The lists differ only with regards to their locking semantics: an elevated reference count is not required to make use of the static list. When kernel modules containing MAC Framework policies are loaded, the policy module will use SYSINIT to invoke a registration function; when a policy module is unloaded, SYSINIT will likewise invoke a de-registration function. Registration may fail if a policy module is loaded more than once, if insufficient resources are available for the registration (for example, the policy might require labeling and insufficient labeling state might be available), or other policy prerequisites might not be met (some policies may only be loaded prior to boot). Likewise, de-registration may fail if a policy is flagged as not unloadable.

6.5.7 Entry Points

Kernel services interact with the MAC Framework in two ways: they invoke a series of APIs to notify the framework of relevant events, and they provide a policy-agnostic label structure pointer in security-relevant objects. The label pointer is maintained by the MAC Framework via label management entry points, and permits the Framework to offer a labeling service to policy modules through relatively non-invasive changes to the kernel subsystem maintaining the object. For example, label pointers have been added to processes, process credentials, sockets, pipes, vnodes, Mbufs, network interfaces, IP reassembly queues, and a variety of other security-relevant structures. Kernel services also invoke the MAC Framework when they perform important security decisions, permitting policy modules to augment those decisions based on their own criteria (possibly including data stored in security labels). Most of these security critical decisions will be explicit access control checks; however, some affect more general decision functions such as packet matching for sockets and label transition at program execution.

6.5.8 Policy Composition

When more than one policy module is loaded into the kernel at a time, the results of the policy modules will be composed by the framework using a composition operator. This operator is currently hard-coded, and requires that all active policies must approve a request for it to return success. As policies may return a variety of error conditions (success, access denied, object does not exist, ...), a precedence operator selects the resulting error from the set of errors returned by policies. In general, errors indicating that an object does not exist will be preferred to errors indicating that access to an object is denied. While it is not guaranteed that the resulting composition will be useful or secure, we have found that it is for many useful selections of policies. For example, traditional trusted systems often ship with two or more policies using a similar composition.

6.5.9 Labeling Support

As many interesting access control extensions rely on security labels on objects, the MAC Framework provides a set of policy-agnostic label management system calls covering a variety of user-exposed objects. Common label types include partition identifiers, sensitivity labels, integrity labels, compartments, domains, roles, and types. By policy agnostic, we mean that policy modules are able to completely define the semantics of meta-data associated with an object. Policy modules participate in the internalization and externalization of string-based labels provides by user applications, and can expose multiple label elements to applications if desired.

In-memory labels are stored in slab-allocated struct label, which consists of a fixed-length array of unions, each holding a void * pointer and a long. Policies registering for label storage will be assigned a "slot" identifier, which may be used to dereference the label storage. The semantics of the storage are left entirely up to the policy module: modules are provided with a variety of entry points associated with the kernel object life cycle, including initialization, association/creation, and destruction. Using these interfaces, it is possible to implement reference counting and other storage models. Direct access to the object structure is generally not required by policy modules to retrieve a label, as the MAC Framework generally passes both a pointer to the object and a direct pointer to the object's label into entry points. The primary exception to this rule is the process credential, which must be manually dereferenced to access the credential label. This may change in future revisions of the MAC Framework.

Initialization entry points frequently include a sleeping disposition flag indicating whether or not an initialization is permitted to sleep; if sleeping is not permitted, a failure may be returned to cancel allocation of the label (and hence object). This may occur, for example, in the network stack during interrupt handling, where sleeping is not permitted, or while the caller holds a mutex. Due to the performance cost of maintaining labels on in-flight network packets (Mbufs), policies must specifically declare a requirement that Mbuf labels be allocated. Dynamically loaded policies making use of labels must be able to handle the case where their init function has not been called on an object, as objects may already exist when the policy is loaded. The MAC Framework guarantees that uninitialized label slots will hold a 0 or NULL value, which policies may use to detect uninitialized values. However, as allocation of Mbuf labels is conditional, policies must also be able to handle a NULL label pointer for Mbufs if they have been loaded dynamically.

