2075 lines
52 KiB
C
2075 lines
52 KiB
C
/*-
|
|
* Copyright (c) 2008-2009 Apple Inc.
|
|
* All rights reserved.
|
|
*
|
|
* Redistribution and use in source and binary forms, with or without
|
|
* modification, are permitted provided that the following conditions
|
|
* are met:
|
|
* 1. Redistributions of source code must retain the above copyright
|
|
* notice, this list of conditions and the following disclaimer.
|
|
* 2. Redistributions in binary form must reproduce the above copyright
|
|
* notice, this list of conditions and the following disclaimer in the
|
|
* documentation and/or other materials provided with the distribution.
|
|
* 3. Neither the name of Apple Inc. ("Apple") nor the names of
|
|
* its contributors may be used to endorse or promote products derived
|
|
* from this software without specific prior written permission.
|
|
*
|
|
* THIS SOFTWARE IS PROVIDED BY APPLE AND ITS CONTRIBUTORS "AS IS" AND
|
|
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
|
|
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
|
|
* ARE DISCLAIMED. IN NO EVENT SHALL APPLE OR ITS CONTRIBUTORS BE LIABLE FOR
|
|
* ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
|
|
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
|
|
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
|
|
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
|
|
* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING
|
|
* IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
|
|
* POSSIBILITY OF SUCH DAMAGE.
|
|
*/
|
|
|
|
#include <stdarg.h>
|
|
|
|
#include <sys/kernel.h>
|
|
#include <sys/fcntl.h>
|
|
#include <sys/kauth.h>
|
|
#include <sys/conf.h>
|
|
#include <sys/poll.h>
|
|
#include <sys/priv.h>
|
|
#include <sys/queue.h>
|
|
#include <sys/signalvar.h>
|
|
#include <sys/syscall.h>
|
|
#include <sys/sysent.h>
|
|
#include <sys/sysproto.h>
|
|
#include <sys/systm.h>
|
|
#include <sys/ucred.h>
|
|
#include <sys/user.h>
|
|
|
|
#include <miscfs/devfs/devfs.h>
|
|
|
|
#include <libkern/OSAtomic.h>
|
|
|
|
#include <bsm/audit.h>
|
|
#include <bsm/audit_internal.h>
|
|
#include <bsm/audit_kevents.h>
|
|
|
|
#include <security/audit/audit.h>
|
|
#include <security/audit/audit_bsd.h>
|
|
#include <security/audit/audit_ioctl.h>
|
|
#include <security/audit/audit_private.h>
|
|
|
|
#include <vm/vm_protos.h>
|
|
#include <mach/mach_port.h>
|
|
#include <kern/audit_sessionport.h>
|
|
|
|
#include <libkern/OSDebug.h>
|
|
|
|
/*
|
|
* Audit Session Entry. This is treated as an object with public and private
|
|
* data. The se_auinfo field is the only information that is public and
|
|
* needs to be the first entry.
|
|
*/
|
|
struct au_sentry {
|
|
auditinfo_addr_t se_auinfo; /* Public audit session data. */
|
|
#define se_asid se_auinfo.ai_asid
|
|
#define se_auid se_auinfo.ai_auid
|
|
#define se_mask se_auinfo.ai_mask
|
|
#define se_termid se_auinfo.ai_termid
|
|
#define se_flags se_auinfo.ai_flags
|
|
|
|
long se_refcnt; /* Reference count. */
|
|
long se_procnt; /* Processes in session. */
|
|
ipc_port_t se_port; /* Session port. */
|
|
union {
|
|
LIST_ENTRY(au_sentry) se_link; /* Hash bucket link list (1) */
|
|
struct smr_node se_smr_node;
|
|
};
|
|
};
|
|
typedef struct au_sentry au_sentry_t;
|
|
|
|
#define AU_SENTRY_PTR(aia_p) ((au_sentry_t *)(aia_p))
|
|
|
|
/*
|
|
* The default au_sentry/auditinfo_addr entry for ucred.
|
|
*/
|
|
|
|
static au_sentry_t audit_default_se = {
|
|
.se_auinfo = {
|
|
.ai_auid = AU_DEFAUDITID,
|
|
.ai_asid = AU_DEFAUDITSID,
|
|
.ai_termid = { .at_type = AU_IPv4, },
|
|
},
|
|
.se_refcnt = 1,
|
|
.se_procnt = 1,
|
|
};
|
|
|
|
struct auditinfo_addr * const audit_default_aia_p = &audit_default_se.se_auinfo;
|
|
|
|
/* Copied from <ipc/ipc_object.h> */
|
|
#define IPC_OBJECT_COPYIN_FLAGS_ALLOW_IMMOVABLE_SEND 0x1
|
|
kern_return_t ipc_object_copyin(ipc_space_t, mach_port_name_t,
|
|
mach_msg_type_name_t, ipc_port_t *, mach_port_context_t, mach_msg_guard_flags_t *, uint32_t);
|
|
void ipc_port_release_send(ipc_port_t);
|
|
|
|
#if CONFIG_AUDIT
|
|
|
|
|
|
/*
|
|
* Currently the hash table is a fixed size.
|
|
*/
|
|
#define HASH_TABLE_SIZE 97
|
|
#define HASH_ASID(asid) (audit_session_hash(asid) % HASH_TABLE_SIZE)
|
|
|
|
static struct rwlock se_entry_lck; /* (1) lock for se_link above */
|
|
|
|
LIST_HEAD(au_sentry_head, au_sentry);
|
|
static struct au_sentry_head *au_sentry_bucket = NULL;
|
|
|
|
#define AU_HISTORY_LOGGING 0
|
|
#if AU_HISTORY_LOGGING
|
|
typedef enum au_history_event {
|
|
AU_HISTORY_EVENT_UNKNOWN = 0,
|
|
AU_HISTORY_EVENT_REF = 1,
|
|
AU_HISTORY_EVENT_UNREF = 2,
|
|
AU_HISTORY_EVENT_BIRTH = 3,
|
|
AU_HISTORY_EVENT_DEATH = 4,
|
|
AU_HISTORY_EVENT_FIND = 5
|
|
} au_history_event_t;
|
|
|
|
#define AU_HISTORY_MAX_STACK_DEPTH 8
|
|
|
|
struct au_history {
|
|
struct au_sentry *ptr;
|
|
struct au_sentry se;
|
|
void *stack[AU_HISTORY_MAX_STACK_DEPTH];
|
|
unsigned int stack_depth;
|
|
au_history_event_t event;
|
|
};
|
|
|
|
static struct au_history *au_history;
|
|
static size_t au_history_size = 65536;
|
|
static unsigned int au_history_index;
|
|
|
|
static inline unsigned int
|
|
au_history_entries(void)
|
|
{
|
|
if (au_history_index >= au_history_size) {
|
|
return au_history_size;
|
|
} else {
|
|
return au_history_index;
|
|
}
|
|
}
|
|
|
|
static inline void
|
|
au_history_record(au_sentry_t *se, au_history_event_t event)
|
|
{
|
|
struct au_history *p;
|
|
unsigned int i;
|
|
|
|
i = OSAddAtomic(1, &au_history_index);
|
|
p = &au_history[i % au_history_size];
|
|
|
|
bzero(p, sizeof(*p));
|
|
p->event = event;
|
|
bcopy(se, &p->se, sizeof(p->se));
|
|
p->stack_depth = OSBacktrace(&p->stack[0], AU_HISTORY_MAX_STACK_DEPTH);
|
|
p->ptr = se;
|
|
}
|
|
#else
|
|
#define au_history_record(se, event) do {} while (0)
|
|
#endif
|
|
|
|
MALLOC_DEFINE(M_AU_SESSION, "audit_session", "Audit session data");
|
|
|
|
static void audit_ref_session(au_sentry_t *se);
|
|
static void audit_unref_session(au_sentry_t *se);
|
|
|
|
static void audit_session_event(int event, auditinfo_addr_t *aia_p);
|
|
|
|
/*
|
|
* Audit session device.
|
|
*/
|
|
|
|
static MALLOC_DEFINE(M_AUDIT_SDEV, "audit_sdev", "Audit sdevs");
|
|
static MALLOC_DEFINE(M_AUDIT_SDEV_ENTRY, "audit_sdevent",
|
|
"Audit sdev entries and buffers");
|
|
|
|
/*
|
|
* Default audit sdev buffer parameters.
|
|
*/
|
|
#define AUDIT_SDEV_QLIMIT_DEFAULT 128
|
|
#define AUDIT_SDEV_QLIMIT_MIN 1
|
|
#define AUDIT_SDEV_QLIMIT_MAX 1024
|
|
|
|
/*
|
|
* Entry structure.
|
|
*/
|
|
struct audit_sdev_entry {
|
|
void *ase_record;
|
|
u_int ase_record_len;
|
|
TAILQ_ENTRY(audit_sdev_entry) ase_queue;
|
|
};
|
|
|
|
/*
|
|
* Per audit sdev structure.
|
|
*/
|
|
|
|
struct audit_sdev {
|
|
int asdev_open;
|
|
|
|
#define AUDIT_SDEV_ASYNC 0x00000001
|
|
#define AUDIT_SDEV_NBIO 0x00000002
|
|
|
|
#define AUDIT_SDEV_ALLSESSIONS 0x00010000
|
|
u_int asdev_flags;
|
|
|
|
struct selinfo asdev_selinfo;
|
|
pid_t asdev_sigio;
|
|
|
|
au_id_t asdev_auid;
|
|
au_asid_t asdev_asid;
|
|
|
|
/* Per-sdev mutex for most fields in this struct. */
|
|
struct mtx asdev_mtx;
|
|
|
|
/*
|
|
* Per-sdev sleep lock serializing user-generated reads and
|
|
* flushes. uiomove() is called to copy out the current head
|
|
* record's data whie the record remains in the queue, so we
|
|
* prevent other threads from removing it using this lock.
|
|
*/
|
|
struct slck asdev_sx;
|
|
|
|
/*
|
|
* Condition variable to signal when data has been delivered to
|
|
* a sdev.
|
|
*/
|
|
struct cv asdev_cv;
|
|
|
|
/* Count and bound of records in the queue. */
|
|
u_int asdev_qlen;
|
|
u_int asdev_qlimit;
|
|
|
|
/* The number of bytes of data across all records. */
|
|
u_int asdev_qbyteslen;
|
|
|
|
/*
|
|
* The amount read so far of the first record in the queue.
|
|
* (The number of bytes available for reading in the queue is
|
|
* qbyteslen - qoffset.)
|
|
*/
|
|
u_int asdev_qoffset;
|
|
|
|
/*
|
|
* Per-sdev operation statistics.
|
|
*/
|
|
u_int64_t asdev_inserts; /* Records added. */
|
|
u_int64_t asdev_reads; /* Records read. */
|
|
u_int64_t asdev_drops; /* Records dropped. */
|
|
|
|
/*
|
|
* Current pending record list. This is protected by a
|
|
* combination of asdev_mtx and asdev_sx. Note that both
|
|
* locks are required to remove a record from the head of the
|
|
* queue, as an in-progress read may sleep while copying and,
|
|
* therefore, cannot hold asdev_mtx.
|
|
*/
|
|
TAILQ_HEAD(, audit_sdev_entry) asdev_queue;
|
|
|
|
/* Global sdev list. */
|
|
TAILQ_ENTRY(audit_sdev) asdev_list;
|
|
};
|
|
|
|
#define AUDIT_SDEV_LOCK(asdev) mtx_lock(&(asdev)->asdev_mtx)
|
|
#define AUDIT_SDEV_LOCK_ASSERT(asdev) mtx_assert(&(asdev)->asdev_mtx, \
|
|
MA_OWNED)
|
|
#define AUDIT_SDEV_LOCK_DESTROY(asdev) mtx_destroy(&(asdev)->asdev_mtx)
|
|
#define AUDIT_SDEV_LOCK_INIT(asdev) mtx_init(&(asdev)->asdev_mtx, \
|
|
"audit_sdev_mtx", NULL, MTX_DEF)
|
|
#define AUDIT_SDEV_UNLOCK(asdev) mtx_unlock(&(asdev)->asdev_mtx)
|
|
#define AUDIT_SDEV_MTX(asdev) (&(asdev)->asdev_mtx)
|
|
|
|
#define AUDIT_SDEV_SX_LOCK_DESTROY(asd) slck_destroy(&(asd)->asdev_sx)
|
|
#define AUDIT_SDEV_SX_LOCK_INIT(asd) slck_init(&(asd)->asdev_sx, \
|
|
"audit_sdev_sx")
|
|
#define AUDIT_SDEV_SX_XLOCK_ASSERT(asd) slck_assert(&(asd)->asdev_sx, \
|
|
SA_XLOCKED)
|
|
#define AUDIT_SDEV_SX_XLOCK_SIG(asd) slck_lock_sig(&(asd)->asdev_sx)
|
|
#define AUDIT_SDEV_SX_XUNLOCK(asd) slck_unlock(&(asd)->asdev_sx)
|
|
|
|
/*
|
|
* Cloning variables and constants.
|
|
*/
|
|
#define AUDIT_SDEV_NAME "auditsessions"
|
|
#define MAX_AUDIT_SDEVS 32
|
|
|
|
static int audit_sdev_major;
|
|
static void *devnode;
|
|
|
|
/*
|
|
* Global list of audit sdevs. The list is protected by a rw lock.
|
|
* Individaul record queues are protected by per-sdev locks. These
|
|
* locks synchronize between threads walking the list to deliver to
|
|
* individual sdevs and adds/removes of sdevs.
|
|
*/
|
|
static TAILQ_HEAD(, audit_sdev) audit_sdev_list;
|
|
static struct rwlock audit_sdev_lock;
|
|
|
|
#define AUDIT_SDEV_LIST_LOCK_INIT() rw_init(&audit_sdev_lock, \
|
|
"audit_sdev_list_lock")
|
|
#define AUDIT_SDEV_LIST_RLOCK() rw_rlock(&audit_sdev_lock)
|
|
#define AUDIT_SDEV_LIST_RUNLOCK() rw_runlock(&audit_sdev_lock)
|
|
#define AUDIT_SDEV_LIST_WLOCK() rw_wlock(&audit_sdev_lock)
|
|
#define AUDIT_SDEV_LIST_WLOCK_ASSERT() rw_assert(&audit_sdev_lock, \
|
|
RA_WLOCKED)
|
|
#define AUDIT_SDEV_LIST_WUNLOCK() rw_wunlock(&audit_sdev_lock)
|
|
|
|
/*
|
|
* dev_t doesn't have a pointer for "softc" data so we have to keep track of
|
|
* it with the following global array (indexed by the minor number).
|
|
*
|
|
* XXX We may want to dynamically grow this as need.
|
|
*/
|
|
static struct audit_sdev *audit_sdev_dtab[MAX_AUDIT_SDEVS];
|
|
|
|
/*
|
|
* Special device methods and definition.
|
|
*/
|
|
static open_close_fcn_t audit_sdev_open;
|
|
static open_close_fcn_t audit_sdev_close;
|
|
static read_write_fcn_t audit_sdev_read;
|
|
static ioctl_fcn_t audit_sdev_ioctl;
|
|
static select_fcn_t audit_sdev_poll;
|
|
|
|
static const struct cdevsw audit_sdev_cdevsw = {
|
|
.d_open = audit_sdev_open,
|
|
.d_close = audit_sdev_close,
|
|
.d_read = audit_sdev_read,
|
|
.d_write = eno_rdwrt,
|
|
.d_ioctl = audit_sdev_ioctl,
|
|
.d_stop = eno_stop,
|
|
.d_reset = eno_reset,
|
|
.d_ttys = NULL,
|
|
.d_select = audit_sdev_poll,
|
|
.d_mmap = eno_mmap,
|
|
.d_strategy = eno_strat,
|
|
.d_type = 0
|
|
};
|
|
|
|
/*
|
|
* Global statistics on audit sdevs.
|
|
*/
|
|
static int audit_sdev_count; /* Current number of sdevs. */
|
|
static u_int64_t audit_sdev_ever; /* Sdevs ever allocated. */
|
|
static u_int64_t audit_sdev_records; /* Total records seen. */
|
|
static u_int64_t audit_sdev_drops; /* Global record drop count. */
|
|
|
|
static int audit_sdev_init(void);
|
|
|
|
#define AUDIT_SENTRY_RWLOCK_INIT() rw_init(&se_entry_lck, \
|
|
"se_entry_lck")
|
|
#define AUDIT_SENTRY_RLOCK() rw_rlock(&se_entry_lck)
|
|
#define AUDIT_SENTRY_WLOCK() rw_wlock(&se_entry_lck)
|
|
#define AUDIT_SENTRY_RWLOCK_ASSERT() rw_assert(&se_entry_lck, RA_LOCKED)
|
|
#define AUDIT_SENTRY_RUNLOCK() rw_runlock(&se_entry_lck)
|
|
#define AUDIT_SENTRY_WUNLOCK() rw_wunlock(&se_entry_lck)
|
|
|
|
/*
|
|
* Access control on the auditinfo_addr.ai_flags member.
|
|
*/
|
|
static const uint64_t audit_session_superuser_set_sflags_mask = AU_SESSION_FLAG_HAS_GRAPHIC_ACCESS | AU_SESSION_FLAG_HAS_CONSOLE_ACCESS | AU_SESSION_FLAG_HAS_AUTHENTICATED;
|
|
static const uint64_t audit_session_superuser_clear_sflags_mask = AU_SESSION_FLAG_HAS_GRAPHIC_ACCESS | AU_SESSION_FLAG_HAS_CONSOLE_ACCESS | AU_SESSION_FLAG_HAS_AUTHENTICATED;
|
|
static const uint64_t audit_session_member_set_sflags_mask = 0;
|
|
static const uint64_t audit_session_member_clear_sflags_mask = AU_SESSION_FLAG_HAS_AUTHENTICATED;
|
|
|
|
#define AUDIT_SESSION_DEBUG 0
|
|
#if AUDIT_SESSION_DEBUG
|
|
/*
|
|
* The following is debugging code that can be used to get a snapshot of the
|
|
* session state. The audit session information is read out using sysctl:
|
|
*
|
|
* error = sysctlbyname("kern.audit_session_debug", buffer_ptr, &buffer_len,
|
|
* NULL, 0);
|
|
*/
|
|
#include <kern/kalloc.h>
|
|
|
|
/*
|
|
* The per session record structure for the snapshot data.
|
|
*/
|
|
struct au_sentry_debug {
|
|
auditinfo_addr_t se_auinfo;
|
|
int64_t se_refcnt; /* refereence count */
|
|
int64_t se_procnt; /* process count */
|
|
int64_t se_ptcnt; /* process count from
|
|
* proc table */
|
|
};
|
|
typedef struct au_sentry_debug au_sentry_debug_t;
|
|
|
|
static int audit_sysctl_session_debug(struct sysctl_oid *oidp, void *arg1,
|
|
int arg2, struct sysctl_req *req);
|
|
|
|
SYSCTL_PROC(_kern, OID_AUTO, audit_session_debug, CTLFLAG_RD | CTLFLAG_LOCKED,
|
|
NULL, 0, audit_sysctl_session_debug, "S,audit_session_debug",
|
|
"Current session debug info for auditing.");
|
|
|
|
/*
|
|
* Callouts for proc_interate() which is used to reconcile the audit session
|
|
* proc state information with the proc table. We get everything we need
|
|
* in the filterfn while the proc_lock() is held so we really don't need the
|
|
* callout() function.
|
|
*/
|
|
static int
|
|
audit_session_debug_callout(__unused proc_t p, __unused void *arg)
|
|
{
|
|
return PROC_RETURNED_DONE;
|
|
}
|
|
|
|
static int
|
|
audit_session_debug_filterfn(proc_t p, void *st)
|
|
{
|
|
kauth_cred_t cred = kauth_cred_get();
|
|
auditinfo_addr_t *aia_p = cred->cr_audit.as_aia_p;
|
|
au_sentry_debug_t *sed_tab = (au_sentry_debug_t *) st;
|
|
au_sentry_debug_t *sdtp;
|
|
au_sentry_t *se;
|
|
|
|
if (IS_VALID_SESSION(aia_p)) {
|
|
sdtp = &sed_tab[0];
|
|
do {
|
|
if (aia_p->ai_asid == sdtp->se_asid) {
|
|
sdtp->se_ptcnt++;
|
|
|
|
/* Do some santy checks. */
|
|
se = AU_SENTRY_PTR(aia_p);
|
|
if (se->se_refcnt != sdtp->se_refcnt) {
|
|
sdtp->se_refcnt =
|
|
(int64_t)se->se_refcnt;
|
|
}
|
|
if (se->se_procnt != sdtp->se_procnt) {
|
|
sdtp->se_procnt =
|
|
(int64_t)se->se_procnt;
|
|
}
|
|
break;
|
|
}
|
|
sdtp++;
|
|
} while (sdtp->se_asid != 0 && sdtp->se_auid != 0);
|
|
} else {
|
|
/* add it to the default sesison */
|
|
sed_tab->se_ptcnt++;
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Copy out the session debug info via the sysctl interface.