In the case of file system labels, special support is provided for the persistent storage of security labels in extended attributes. Where available, extended attribute transactions are used to permit consistent compound updates of security labels on vnodes--currently this support is present only in the UFS2 file system. Policy authors may choose to implement multilabel file system object labels using one (or more) extended attributes. For efficiency reasons, the vnode label (v_label) is a cache of any on-disk label; policies are able to load values into the cache when the vnode is instantiated, and update the cache as needed. As a result, the extended attribute need not be directly accessed with every access control check.

6.5.10 System Calls

The MAC Framework implements a number of system calls: most of these calls support the policy-agnostic label retrieval and manipulation APIs exposed to user applications.

The label management calls accept a label description structure, struct mac, which contains a series of MAC label elements. Each element contains a character string name, and character string value. Each policy will be given the chance to claim a particular element name, permitting policies to expose multiple independent elements if desired. Policy modules perform the internalization and externalization between kernel labels and user-provided labels via entry points, permitting a variety of semantics. Label management system calls are generally wrapped by user library functions to perform memory allocation and error handling, simplifying user applications that must manage labels.

The following MAC-related system calls are present in the FreeBSD kernel:

• mac_get_proc() may be used to retrieve the label of the current process.

• mac_set_proc() may be used to request a change in the label of the current process.

• mac_get_fd() may be used to retrieve the label of an object (file, socket, pipe, ...) referenced by a file descriptor.

• mac_get_file() may be used to retrieve the label of an object referenced by a file system path.

• mac_set_fd() may be used to request a change in the label of an object (file, socket, pipe, ...) referenced by a file descriptor.

• mac_set_file() may be used to request a change in the label of an object referenced by a file system path.

• mac_syscall() permits policy modules to create new system calls without modifying the system call table; it accepts a target policy name, operation number, and opaque argument for use by the policy.

• mac_get_pid() may be used to request the label of another process by process id.

• mac_get_link() is identical to mac_get_file(), only it will not follow a symbolic link if it is the final entry in the path, so may be used to retrieve the label on a symlink.

• mac_set_link() is identical to mac_set_file(), only it will not follow a symbolic link if it is the final entry in a path, so may be used to manipulate the label on a symlink.

• mac_execve() is identical to the execve() system call, only it also accepts a requested label to set the process label to when beginning execution of a new program. This change in label on execution is referred to as a "transition".

• mac_get_peer(), actually implemented via a socket option, retrieves the label of a remote peer on a socket, if available.

In addition to these system calls, the SIOCSIGMAC and SIOCSIFMAC network interface ioctls permit the labels on network interfaces to be retrieved and set.

6.6.1 Policy Declaration

Modules may be declared using the MAC_POLICY_SET() macro, which names the policy, provides a reference to the MAC entry point vector, provides load-time flags determining how the policy framework should handle the policy, and optionally requests the allocation of label state by the framework.

static struct mac_policy_ops mac_policy_ops =
{                                   
        .mpo_destroy = mac_policy_destroy,
        .mpo_init = mac_policy_init,
        .mpo_init_bpfdesc_label = mac_policy_init_bpfdesc_label,  
        .mpo_init_cred_label = mac_policy_init_label,
/* ... */
        .mpo_check_vnode_setutimes = mac_policy_check_vnode_setutimes,
        .mpo_check_vnode_stat = mac_policy_check_vnode_stat,
        .mpo_check_vnode_write = mac_policy_check_vnode_write,
};

The MAC policy entry point vector, mac_policy_ops in this example, associates functions defined in the module with specific entry points. A complete listing of available entry points and their prototypes may be found in the MAC entry point reference section. Of specific interest during module registration are the .mpo_destroy and .mpo_init entry points. .mpo_init will be invoked once a policy is successfully registered with the module framework but prior to any other entry points becoming active. This permits the policy to perform any policy-specific allocation and initialization, such as initialization of any data or locks. .mpo_destroy will be invoked when a policy module is unloaded to permit releasing of any allocated memory and destruction of locks. Currently, these two entry points are invoked with the MAC policy list mutex held to prevent any other entry points from being invoked: this will be changed, but in the mean time, policies should be careful about what kernel primitives they invoke so as to avoid lock ordering or sleeping problems.

The policy declaration's module name field exists so that the module may be uniquely identified for the purposes of module dependencies. An appropriate string should be selected. The full string name of the policy is displayed to the user via the kernel log during load and unload events, and also exported when providing status information to userland processes.