|
|
*
|
|
*/
|
|
static int
|
|
audit_sysctl_session_debug(__unused struct sysctl_oid *oidp,
|
|
__unused void *arg1, __unused int arg2, struct sysctl_req *req)
|
|
{
|
|
au_sentry_t *se;
|
|
au_sentry_debug_t *sed_tab, *next_sed;
|
|
int i, entry_cnt = 0;
|
|
size_t sz;
|
|
int err = 0;
|
|
|
|
/*
|
|
* This provides a read-only node.
|
|
*/
|
|
if (req->newptr != USER_ADDR_NULL) {
|
|
return EPERM;
|
|
}
|
|
|
|
/*
|
|
* Walk the audit session hash table to determine the size.
|
|
*/
|
|
AUDIT_SENTRY_RLOCK();
|
|
for (i = 0; i < HASH_TABLE_SIZE; i++) {
|
|
LIST_FOREACH(se, &au_sentry_bucket[i], se_link)
|
|
if (se != NULL) {
|
|
entry_cnt++;
|
|
}
|
|
}
|
|
|
|
entry_cnt++; /* add one for the default entry */
|
|
/*
|
|
* If just querying then return the space required. There is an
|
|
* obvious race condition here so we just fudge this by 3 in case
|
|
* the audit session table grows.
|
|
*/
|
|
if (req->oldptr == USER_ADDR_NULL) {
|
|
req->oldidx = (entry_cnt + 3) * sizeof(au_sentry_debug_t);
|
|
AUDIT_SENTRY_RUNLOCK();
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Alloc a temporary buffer.
|
|
*/
|
|
if (req->oldlen < (entry_cnt * sizeof(au_sentry_debug_t))) {
|
|
AUDIT_SENTRY_RUNLOCK();
|
|
return ENOMEM;
|
|
}
|
|
/*
|
|
* We hold the lock over the alloc since we don't want the table to
|
|
* grow on us. Therefore, use the non-blocking version of kalloc().
|
|
*/
|
|
sed_tab = kalloc_data(entry_cnt * sizeof(au_sentry_debug_t),
|
|
Z_NOWAIT | Z_ZERO);
|
|
if (sed_tab == NULL) {
|
|
AUDIT_SENTRY_RUNLOCK();
|
|
return ENOMEM;
|
|
}
|
|
|
|
/*
|
|
* Walk the audit session hash table and build the record array.
|
|
*/
|
|
sz = 0;
|
|
next_sed = sed_tab;
|
|
/* add the first entry for processes not tracked in sessions. */
|
|
bcopy(audit_default_aia_p, &next_sed->se_auinfo, sizeof(au_sentry_t));
|
|
next_sed->se_refcnt = (int64_t)audit_default_se.se_refcnt;
|
|
next_sed->se_procnt = (int64_t)audit_default_se.se_procnt;
|
|
next_sed++;
|
|
sz += sizeof(au_sentry_debug_t);
|
|
for (i = 0; i < HASH_TABLE_SIZE; i++) {
|
|
LIST_FOREACH(se, &au_sentry_bucket[i], se_link) {
|
|
if (se != NULL) {
|
|
next_sed->se_auinfo = se->se_auinfo;
|
|
next_sed->se_refcnt = (int64_t)se->se_refcnt;
|
|
next_sed->se_procnt = (int64_t)se->se_procnt;
|
|
next_sed++;
|
|
sz += sizeof(au_sentry_debug_t);
|
|
}
|
|
}
|
|
}
|
|
AUDIT_SENTRY_RUNLOCK();
|
|
|
|
/* Reconcile with the process table. */
|
|
proc_iterate(PROC_ALLPROCLIST | PROC_ZOMBPROCLIST,
|
|
audit_session_debug_callout, NULL,
|
|
audit_session_debug_filterfn, (void *)&sed_tab[0]);
|
|
|
|
|
|
req->oldlen = sz;
|
|
err = SYSCTL_OUT(req, sed_tab, sz);
|
|
kfree_data(sed_tab, entry_cnt * sizeof(au_sentry_debug_t));
|
|
|
|
return err;
|
|
}
|
|
|
|
#endif /* AUDIT_SESSION_DEBUG */
|
|
|
|
/*
|
|
* Create and commit a session audit event. The proc and se arguments needs to
|
|
* be that of the subject and not necessarily the current process.
|
|
*/
|
|
static void
|
|
audit_session_event(int event, auditinfo_addr_t *aia_p)
|
|
{
|
|
struct kaudit_record *ar;
|
|
|
|
KASSERT(AUE_SESSION_START == event || AUE_SESSION_UPDATE == event ||
|
|
AUE_SESSION_END == event || AUE_SESSION_CLOSE == event,
|
|
("audit_session_event: invalid event: %d", event));
|
|
|
|
if (NULL == aia_p) {
|
|
return;
|
|
}
|
|
|
|
/*
|
|
* Create a new audit record. The record will contain the subject
|
|
* ruid, rgid, egid, pid, auid, asid, amask, and term_addr
|
|
* (implicitly added by audit_new).
|
|
*/
|
|
ar = audit_new(event, PROC_NULL, /* Not used */ NULL);
|
|
if (NULL == ar) {
|
|
return;
|
|
}
|
|
|
|
/*
|
|
* Audit session events are always generated because they are used
|
|
* by some userland consumers so just set the preselect flag.
|
|
*/
|
|
ar->k_ar_commit |= AR_PRESELECT_FILTER;
|
|
|
|
/*
|
|
* Populate the subject information. Note that the ruid, rgid,
|
|
* egid, and pid values are incorrect. We only need the auditinfo_addr
|
|
* information.
|
|
*/
|
|
ar->k_ar.ar_subj_ruid = 0;
|
|
ar->k_ar.ar_subj_rgid = 0;
|
|
ar->k_ar.ar_subj_egid = 0;
|
|
ar->k_ar.ar_subj_pid = 0;
|
|
ar->k_ar.ar_subj_auid = aia_p->ai_auid;
|
|
ar->k_ar.ar_subj_asid = aia_p->ai_asid;
|
|
bcopy(&aia_p->ai_termid, &ar->k_ar.ar_subj_term_addr,
|
|
sizeof(struct au_tid_addr));
|
|
|
|
/* Add the audit masks to the record. */
|
|
ar->k_ar.ar_arg_amask.am_success = aia_p->ai_mask.am_success;
|
|
ar->k_ar.ar_arg_amask.am_failure = aia_p->ai_mask.am_failure;
|
|
ARG_SET_VALID(ar, ARG_AMASK);
|
|
|
|
/* Add the audit session flags to the record. */
|
|
ar->k_ar.ar_arg_value64 = aia_p->ai_flags;
|
|
ARG_SET_VALID(ar, ARG_VALUE64);
|
|
|
|
|
|
/* Commit the record to the queue. */
|
|
audit_commit(ar, 0, 0);
|
|
}
|
|
|
|
/*
|
|
* Hash the audit session ID using a simple 32-bit mix.
|
|
*/
|
|
static inline uint32_t
|
|
audit_session_hash(au_asid_t asid)
|
|
{
|
|
uint32_t a = (uint32_t) asid;
|
|
|
|
a = (a - (a << 6)) ^ (a >> 17);
|
|
a = (a - (a << 9)) ^ (a << 4);
|
|
a = (a - (a << 3)) ^ (a << 10);
|
|
a = a ^ (a >> 15);
|
|
|
|
return a;
|
|
}
|
|
|
|
/*
|
|
* Do an hash lookup and find the session entry for a given ASID. Return NULL
|
|
* if not found. If the session is found then audit_session_find takes a
|
|
* reference.
|
|
*/
|
|
static au_sentry_t *
|
|
audit_session_find(au_asid_t asid)
|
|
{
|
|
uint32_t hkey;
|
|
au_sentry_t *found_se;
|
|
|
|
AUDIT_SENTRY_RWLOCK_ASSERT();
|
|
|
|
hkey = HASH_ASID(asid);
|
|
|
|
LIST_FOREACH(found_se, &au_sentry_bucket[hkey], se_link)
|
|
if (found_se->se_asid == asid) {
|
|
au_history_record(found_se, AU_HISTORY_EVENT_FIND);
|
|
audit_ref_session(found_se);
|
|
return found_se;
|
|
}
|
|
return NULL;
|
|
}
|
|
|
|
static void
|
|
audit_session_free(smr_node_t node)
|
|
{
|
|
au_sentry_t *se = __container_of(node, au_sentry_t, se_smr_node);
|
|
|
|
kfree_type(au_sentry_t, se);
|
|
}
|
|
|
|
/*
|
|
* Remove the given audit_session entry from the hash table.
|
|
*/
|
|
static void
|
|
audit_session_remove(au_sentry_t *se)
|
|
{
|
|
uint32_t hkey;
|
|
au_sentry_t *found_se, *tmp_se;
|
|
|
|
au_history_record(se, AU_HISTORY_EVENT_DEATH);
|
|
KASSERT(se->se_refcnt == 0, ("audit_session_remove: ref count != 0"));
|
|
KASSERT(se != &audit_default_se,
|
|
("audit_session_remove: removing default session"));
|
|
|
|
hkey = HASH_ASID(se->se_asid);
|
|
|
|
AUDIT_SENTRY_WLOCK();
|
|
/*
|
|
* Check and see if someone got a reference before we got the lock.
|
|
*/
|
|
if (se->se_refcnt != 0) {
|
|
AUDIT_SENTRY_WUNLOCK();
|
|
return;
|
|
}
|
|
|
|
audit_session_portdestroy(&se->se_port);
|
|
LIST_FOREACH_SAFE(found_se, &au_sentry_bucket[hkey], se_link, tmp_se) {
|
|
if (found_se == se) {
|
|
/*
|
|
* Generate an audit event to notify userland of the
|
|
* session close.
|
|
*/
|
|
audit_session_event(AUE_SESSION_CLOSE,
|
|
&found_se->se_auinfo);
|
|
|
|
LIST_REMOVE(found_se, se_link);
|
|
AUDIT_SENTRY_WUNLOCK();
|
|
smr_call(&smr_proc_task, &found_se->se_smr_node,
|
|
sizeof(found_se), audit_session_free);
|
|
return;
|
|
}
|
|
}
|
|
AUDIT_SENTRY_WUNLOCK();
|
|
}
|
|
|
|
/*
|
|
* Reference the session by incrementing the sentry ref count.
|
|
*/
|
|
static void
|
|
audit_ref_session(au_sentry_t *se)
|
|
{
|
|
long old_val;
|
|
|
|
if (se == NULL || se == &audit_default_se) {
|
|
return;
|
|
}
|
|
|
|
au_history_record(se, AU_HISTORY_EVENT_REF);
|
|
|
|
old_val = OSAddAtomicLong(1, &se->se_refcnt);
|
|
KASSERT(old_val < 100000,
|
|
("audit_ref_session: Too many references on session."));
|
|
}
|
|
|
|
/*
|
|
* Decrement the sentry ref count and remove the session entry if last one.
|
|
*/
|
|
static void
|
|
audit_unref_session(au_sentry_t *se)
|
|
{
|
|
long old_val;
|
|
|
|
if (se == NULL || se == &audit_default_se) {
|
|
return;
|
|
}
|
|
|
|
au_history_record(se, AU_HISTORY_EVENT_UNREF);
|
|
|
|
old_val = OSAddAtomicLong(-1, &se->se_refcnt);
|
|
if (old_val == 1) {
|
|
audit_session_remove(se);
|
|
}
|
|
KASSERT(old_val > 0,
|
|
("audit_unref_session: Too few references on session."));
|
|
}
|
|
|
|
/*
|
|
* Increment the process count in the session.
|
|
*/
|
|
static void
|
|
audit_inc_procount(au_sentry_t *se)
|
|
{
|
|
long old_val;
|
|
|
|
if (se == NULL || se == &audit_default_se) {
|
|
return;
|
|
}
|
|
|
|
old_val = OSAddAtomicLong(1, &se->se_procnt);
|
|
KASSERT(old_val <= PID_MAX,
|
|
("audit_inc_procount: proc count > PID_MAX"));
|
|
}
|
|
|
|
/*
|
|
* Decrement the process count and add a knote if it is the last process
|
|
* to exit the session.
|
|
*/
|
|
static void
|
|
audit_dec_procount(au_sentry_t *se)
|
|
{
|
|
long old_val;
|
|
|
|
if (se == NULL || se == &audit_default_se) {
|
|
return;
|
|
}
|
|
|
|
old_val = OSAddAtomicLong(-1, &se->se_procnt);
|
|
/*
|
|
* If this was the last process generate an audit event to notify
|
|
* userland of the session ending.
|
|
*/
|
|
if (old_val == 1) {
|
|
audit_session_event(AUE_SESSION_END, &se->se_auinfo);
|
|
}
|
|
KASSERT(old_val >= 1,
|
|
("audit_dec_procount: proc count < 0"));
|
|
}
|
|
|
|
/*
|
|
* Update the session entry and check to see if anything was updated.
|
|
* Returns:
|
|
* 0 Nothing was updated (We don't care about process preselection masks)
|
|
* 1 Something was updated.
|
|
*/
|
|
static int
|
|
audit_update_sentry(au_sentry_t *se, auditinfo_addr_t *new_aia)
|
|
{
|
|
auditinfo_addr_t *aia = &se->se_auinfo;
|
|
int update;
|
|
|
|
KASSERT(new_aia != audit_default_aia_p,
|
|
("audit_update_sentry: Trying to update the default aia."));
|
|
|
|
update = (aia->ai_auid != new_aia->ai_auid ||
|
|
bcmp(&aia->ai_termid, &new_aia->ai_termid,
|
|
sizeof(new_aia->ai_termid)) ||
|
|
aia->ai_flags != new_aia->ai_flags);
|
|
|
|
if (update) {
|
|
bcopy(new_aia, aia, sizeof(*aia));
|
|
}
|
|
|
|
return update;
|
|
}
|
|
|
|
/*
|
|
* Return the next session ID. The range of kernel generated audit session IDs
|
|
* is ASSIGNED_ASID_MIN to ASSIGNED_ASID_MAX.
|
|
*/
|
|
static uint32_t
|
|
audit_session_nextid(void)
|
|
{
|
|
static uint32_t next_asid = ASSIGNED_ASID_MIN;
|
|
|
|
AUDIT_SENTRY_RWLOCK_ASSERT();
|
|
|
|
if (next_asid > ASSIGNED_ASID_MAX) {
|
|
next_asid = ASSIGNED_ASID_MIN;
|
|
}
|
|
|
|
return next_asid++;
|
|
}
|
|
|
|
/*
|
|
* Allocated a new audit_session entry and add it to the hash table. If the
|
|
* given ASID is set to AU_ASSIGN_ASID then audit_session_new() will pick an
|
|
* audit session ID. Otherwise, it attempts use the one given. It creates a
|
|
* reference to the entry that must be unref'ed.
|
|
*/
|
|
static auditinfo_addr_t *
|
|
audit_session_new(auditinfo_addr_t *new_aia_p, auditinfo_addr_t *old_aia_p)
|
|
{
|
|
au_asid_t new_asid;
|
|
au_sentry_t *se = NULL;
|
|
au_sentry_t *found_se = NULL;
|
|
auditinfo_addr_t *aia = NULL;
|
|
|
|
KASSERT(new_aia_p != NULL, ("audit_session_new: new_aia_p == NULL"));
|
|
|
|
new_asid = new_aia_p->ai_asid;
|
|
|
|
/*
|
|
* Alloc a new session entry now so we don't wait holding the lock.
|
|
*/
|
|
se = kalloc_type(au_sentry_t, Z_WAITOK | Z_ZERO | Z_NOFAIL);
|
|
|
|
/*
|
|
* Find an unique session ID, if desired.
|
|
*/
|
|
AUDIT_SENTRY_WLOCK();
|
|
if (new_asid == AU_ASSIGN_ASID) {
|
|
do {
|
|
new_asid = (au_asid_t)audit_session_nextid();
|
|
found_se = audit_session_find(new_asid);
|
|
|
|
/*
|
|
* If the session ID is currently active then drop the
|
|
* reference and try again.
|
|
*/
|
|
if (found_se != NULL) {
|
|
audit_unref_session(found_se);
|
|
} else {
|
|
break;
|
|
}
|
|
} while (1);
|
|
} else {
|
|
/*
|
|
* Check to see if the requested ASID is already in the
|
|
* hash table. If so, update it with the new auditinfo.
|
|
*/
|
|
if ((found_se = audit_session_find(new_asid)) != NULL) {
|
|
int updated;
|
|
|
|
updated = audit_update_sentry(found_se, new_aia_p);
|
|
|
|
AUDIT_SENTRY_WUNLOCK();
|
|
kfree_type(au_sentry_t, se);
|
|
|
|
/* If a different session then add this process in. */
|
|
if (new_aia_p != old_aia_p) {
|
|
audit_inc_procount(found_se);
|
|
}
|
|
|
|
/*
|
|
* If the session information was updated then
|
|
* generate an audit event to notify userland.
|
|
*/
|
|
if (updated) {
|
|
audit_session_event(AUE_SESSION_UPDATE,
|
|
&found_se->se_auinfo);
|
|
}
|
|
|
|
return &found_se->se_auinfo;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Start the reference and proc count at 1 to account for the process
|
|
* that invoked this via setaudit_addr() (or friends).
|
|
*/
|
|
se->se_refcnt = se->se_procnt = 1;
|
|
|
|
/*
|
|
* Populate the new session entry. Note that process masks are stored
|
|
* in kauth ucred so just zero them here.
|
|
*/
|
|
se->se_port = IPC_PORT_NULL;
|
|
aia = &se->se_auinfo;
|
|
aia->ai_asid = new_asid;
|
|
aia->ai_auid = new_aia_p->ai_auid;
|
|
bzero(&new_aia_p->ai_mask, sizeof(new_aia_p->ai_mask));
|
|
bcopy(&new_aia_p->ai_termid, &aia->ai_termid, sizeof(aia->ai_termid));
|
|
aia->ai_flags = new_aia_p->ai_flags;
|
|
|
|
/*
|
|
* Add it to the hash table.
|
|
*/
|
|
LIST_INSERT_HEAD(&au_sentry_bucket[HASH_ASID(new_asid)], se, se_link);
|
|
AUDIT_SENTRY_WUNLOCK();
|
|
|
|
/*
|
|
* Generate an audit event to notify userland of the new session.
|
|
*/
|
|
audit_session_event(AUE_SESSION_START, aia);
|
|
au_history_record(se, AU_HISTORY_EVENT_BIRTH);
|
|
return aia;
|
|
}
|
|
|
|
/*
|
|
* Lookup an existing session. A copy of the audit session info for a given
|
|
* ASID is returned in ret_aia. Returns 0 on success.
|
|
*/
|
|
int
|
|
audit_session_lookup(au_asid_t asid, auditinfo_addr_t *ret_aia)
|
|
{
|
|
au_sentry_t *se = NULL;
|
|
|
|
if ((uint32_t)asid > ASSIGNED_ASID_MAX) {
|
|
return -1;
|
|
}
|
|
AUDIT_SENTRY_RLOCK();
|
|
if ((se = audit_session_find(asid)) == NULL) {
|
|
AUDIT_SENTRY_RUNLOCK();
|
|
return 1;
|
|
}
|
|
/* We have a reference on the session so it is safe to drop the lock. */
|
|
AUDIT_SENTRY_RUNLOCK();
|
|
if (ret_aia != NULL) {
|
|
bcopy(&se->se_auinfo, ret_aia, sizeof(*ret_aia));
|
|
}
|
|
audit_unref_session(se);
|
|
|
|
return 0;
|
|
}
|
|
|
|
void
|
|
audit_session_aiaref(auditinfo_addr_t *aia_p)
|
|
{
|
|
audit_ref_session(AU_SENTRY_PTR(aia_p));
|
|
}
|
|
|
|
/*
|
|
* Add a reference to the session entry.