6.6.2 Policy Flags

The policy declaration flags field permits the module to provide the framework with information about its capabilities at the time the module is loaded. Currently, three flags are defined:

6.6.3 Policy Entry Points

Four classes of entry points are offered to policies registered with the framework: entry points associated with the registration and management of policies, entry points denoting initialization, creation, destruction, and other life cycle events for kernel objects, events associated with access control decisions that the policy module may influence, and calls associated with the management of labels on objects. In addition, a mac_syscall() entry point is provided so that policies may extend the kernel interface without registering new system calls.

Policy module writers should be aware of the kernel locking strategy, as well as what object locks are available during which entry points. Writers should attempt to avoid deadlock scenarios by avoiding grabbing non-leaf locks inside of entry points, and also follow the locking protocol for object access and modification. In particular, writers should be aware that while necessary locks to access objects and their labels are generally held, sufficient locks to modify an object or its label may not be present for all entry points. Locking information for arguments is documented in the MAC framework entry point document.

Policy entry points will pass a reference to the object label along with the object itself. This permits labeled policies to be unaware of the internals of the object yet still make decisions based on the label. The exception to this is the process credential, which is assumed to be understood by policies as a first class security object in the kernel.

6.7.1 General-Purpose Module Entry Points

6.7.2 Label Operations

6.7.3 Label Events

This class of entry points is used by the MAC framework to permit policies to maintain label information on kernel objects. For each labeled kernel object of interest to a MAC policy, entry points may be registered for relevant life cycle events. All objects implement initialization, creation, and destruction hooks. Some objects will also implement relabeling, allowing user processes to change the labels on objects. Some objects will also implement object-specific events, such as label events associated with IP reassembly. A typical labeled object will have the following life cycle of entry points:

Label initialization          o
(object-specific wait)         \
Label creation                  o
                                 \
Relabel events,                   o--<--.
Various object-specific,          |     |
Access control events             ~-->--o
                                         \
Label destruction                         o

Label initialization permits policies to allocate memory and set initial values for labels without context for the use of the object. The label slot allocated to a policy will be zeroed by default, so some policies may not need to perform initialization.

Label creation occurs when the kernel structure is associated with an actual kernel object. For example, Mbufs may be allocated and remain unused in a pool until they are required. mbuf allocation causes label initialization on the mbuf to take place, but mbuf creation occurs when the mbuf is associated with a datagram. Typically, context will be provided for a creation event, including the circumstances of the creation, and labels of other relevant objects in the creation process. For example, when an mbuf is created from a socket, the socket and its label will be presented to registered policies in addition to the new mbuf and its label. Memory allocation in creation events is discouraged, as it may occur in performance sensitive ports of the kernel; in addition, creation calls are not permitted to fail so a failure to allocate memory cannot be reported.

Object specific events do not generally fall into the other broad classes of label events, but will generally provide an opportunity to modify or update the label on an object based on additional context. For example, the label on an IP fragment reassembly queue may be updated during the MAC_UPDATE_IPQ entry point as a result of the acceptance of an additional mbuf to that queue.

Access control events are discussed in detail in the following section.

Label destruction permits policies to release storage or state associated with a label during its association with an object so that the kernel data structures supporting the object may be reused or released.

In addition to labels associated with specific kernel objects, an additional class of labels exists: temporary labels. These labels are used to store update information submitted by user processes. These labels are initialized and destroyed as with other label types, but the creation event is MAC_INTERNALIZE, which accepts a user label to be converted to an in-kernel representation.

6.7.4 Access Control Checks

Access control entry points permit policy modules to influence access control decisions made by the kernel. Generally, although not always, arguments to an access control entry point will include one or more authorizing credentials, information (possibly including a label) for any other objects involved in the operation. An access control entry point may return 0 to permit the operation, or an errno(2) error value. The results of invoking the entry point across various registered policy modules will be composed as follows: if all modules permit the operation to succeed, success will be returned. If one or modules returns a failure, a failure will be returned. If more than one module returns a failure, the errno value to return to the user will be selected using the following precedence, implemented by the error_select() function in kern_mac.c:

If none of the error values returned by all modules are listed in the precedence chart then an arbitrarily selected value from the set will be returned. In general, the rules provide precedence to errors in the following order: kernel failures, invalid arguments, object not present, access not permitted, other.