|
|
*/
|
|
void
|
|
audit_session_ref(kauth_cred_t cred)
|
|
{
|
|
auditinfo_addr_t *aia_p;
|
|
|
|
KASSERT(IS_VALID_CRED(cred),
|
|
("audit_session_ref: Invalid kauth_cred."));
|
|
|
|
aia_p = cred->cr_audit.as_aia_p;
|
|
audit_session_aiaref(aia_p);
|
|
}
|
|
|
|
void
|
|
audit_session_aiaunref(auditinfo_addr_t *aia_p)
|
|
{
|
|
audit_unref_session(AU_SENTRY_PTR(aia_p));
|
|
}
|
|
|
|
/*
|
|
* Remove a reference to the session entry.
|
|
*/
|
|
void
|
|
audit_session_unref(kauth_cred_t cred)
|
|
{
|
|
auditinfo_addr_t *aia_p;
|
|
|
|
KASSERT(IS_VALID_CRED(cred),
|
|
("audit_session_unref: Invalid kauth_cred."));
|
|
|
|
aia_p = cred->cr_audit.as_aia_p;
|
|
audit_session_aiaunref(aia_p);
|
|
}
|
|
|
|
/*
|
|
* Increment the per audit session process count. Assumes that the caller has
|
|
* a reference on the process' cred.
|
|
*/
|
|
void
|
|
audit_session_procnew(proc_t p)
|
|
{
|
|
kauth_cred_t cred = proc_ucred_unsafe(p); /* during create */
|
|
auditinfo_addr_t *aia_p;
|
|
|
|
KASSERT(IS_VALID_CRED(cred),
|
|
("audit_session_procnew: Invalid kauth_cred."));
|
|
|
|
aia_p = cred->cr_audit.as_aia_p;
|
|
|
|
audit_inc_procount(AU_SENTRY_PTR(aia_p));
|
|
}
|
|
|
|
/*
|
|
* Decrement the per audit session process count. Assumes that the caller has
|
|
* a reference on the cred.
|
|
*/
|
|
void
|
|
audit_session_procexit(proc_t p)
|
|
{
|
|
kauth_cred_t cred = proc_ucred_unsafe(p); /* during exit */
|
|
auditinfo_addr_t *aia_p;
|
|
|
|
KASSERT(IS_VALID_CRED(cred),
|
|
("audit_session_procexit: Invalid kauth_cred."));
|
|
|
|
aia_p = cred->cr_audit.as_aia_p;
|
|
|
|
audit_dec_procount(AU_SENTRY_PTR(aia_p));
|
|
}
|
|
|
|
/*
|
|
* Init the audit session code.
|
|
*/
|
|
void
|
|
audit_session_init(void)
|
|
{
|
|
int i;
|
|
|
|
KASSERT((ASSIGNED_ASID_MAX - ASSIGNED_ASID_MIN) > PID_MAX,
|
|
("audit_session_init: ASSIGNED_ASID_MAX is not large enough."));
|
|
|
|
AUDIT_SENTRY_RWLOCK_INIT();
|
|
|
|
au_sentry_bucket = zalloc_permanent(sizeof(struct au_sentry) *
|
|
HASH_TABLE_SIZE, ZALIGN_PTR);
|
|
|
|
for (i = 0; i < HASH_TABLE_SIZE; i++) {
|
|
LIST_INIT(&au_sentry_bucket[i]);
|
|
}
|
|
|
|
(void)audit_sdev_init();
|
|
#if AU_HISTORY_LOGGING
|
|
au_history = zalloc_permanent(sizeof(struct au_history) * au_history_size,
|
|
ZALIGN_PTR);
|
|
#endif
|
|
}
|
|
|
|
static int
|
|
audit_session_update_check(kauth_cred_t cred, auditinfo_addr_t *old,
|
|
auditinfo_addr_t *new)
|
|
{
|
|
uint64_t n;
|
|
|
|
/* If the current audit ID is not the default then it is immutable. */
|
|
if (old->ai_auid != AU_DEFAUDITID && old->ai_auid != new->ai_auid) {
|
|
return EINVAL;
|
|
}
|
|
|
|
/* If the current termid is not the default then it is immutable. */
|
|
if ((old->ai_termid.at_type != AU_IPv4 ||
|
|
old->ai_termid.at_port != 0 ||
|
|
old->ai_termid.at_addr[0] != 0) &&
|
|
(old->ai_termid.at_port != new->ai_termid.at_port ||
|
|
old->ai_termid.at_type != new->ai_termid.at_type ||
|
|
0 != bcmp(&old->ai_termid.at_addr, &new->ai_termid.at_addr,
|
|
sizeof(old->ai_termid.at_addr)))) {
|
|
return EINVAL;
|
|
}
|
|
|
|
/* The flags may be set only according to the
|
|
* audit_session_*_set_sflags_masks.
|
|
*/
|
|
n = ~old->ai_flags & new->ai_flags;
|
|
if (0 != n &&
|
|
!((n == (audit_session_superuser_set_sflags_mask & n) &&
|
|
kauth_cred_issuser(cred)) ||
|
|
(n == (audit_session_member_set_sflags_mask & n) &&
|
|
old->ai_asid == new->ai_asid))) {
|
|
return EINVAL;
|
|
}
|
|
|
|
/* The flags may be cleared only according to the
|
|
* audit_session_*_clear_sflags_masks.
|
|
*/
|
|
n = ~new->ai_flags & old->ai_flags;
|
|
if (0 != n &&
|
|
!((n == (audit_session_superuser_clear_sflags_mask & n) &&
|
|
kauth_cred_issuser(cred)) ||
|
|
(n == (audit_session_member_clear_sflags_mask & n) &&
|
|
old->ai_asid == new->ai_asid))) {
|
|
return EINVAL;
|
|
}
|
|
|
|
/* The audit masks are mutable. */
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Protect updates to proc->cred->session
|
|
*
|
|
* The lifecycle of sessions and kauth creds do not compose well,
|
|
* so this lock makes sure that even in the presence of concurrent
|
|
* updates to the proc's credential, sessions stay stable.
|
|
*
|
|
* This lock is only used to serialize audit_session_setaia()
|
|
* and audit_session_join_internal() with each other,
|
|
* which are called from posix_spawn() or regular syscall context.
|
|
*
|
|
* Once the session is established in the cred, this lock
|
|
* is no longer required, it is only about avoiding racing
|
|
* updates and lifetime bugs due to the discrepancy between
|
|
* audit sessions and creds.
|
|
*/
|
|
static void
|
|
proc_audit_session_lock(proc_t p)
|
|
{
|
|
lck_mtx_lock(&p->p_audit_mlock);
|
|
}
|
|
|
|
static void
|
|
proc_audit_session_unlock(proc_t p)
|
|
{
|
|
lck_mtx_unlock(&p->p_audit_mlock);
|
|
}
|
|
|
|
/*
|
|
* Safely update kauth cred of the given process with new the given audit info.
|
|
*/
|
|
int
|
|
audit_session_setaia(proc_t p, auditinfo_addr_t *new_aia_p)
|
|
{
|
|
kauth_cred_t my_cred;
|
|
struct au_session as, *asp = &as;
|
|
auditinfo_addr_t caia, *old_aia_p;
|
|
int ret;
|
|
|
|
proc_audit_session_lock(p);
|
|
my_cred = kauth_cred_proc_ref(p);
|
|
|
|
/*
|
|
* If this is going to modify an existing session then do some
|
|
* immutable checks.
|
|
*/
|
|
if (audit_session_lookup(new_aia_p->ai_asid, &caia) == 0) {
|
|
ret = audit_session_update_check(my_cred, &caia, new_aia_p);
|
|
if (ret) {
|
|
proc_audit_session_unlock(p);
|
|
kauth_cred_unref(&my_cred);
|
|
return ret;
|
|
}
|
|
}
|
|
|
|
bcopy(&new_aia_p->ai_mask, &as.as_mask, sizeof(as.as_mask));
|
|
old_aia_p = my_cred->cr_audit.as_aia_p;
|
|
/* audit_session_new() adds a reference on the session */
|
|
as.as_aia_p = audit_session_new(new_aia_p, old_aia_p);
|
|
|
|
kauth_cred_proc_update(p, PROC_SETTOKEN_LAZY,
|
|
^bool (kauth_cred_t parent __unused, kauth_cred_t model) {
|
|
return kauth_cred_model_setauditinfo(model, asp);
|
|
});
|
|
|
|
proc_audit_session_unlock(p);
|
|
kauth_cred_unref(&my_cred);
|
|
|
|
/* If the process left a session then update the process count. */
|
|
if (old_aia_p != new_aia_p) {
|
|
audit_dec_procount(AU_SENTRY_PTR(old_aia_p));
|
|
}
|
|
|
|
/* Drop the reference taken by audit_session_new() above. */
|
|
audit_unref_session(AU_SENTRY_PTR(as.as_aia_p));
|
|
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* audit_session_self (system call)
|
|
*
|
|
* Description: Obtain a Mach send right for the current session.
|
|
*
|
|
* Parameters: p Process calling audit_session_self().
|
|
*
|
|
* Returns: *ret_port Named Mach send right, which may be
|
|
* MACH_PORT_NULL in the failure case.
|
|
*
|
|
* Errno: 0 Success
|
|
* EINVAL The calling process' session has not be set.
|
|
* ESRCH Bad process, can't get valid cred for process.
|
|
* ENOMEM Port allocation failed due to no free memory.
|
|
*/
|
|
int
|
|
audit_session_self(proc_t p, __unused struct audit_session_self_args *uap,
|
|
mach_port_name_t *ret_port)
|
|
{
|
|
ipc_port_t sendport = IPC_PORT_NULL;
|
|
kauth_cred_t cred = NULL;
|
|
auditinfo_addr_t *aia_p;
|
|
au_sentry_t *se;
|
|
int err = 0;
|
|
|
|
cred = kauth_cred_proc_ref(p);
|
|
if (!IS_VALID_CRED(cred)) {
|
|
err = ESRCH;
|
|
goto done;
|
|
}
|
|
|
|
aia_p = cred->cr_audit.as_aia_p;
|
|
if (!IS_VALID_SESSION(aia_p)) {
|
|
/* Can't join the default session. */
|
|
err = EINVAL;
|
|
goto done;
|
|
}
|
|
|
|
se = AU_SENTRY_PTR(aia_p);
|
|
|
|
/*
|
|
* Processes that join using this mach port will inherit this process'
|
|
* pre-selection masks.
|
|
*/
|
|
if (se->se_port == IPC_PORT_NULL) {
|
|
bcopy(&cred->cr_audit.as_mask, &se->se_mask,
|
|
sizeof(se->se_mask));
|
|
}
|
|
|
|
/*
|
|
* Get a send right to the session's Mach port and insert it in the
|
|
* process' mach port namespace.
|
|
*/
|
|
sendport = audit_session_mksend(aia_p, &se->se_port);
|
|
*ret_port = ipc_port_copyout_send(sendport, get_task_ipcspace(proc_task(p)));
|
|
|
|
done:
|
|
if (cred != NULL) {
|
|
kauth_cred_unref(&cred);
|
|
}
|
|
if (err != 0) {
|
|
*ret_port = MACH_PORT_NULL;
|
|
}
|
|
return err;
|
|
}
|
|
|
|
/*
|
|
* audit_session_port (system call)
|
|
*
|
|
* Description: Obtain a Mach send right for the given session ID.
|
|
*
|
|
* Parameters: p Process calling audit_session_port().
|
|
* uap->asid The target audit session ID. The special
|
|
* value -1 can be used to target the process's
|
|
* own session.
|
|
* uap->portnamep User address at which to place port name.
|
|
*
|
|
* Returns: 0 Success
|
|
* EINVAL The calling process' session has not be set.
|
|
* EINVAL The given session ID could not be found.
|
|
* EINVAL The Mach port right could not be copied out.
|
|
* ESRCH Bad process, can't get valid cred for process.
|
|
* EPERM Only the superuser can reference sessions other
|
|
* than the process's own.
|
|
* ENOMEM Port allocation failed due to no free memory.
|
|
*/
|
|
int
|
|
audit_session_port(proc_t p, struct audit_session_port_args *uap,
|
|
__unused int *retval)
|
|
{
|
|
ipc_port_t sendport = IPC_PORT_NULL;
|
|
mach_port_name_t portname = MACH_PORT_NULL;
|
|
kauth_cred_t cred = NULL;
|
|
auditinfo_addr_t *aia_p = NULL;
|
|
au_sentry_t *se = NULL;
|
|
int err = 0;
|
|
|
|
/* Note: Currently this test will never be true, because
|
|
* ASSIGNED_ASID_MAX is effectively (uint32_t)-2.
|
|
*/
|
|
if (uap->asid != -1 && (uint32_t)uap->asid > ASSIGNED_ASID_MAX) {
|
|
err = EINVAL;
|
|
goto done;
|
|
}
|
|
cred = kauth_cred_proc_ref(p);
|
|
if (!IS_VALID_CRED(cred)) {
|
|
err = ESRCH;
|
|
goto done;
|
|
}
|
|
aia_p = cred->cr_audit.as_aia_p;
|
|
|
|
/* Find the session corresponding to the requested audit
|
|
* session ID. If found, take a reference on it so that
|
|
* the session is not dropped until the join is later done.
|
|
*/
|
|
if (uap->asid == (au_asid_t)-1 ||
|
|
uap->asid == aia_p->ai_asid) {
|
|
if (!IS_VALID_SESSION(aia_p)) {
|
|
/* Can't join the default session. */
|
|
err = EINVAL;
|
|
goto done;
|
|
}
|
|
|
|
/* No privilege is required to obtain a port for our
|
|
* own session.
|
|
*/
|
|
se = AU_SENTRY_PTR(aia_p);
|
|
audit_ref_session(se);
|
|
} else {
|
|
/*
|
|
* Only privileged processes may obtain a port for
|
|
* any existing session.
|
|
*/
|
|
err = priv_check_cred(cred, PRIV_AUDIT_SESSION_PORT, 0);
|
|
if (err != 0) {
|
|
goto done;
|
|
}
|
|
AUDIT_SENTRY_RLOCK();
|
|
se = audit_session_find(uap->asid);
|
|
AUDIT_SENTRY_RUNLOCK();
|
|
if (NULL == se) {
|
|
err = EINVAL;
|
|
goto done;
|
|
}
|
|
aia_p = &se->se_auinfo;
|
|
}
|
|
|
|
/*
|
|
* Processes that join using this mach port will inherit this process'
|
|
* pre-selection masks.
|
|
*/
|
|
if (se->se_port == IPC_PORT_NULL) {
|
|
bcopy(&cred->cr_audit.as_mask, &se->se_mask,
|
|
sizeof(se->se_mask));
|
|
}
|
|
|
|
/*
|
|
* Use the session reference to create a mach port reference for the
|
|
* session (at which point we are free to drop the session reference)
|
|
* and then copy out the mach port to the process' mach port namespace.
|
|
*/
|
|
sendport = audit_session_mksend(aia_p, &se->se_port);
|
|
portname = ipc_port_copyout_send(sendport, get_task_ipcspace(proc_task(p)));
|
|
if (!MACH_PORT_VALID(portname)) {
|
|
err = EINVAL;
|
|
goto done;
|
|
}
|
|
err = copyout(&portname, uap->portnamep, sizeof(mach_port_name_t));
|
|
done:
|
|
if (cred != NULL) {
|
|
kauth_cred_unref(&cred);
|
|
}
|
|
if (NULL != se) {
|
|
audit_unref_session(se);
|
|
}
|
|
if (MACH_PORT_VALID(portname) && 0 != err) {
|
|
(void)mach_port_deallocate(get_task_ipcspace(proc_task(p)),
|
|
portname);
|
|
}
|
|
|
|
return err;
|
|
}
|
|
|
|
static int
|
|
audit_session_join_internal(proc_t p, ipc_port_t port, au_asid_t *new_asid)
|
|
{
|
|
__block auditinfo_addr_t *old_aia_p = NULL;
|
|
auditinfo_addr_t *new_aia_p;
|
|
int err = 0;
|
|
|
|
if ((new_aia_p = audit_session_porttoaia(port)) == NULL) {
|
|
err = EINVAL;
|
|
*new_asid = AU_DEFAUDITSID;
|
|
goto done;
|
|
}
|
|
|
|
/* Increment the proc count of new session */
|
|
audit_inc_procount(AU_SENTRY_PTR(new_aia_p));
|
|
|
|
proc_audit_session_lock(p);
|
|
|
|
kauth_cred_proc_update(p, PROC_SETTOKEN_LAZY,
|
|
^bool (kauth_cred_t parent __unused, kauth_cred_t model) {
|
|
struct au_session new_as;
|
|
|
|
old_aia_p = model->cr_audit.as_aia_p;
|
|
|
|
if (old_aia_p->ai_asid == new_aia_p->ai_asid) {
|
|
return false;
|
|
}
|
|
|
|
bcopy(&new_aia_p->ai_mask, &new_as.as_mask,
|
|
sizeof(new_as.as_mask));
|
|
new_as.as_aia_p = new_aia_p;
|
|
|
|
return kauth_cred_model_setauditinfo(model, &new_as);
|
|
});
|
|
|
|
proc_audit_session_unlock(p);
|
|
|
|
/* Decrement the process count of the former session. */
|
|
audit_dec_procount(AU_SENTRY_PTR(old_aia_p));
|
|
|
|
*new_asid = new_aia_p->ai_asid;
|
|
|
|
done:
|
|
if (port != IPC_PORT_NULL) {
|
|
ipc_port_release_send(port);
|
|
}
|
|
|
|
return err;
|
|
}
|
|
|
|
/*
|
|
* audit_session_spawnjoin
|
|
*
|
|
* Description: posix_spawn() interface to audit_session_join_internal().
|
|
*
|
|
* Returns: 0 Success
|
|
* EINVAL Invalid Mach port name.
|
|
* ESRCH Invalid calling process/cred.
|
|
*/
|
|
int
|
|
audit_session_spawnjoin(proc_t p, ipc_port_t port)
|
|
{
|
|
au_asid_t new_asid;
|
|
|
|
return audit_session_join_internal(p, port, &new_asid);
|
|
}
|
|
|
|
/*
|
|
* audit_session_join (system call)
|
|
*
|
|
* Description: Join the session for a given Mach port send right.
|
|
*
|
|
* Parameters: p Process calling session join.
|
|
* uap->port A Mach send right.
|
|
*
|
|
* Returns: *ret_asid Audit session ID of new session.
|
|
* In the failure case the return value will be -1
|
|
* and 'errno' will be set to a non-zero value
|
|
* described below.
|
|
*
|
|
* Errno: 0 Success
|
|
* EINVAL Invalid Mach port name.
|
|
* ESRCH Invalid calling process/cred.
|
|
*/
|
|
int
|
|
audit_session_join(proc_t p, struct audit_session_join_args *uap,
|
|
au_asid_t *ret_asid)
|
|
{
|
|
ipc_port_t port = IPC_PORT_NULL;
|
|
mach_port_name_t send = uap->port;
|
|
int err = 0;
|
|
|
|
|
|
if (ipc_object_copyin(get_task_ipcspace(proc_task(p)), send,
|
|
MACH_MSG_TYPE_COPY_SEND, &port, 0, NULL,
|
|
IPC_OBJECT_COPYIN_FLAGS_ALLOW_IMMOVABLE_SEND) != KERN_SUCCESS) {
|
|
*ret_asid = AU_DEFAUDITSID;
|
|
err = EINVAL;
|
|
} else {
|
|
err = audit_session_join_internal(p, port, ret_asid);
|
|
}
|
|
|
|
return err;
|
|
}
|
|
|
|
/*
|
|
* Audit session device.
|
|
*/
|
|
|
|
/*
|
|
* Free an audit sdev entry.
|
|
*/
|
|
static void
|
|
audit_sdev_entry_free(struct audit_sdev_entry *ase)
|
|
{
|
|
kfree_data(ase->ase_record, ase->ase_record_len);
|
|
kfree_type(struct audit_sdev_entry, ase);
|
|
}
|
|
|
|
/*
|
|
* Append individual record to a queue. Allocate queue-local buffer and
|
|
* add to the queue. If the queue is full or we can't allocate memory,
|
|
* drop the newest record.
|
|
*/
|
|
static void
|
|
audit_sdev_append(struct audit_sdev *asdev, void *record, u_int record_len)
|
|
{
|
|
struct audit_sdev_entry *ase;
|
|
|
|
AUDIT_SDEV_LOCK_ASSERT(asdev);
|
|
|
|
if (asdev->asdev_qlen >= asdev->asdev_qlimit) {
|
|
asdev->asdev_drops++;
|
|
audit_sdev_drops++;
|
|
return;
|
|
}
|
|
|
|
ase = kalloc_type(struct audit_sdev_entry, Z_NOWAIT | Z_ZERO);
|
|
if (NULL == ase) {
|
|
asdev->asdev_drops++;
|
|
audit_sdev_drops++;
|
|
return;
|
|
}
|
|
|
|
ase->ase_record = kalloc_data(record_len, Z_NOWAIT);
|
|
if (NULL == ase->ase_record) {
|
|
kfree_type(struct audit_sdev_entry, ase);
|
|
asdev->asdev_drops++;
|
|
audit_sdev_drops++;
|
|
return;
|
|
}
|
|
|
|
bcopy(record, ase->ase_record, record_len);
|
|
ase->ase_record_len = record_len;
|
|
|
|
TAILQ_INSERT_TAIL(&asdev->asdev_queue, ase, ase_queue);
|
|
asdev->asdev_inserts++;
|
|
asdev->asdev_qlen++;
|
|
asdev->asdev_qbyteslen += ase->ase_record_len;
|
|
selwakeup(&asdev->asdev_selinfo);
|
|
if (asdev->asdev_flags & AUDIT_SDEV_ASYNC) {
|
|
pgsigio(asdev->asdev_sigio, SIGIO);
|
|
}
|
|
|
|
cv_broadcast(&asdev->asdev_cv);
|
|
}
|
|
|
|
/*
|
|
* Submit an audit record to be queued in the audit session device.
|
|
*/
|
|
void
|
|
audit_sdev_submit(__unused au_id_t auid, __unused au_asid_t asid, void *record,
|
|
u_int record_len)
|
|
{
|
|
struct audit_sdev *asdev;
|
|
|
|
/*
|
|
* Lockless read to avoid lock overhead if sessio devices are not in
|
|
* use.
|
|
*/
|
|
if (NULL == TAILQ_FIRST(&audit_sdev_list)) {
|
|
return;
|
|
}
|
|
|
|
AUDIT_SDEV_LIST_RLOCK();
|
|
TAILQ_FOREACH(asdev, &audit_sdev_list, asdev_list) {
|
|
AUDIT_SDEV_LOCK(asdev);
|
|
|
|
/*
|
|
* Only append to the sdev queue if the AUID and ASID match that
|
|
* of the process that opened this session device or if the
|
|
* ALLSESSIONS flag is set.
|
|
*/
|
|
if ((/* XXXss auid == asdev->asdev_auid && */
|
|
asid == asdev->asdev_asid) ||
|
|
(asdev->asdev_flags & AUDIT_SDEV_ALLSESSIONS) != 0) {
|
|
audit_sdev_append(asdev, record, record_len);
|
|
}
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
}
|
|
AUDIT_SDEV_LIST_RUNLOCK();
|
|
|
|
/* Unlocked increment. */
|
|
audit_sdev_records++;
|
|
}
|
|
|
|
/*
|
|
* Allocate a new audit sdev. Connects the sdev, on succes, to the global
|
|
* list and updates statistics.
|
|
*/
|
|
static struct audit_sdev *
|
|
audit_sdev_alloc(void)
|
|
{
|
|
struct audit_sdev *asdev;
|
|
|
|
AUDIT_SDEV_LIST_WLOCK_ASSERT();
|
|
|
|
asdev = kalloc_type(struct audit_sdev, Z_ZERO | Z_WAITOK | Z_NOFAIL);
|
|
asdev->asdev_qlimit = AUDIT_SDEV_QLIMIT_DEFAULT;
|
|
TAILQ_INIT(&asdev->asdev_queue);
|
|
AUDIT_SDEV_LOCK_INIT(asdev);
|
|
AUDIT_SDEV_SX_LOCK_INIT(asdev);
|
|
cv_init(&asdev->asdev_cv, "audit_sdev_cv");
|
|
|
|
/*
|
|
* Add to global list and update global statistics.
|
|
*/
|
|
TAILQ_INSERT_HEAD(&audit_sdev_list, asdev, asdev_list);
|
|
audit_sdev_count++;
|
|
audit_sdev_ever++;
|
|
|
|
return asdev;
|
|
}
|
|
|
|
/*
|
|
* Flush all records currently present in an audit sdev.
|
|
*/
|
|
static void
|
|
audit_sdev_flush(struct audit_sdev *asdev)
|
|
{
|
|
struct audit_sdev_entry *ase;
|
|
|
|
AUDIT_SDEV_LOCK_ASSERT(asdev);
|
|
|
|
while ((ase = TAILQ_FIRST(&asdev->asdev_queue)) != NULL) {
|
|
TAILQ_REMOVE(&asdev->asdev_queue, ase, ase_queue);
|
|
asdev->asdev_qbyteslen -= ase->ase_record_len;
|
|
audit_sdev_entry_free(ase);
|
|
asdev->asdev_qlen--;
|
|
}
|
|
asdev->asdev_qoffset = 0;
|
|
|
|
KASSERT(0 == asdev->asdev_qlen, ("audit_sdev_flush: asdev_qlen"));
|
|
KASSERT(0 == asdev->asdev_qbyteslen,
|
|
("audit_sdev_flush: asdev_qbyteslen"));
|
|
}
|
|
|
|
/*
|
|
* Free an audit sdev.
|
|
*/
|
|
static void
|
|
audit_sdev_free(struct audit_sdev *asdev)
|
|
{
|
|
AUDIT_SDEV_LIST_WLOCK_ASSERT();
|
|
AUDIT_SDEV_LOCK_ASSERT(asdev);
|
|
|
|
/* XXXss - preselect hook here */
|
|
audit_sdev_flush(asdev);
|
|
cv_destroy(&asdev->asdev_cv);
|
|
AUDIT_SDEV_SX_LOCK_DESTROY(asdev);
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
AUDIT_SDEV_LOCK_DESTROY(asdev);
|
|
|
|
TAILQ_REMOVE(&audit_sdev_list, asdev, asdev_list);
|
|
kfree_type(struct audit_sdev, asdev);
|
|
audit_sdev_count--;
|
|
}
|
|
|
|
/*
|
|
* Get the auditinfo_addr of the proc and check to see if suser. Will return
|
|
* non-zero if not suser.
|
|
*/
|
|
static int
|
|
audit_sdev_get_aia(proc_t p, struct auditinfo_addr *aia_p)
|
|
{
|
|
int error;
|
|
kauth_cred_t scred;
|
|
|
|
scred = kauth_cred_proc_ref(p);
|
|
error = suser(scred, &p->p_acflag);
|
|
|
|
if (NULL != aia_p) {
|
|
bcopy(scred->cr_audit.as_aia_p, aia_p, sizeof(*aia_p));
|
|
}
|
|
kauth_cred_unref(&scred);
|
|
|
|
return error;
|
|
}
|
|
|
|
/*
|
|
* Audit session dev open method.
|
|
*/
|
|
static int
|
|
audit_sdev_open(dev_t dev, __unused int flags, __unused int devtype, proc_t p)
|
|
{
|
|
struct audit_sdev *asdev;
|
|
struct auditinfo_addr aia;
|
|
int u;
|
|
|
|
u = minor(dev);
|
|
if (u < 0 || u >= MAX_AUDIT_SDEVS) {
|
|
return ENXIO;
|
|
}
|
|
|
|
(void) audit_sdev_get_aia(p, &aia);
|
|
|
|
AUDIT_SDEV_LIST_WLOCK();
|
|
asdev = audit_sdev_dtab[u];
|
|
if (NULL == asdev) {
|
|
asdev = audit_sdev_alloc();
|
|
if (NULL == asdev) {
|
|
AUDIT_SDEV_LIST_WUNLOCK();
|
|
return ENOMEM;
|
|
}
|
|
audit_sdev_dtab[u] = asdev;
|
|
} else {
|
|
KASSERT(asdev->asdev_open, ("audit_sdev_open: Already open"));
|
|
AUDIT_SDEV_LIST_WUNLOCK();
|
|
return EBUSY;
|
|
}
|
|
asdev->asdev_open = 1;
|
|
asdev->asdev_auid = aia.ai_auid;
|
|
asdev->asdev_asid = aia.ai_asid;
|
|
asdev->asdev_flags = 0;
|
|
|
|
AUDIT_SDEV_LIST_WUNLOCK();
|
|
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Audit session dev close method.
|
|
*/
|
|
static int
|
|
audit_sdev_close(dev_t dev, __unused int flags, __unused int devtype,
|
|
__unused proc_t p)
|
|
{
|
|
struct audit_sdev *asdev;
|
|
int u;
|
|
|
|
u = minor(dev);
|
|
asdev = audit_sdev_dtab[u];
|
|
|
|
KASSERT(asdev != NULL, ("audit_sdev_close: asdev == NULL"));
|
|
KASSERT(asdev->asdev_open, ("audit_sdev_close: !asdev_open"));
|
|
|
|
AUDIT_SDEV_LIST_WLOCK();
|
|
AUDIT_SDEV_LOCK(asdev);
|
|
asdev->asdev_open = 0;
|
|
audit_sdev_free(asdev); /* sdev lock unlocked in audit_sdev_free() */
|
|
audit_sdev_dtab[u] = NULL;
|
|
AUDIT_SDEV_LIST_WUNLOCK();
|
|
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Audit session dev ioctl method.
|
|
*/
|
|
static int
|
|
audit_sdev_ioctl(dev_t dev, u_long cmd, caddr_t data,
|
|
__unused int flag, proc_t p)
|
|
{
|
|
struct audit_sdev *asdev;
|
|
int error;
|
|
|
|
asdev = audit_sdev_dtab[minor(dev)];
|
|
KASSERT(asdev != NULL, ("audit_sdev_ioctl: asdev == NULL"));
|
|
|
|
error = 0;
|
|
|
|
switch (cmd) {
|
|
case FIONBIO:
|
|
AUDIT_SDEV_LOCK(asdev);
|
|
if (*(int *)data) {
|
|
asdev->asdev_flags |= AUDIT_SDEV_NBIO;
|
|
} else {
|
|
asdev->asdev_flags &= ~AUDIT_SDEV_NBIO;
|
|
}
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
break;
|
|
|
|
case FIONREAD:
|
|
AUDIT_SDEV_LOCK(asdev);
|
|
*(int *)data = asdev->asdev_qbyteslen - asdev->asdev_qoffset;
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
break;
|
|
|
|
case AUDITSDEV_GET_QLEN:
|
|
*(u_int *)data = asdev->asdev_qlen;
|
|
break;
|
|
|
|
case AUDITSDEV_GET_QLIMIT:
|
|
*(u_int *)data = asdev->asdev_qlimit;
|
|
break;
|
|
|
|
case AUDITSDEV_SET_QLIMIT:
|
|
if (*(u_int *)data >= AUDIT_SDEV_QLIMIT_MIN ||
|
|
*(u_int *)data <= AUDIT_SDEV_QLIMIT_MAX) {
|
|
asdev->asdev_qlimit = *(u_int *)data;
|
|
} else {
|
|
error = EINVAL;
|
|
}
|
|
break;
|
|
|
|
case AUDITSDEV_GET_QLIMIT_MIN:
|
|
*(u_int *)data = AUDIT_SDEV_QLIMIT_MIN;
|
|
break;
|
|
|
|
case AUDITSDEV_GET_QLIMIT_MAX:
|
|
*(u_int *)data = AUDIT_SDEV_QLIMIT_MAX;
|
|
break;
|
|
|
|
case AUDITSDEV_FLUSH:
|
|
if (AUDIT_SDEV_SX_XLOCK_SIG(asdev) != 0) {
|
|
return EINTR;
|
|
}
|
|
AUDIT_SDEV_LOCK(asdev);
|
|
audit_sdev_flush(asdev);
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
AUDIT_SDEV_SX_XUNLOCK(asdev);
|
|
break;
|
|
|
|
case AUDITSDEV_GET_MAXDATA:
|
|
*(u_int *)data = MAXAUDITDATA;
|
|
break;
|
|
|
|
/* XXXss these should be 64 bit, maybe. */
|
|
case AUDITSDEV_GET_INSERTS:
|
|
*(u_int *)data = asdev->asdev_inserts;
|
|
break;
|
|
|
|
case AUDITSDEV_GET_READS:
|
|
*(u_int *)data = asdev->asdev_reads;
|
|
break;
|
|
|
|
case AUDITSDEV_GET_DROPS:
|
|
*(u_int *)data = asdev->asdev_drops;
|
|
break;
|
|
|
|
case AUDITSDEV_GET_ALLSESSIONS:
|
|
error = audit_sdev_get_aia(p, NULL);
|
|
if (error) {
|
|
break;
|
|
}
|
|
*(u_int *)data = (asdev->asdev_flags & AUDIT_SDEV_ALLSESSIONS) ?
|
|
1 : 0;
|
|
break;
|
|
|
|
case AUDITSDEV_SET_ALLSESSIONS:
|
|
error = audit_sdev_get_aia(p, NULL);
|
|
if (error) {
|
|
break;
|
|
}
|
|
|
|
AUDIT_SDEV_LOCK(asdev);
|
|
if (*(int *)data) {
|
|
asdev->asdev_flags |= AUDIT_SDEV_ALLSESSIONS;
|
|
} else {
|
|
asdev->asdev_flags &= ~AUDIT_SDEV_ALLSESSIONS;
|
|
}
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
break;
|
|
|
|
default:
|
|
error = ENOTTY;
|
|
}
|
|
|
|
return error;
|
|
}
|
|
|
|
/*
|
|
* Audit session dev read method.
|
|
*/
|
|
static int
|
|
audit_sdev_read(dev_t dev, struct uio *uio, __unused int flag)
|
|
{
|
|
struct audit_sdev_entry *ase;
|
|
struct audit_sdev *asdev;
|
|
u_int toread;
|
|
int error;
|
|
|
|
asdev = audit_sdev_dtab[minor(dev)];
|
|
KASSERT(NULL != asdev, ("audit_sdev_read: asdev == NULL"));
|
|
|
|
/*
|
|
* We hold a sleep lock over read and flush because we rely on the
|
|
* stability of a record in the queue during uiomove.
|
|
*/
|
|
if (0 != AUDIT_SDEV_SX_XLOCK_SIG(asdev)) {
|
|
return EINTR;
|
|
}
|
|
AUDIT_SDEV_LOCK(asdev);
|
|
while (TAILQ_EMPTY(&asdev->asdev_queue)) {
|
|
if (asdev->asdev_flags & AUDIT_SDEV_NBIO) {
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
AUDIT_SDEV_SX_XUNLOCK(asdev);
|
|
return EAGAIN;
|
|
}
|
|
error = cv_wait_sig(&asdev->asdev_cv, AUDIT_SDEV_MTX(asdev));
|
|
if (error) {
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
AUDIT_SDEV_SX_XUNLOCK(asdev);
|
|
return error;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Copy as many remaining bytes from the current record to userspace
|
|
* as we can. Keep processing records until we run out of records in
|
|
* the queue or until the user buffer runs out of space.
|
|
*
|
|
* We rely on the sleep lock to maintain ase's stability here.
|
|
*/
|
|
asdev->asdev_reads++;
|
|
while ((ase = TAILQ_FIRST(&asdev->asdev_queue)) != NULL &&
|
|
uio_resid(uio) > 0) {
|
|
AUDIT_SDEV_LOCK_ASSERT(asdev);
|
|
|
|
KASSERT(ase->ase_record_len > asdev->asdev_qoffset,
|
|
("audit_sdev_read: record_len > qoffset (1)"));
|
|
toread = MIN((int)(ase->ase_record_len - asdev->asdev_qoffset),
|
|
uio_resid(uio));
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
error = uiomove((char *) ase->ase_record + asdev->asdev_qoffset,
|
|
toread, uio);
|
|
if (error) {
|
|
AUDIT_SDEV_SX_XUNLOCK(asdev);
|
|
return error;
|
|
}
|
|
|
|
/*
|
|
* If the copy succeeded then update book-keeping, and if no
|
|
* bytes remain in the current record then free it.
|
|
*/
|
|
AUDIT_SDEV_LOCK(asdev);
|
|
KASSERT(TAILQ_FIRST(&asdev->asdev_queue) == ase,
|
|
("audit_sdev_read: queue out of sync after uiomove"));
|
|
asdev->asdev_qoffset += toread;
|
|
KASSERT(ase->ase_record_len >= asdev->asdev_qoffset,
|
|
("audit_sdev_read: record_len >= qoffset (2)"));
|
|
if (asdev->asdev_qoffset == ase->ase_record_len) {
|
|
TAILQ_REMOVE(&asdev->asdev_queue, ase, ase_queue);
|
|
asdev->asdev_qbyteslen -= ase->ase_record_len;
|
|
audit_sdev_entry_free(ase);
|
|
asdev->asdev_qlen--;
|
|
asdev->asdev_qoffset = 0;
|
|
}
|
|
}
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
AUDIT_SDEV_SX_XUNLOCK(asdev);
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Audit session device poll method.
|
|
*/
|
|
static int
|
|
audit_sdev_poll(dev_t dev, int events, void *wql, struct proc *p)
|
|
{
|
|
struct audit_sdev *asdev;
|
|
int revents;
|
|
|
|
revents = 0;
|
|
asdev = audit_sdev_dtab[minor(dev)];
|
|
KASSERT(NULL != asdev, ("audit_sdev_poll: asdev == NULL"));
|
|
|
|
if (events & (POLLIN | POLLRDNORM)) {
|
|
AUDIT_SDEV_LOCK(asdev);
|
|
if (NULL != TAILQ_FIRST(&asdev->asdev_queue)) {
|
|
revents |= events & (POLLIN | POLLRDNORM);
|
|
} else {
|
|
selrecord(p, &asdev->asdev_selinfo, wql);
|
|
}
|
|
AUDIT_SDEV_UNLOCK(asdev);
|
|
}
|
|
return revents;
|
|
}
|
|
|
|
/*
|
|
* Audit sdev clone routine. Provides a new minor number or returns -1.
|
|
* This called with DEVFS_LOCK held.
|
|
*/
|
|
static int
|
|
audit_sdev_clone(__unused dev_t dev, int action)
|
|
{
|
|
int i;
|
|
|
|
if (DEVFS_CLONE_ALLOC == action) {
|
|
for (i = 0; i < MAX_AUDIT_SDEVS; i++) {
|
|
if (NULL == audit_sdev_dtab[i]) {
|
|
return i;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* This really should return -1 here but that seems to
|
|
* hang things in devfs. We instead return 0 and let
|
|
* audit_sdev_open tell userland the bad news.
|
|
*/
|
|
return 0;
|
|
}
|
|
|
|
return -1;
|
|
}
|
|
|
|
static int
|
|
audit_sdev_init(void)
|
|
{
|
|
dev_t dev;
|
|
|
|
TAILQ_INIT(&audit_sdev_list);
|
|
AUDIT_SDEV_LIST_LOCK_INIT();
|
|
|
|
audit_sdev_major = cdevsw_add(-1, &audit_sdev_cdevsw);
|
|
if (audit_sdev_major < 0) {
|
|
return KERN_FAILURE;
|
|
}
|
|
|
|
dev = makedev(audit_sdev_major, 0);
|
|
devnode = devfs_make_node_clone(dev, DEVFS_CHAR, UID_ROOT, GID_WHEEL,
|
|
0644, audit_sdev_clone, AUDIT_SDEV_NAME);
|
|
|
|
if (NULL == devnode) {
|
|
return KERN_FAILURE;
|
|
}
|
|
|
|
return KERN_SUCCESS;
|
|
}
|
|
|
|
/* XXXss
|
|
* static int
|
|
* audit_sdev_shutdown(void)
|
|
* {
|
|
*
|
|
* devfs_remove(devnode);
|
|
* (void) cdevsw_remove(audit_sdev_major, &audit_sdev_cdevsw);
|
|
*
|
|
* return (KERN_SUCCESS);
|
|
* }
|
|
*/
|
|
|
|
#else
|
|
|
|
int
|
|
audit_session_self(proc_t p, struct audit_session_self_args *uap,
|
|
mach_port_name_t *ret_port)
|
|
{
|
|
#pragma unused(p, uap, ret_port)
|
|
|
|
return ENOSYS;
|
|
}
|
|
|
|
int
|
|
audit_session_join(proc_t p, struct audit_session_join_args *uap,
|
|
au_asid_t *ret_asid)
|
|
{
|
|
#pragma unused(p, uap, ret_asid)
|
|
|
|
return ENOSYS;
|
|
}
|
|
|
|
int
|
|
audit_session_port(proc_t p, struct audit_session_port_args *uap, int *retval)
|
|
{
|
|
#pragma unused(p, uap, retval)
|
|
|
|
return ENOSYS;
|
|
}
|
|
|
|
#endif /* CONFIG_AUDIT */
|