/*
* Function: unix_syscall
*
* Inputs: regs - pointer to i386 save area
*
* Outputs: none
*/
void
unix_syscall(x86_saved_state_t *state)
{
thread_t thread;
void *vt;
unsigned int code;
struct sysent *callp;
int error;
vm_offset_t params;
struct proc *p;
struct uthread *uthread;
x86_saved_state32_t *regs;
boolean_t is_vfork;
assert(is_saved_state32(state));
regs = saved_state32(state);
#if DEBUG
if (regs->eax == 0x800)
thread_exception_return();
#endif
thread = current_thread();
uthread = get_bsdthread_info(thread);
/* Get the approriate proc; may be different from task's for vfork() */
is_vfork = uthread->uu_flag & UT_VFORK;
if (__improbable(is_vfork != 0))
p = current_proc();
else
p = (struct proc *)get_bsdtask_info(current_task());
/* Verify that we are not being called from a task without a proc */
if (__improbable(p == NULL)) {
regs->eax = EPERM;
regs->efl |= EFL_CF;
task_terminate_internal(current_task());
thread_exception_return();
/* NOTREACHED */
}
code = regs->eax & I386_SYSCALL_NUMBER_MASK;
DEBUG_KPRINT_SYSCALL_UNIX("unix_syscall: code=%d(%s) eip=%u\n",
code, syscallnames[code >= NUM_SYSENT ? 63 : code], (uint32_t)regs->eip);
params = (vm_offset_t) (regs->uesp + sizeof (int));
regs->efl &= ~(EFL_CF);
callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
if (__improbable(callp == sysent)) {
code = fuword(params);
params += sizeof(int);
callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
}
vt = (void *)uthread->uu_arg;
if (callp->sy_arg_bytes != 0) {
#if CONFIG_REQUIRES_U32_MUNGING
sy_munge_t *mungerp;
#else
#error U32 syscalls on x86_64 kernel requires munging
#endif
uint32_t nargs;
assert((unsigned) callp->sy_arg_bytes <= sizeof (uthread->uu_arg));
nargs = callp->sy_arg_bytes;
error = copyin((user_addr_t) params, (char *) vt, nargs);
if (error) {
regs->eax = error;
regs->efl |= EFL_CF;
thread_exception_return();
/* NOTREACHED */
}
if (__probable(code != 180)) {
int *ip = (int *)vt;
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
*ip, *(ip+1), *(ip+2), *(ip+3), 0);
}
#if CONFIG_REQUIRES_U32_MUNGING
mungerp = callp->sy_arg_munge32;
if (mungerp != NULL)
(*mungerp)(vt);
#endif
} else
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
0, 0, 0, 0, 0);
/*
* Delayed binding of thread credential to process credential, if we
* are not running with an explicitly set thread credential.
*/
kauth_cred_uthread_update(uthread, p);
uthread->uu_rval[0] = 0;
uthread->uu_rval[1] = 0;
uthread->uu_flag |= UT_NOTCANCELPT;
uthread->syscall_code = code;
#ifdef JOE_DEBUG
uthread->uu_iocount = 0;
uthread->uu_vpindex = 0;
#endif
AUDIT_SYSCALL_ENTER(code, p, uthread);
error = (*(callp->sy_call))((void *) p, (void *) vt, &(uthread->uu_rval[0]));
AUDIT_SYSCALL_EXIT(code, p, uthread, error);
#ifdef JOE_DEBUG
if (uthread->uu_iocount)
printf("system call returned with uu_iocount != 0\n");
#endif
#if CONFIG_DTRACE
uthread->t_dtrace_errno = error;
#endif /* CONFIG_DTRACE */
if (__improbable(error == ERESTART)) {
/*
* Move the user's pc back to repeat the syscall:
* 5 bytes for a sysenter, or 2 for an int 8x.
* The SYSENTER_TF_CS covers single-stepping over a sysenter
* - see debug trap handler in idt.s/idt64.s
*/
pal_syscall_restart(thread, state);
}
else if (error != EJUSTRETURN) {
if (__improbable(error)) {
regs->eax = error;
regs->efl |= EFL_CF; /* carry bit */
} else { /* (not error) */
/*
* We split retval across two registers, in case the
* syscall had a 64-bit return value, in which case
* eax/edx matches the function call ABI.
*/
regs->eax = uthread->uu_rval[0];
regs->edx = uthread->uu_rval[1];
}
}
DEBUG_KPRINT_SYSCALL_UNIX(
"unix_syscall: error=%d retval=(%u,%u)\n",
error, regs->eax, regs->edx);
uthread->uu_flag &= ~UT_NOTCANCELPT;
if (__improbable(uthread->uu_lowpri_window)) {
/*
* task is marked as a low priority I/O type
* and the I/O we issued while in this system call
* collided with normal I/O operations... we'll
* delay in order to mitigate the impact of this
* task on the normal operation of the system
*/
throttle_lowpri_io(1);
}
if (__probable(code != 180))
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
error, uthread->uu_rval[0], uthread->uu_rval[1], p->p_pid, 0);
if (__improbable(!is_vfork && callp->sy_call == (sy_call_t *)execve && !error)) {
pal_execve_return(thread);
}
thread_exception_return();
/* NOTREACHED */
}
static inline int
syscallenter(struct thread *td, struct syscall_args *sa)
{
struct proc *p;
int error, traced;
PCPU_INC(cnt.v_syscall);
p = td->td_proc;
td->td_pticks = 0;
if (td->td_ucred != p->p_ucred)
cred_update_thread(td);
if (p->p_flag & P_TRACED) {
traced = 1;
PROC_LOCK(p);
td->td_dbgflags &= ~TDB_USERWR;
td->td_dbgflags |= TDB_SCE;
PROC_UNLOCK(p);
} else
traced = 0;
error = (p->p_sysent->sv_fetch_syscall_args)(td, sa);
#ifdef KTRACE
if (KTRPOINT(td, KTR_SYSCALL))
ktrsyscall(sa->code, sa->narg, sa->args);
#endif
CTR6(KTR_SYSC,
"syscall: td=%p pid %d %s (%#lx, %#lx, %#lx)",
td, td->td_proc->p_pid, syscallname(p, sa->code),
sa->args[0], sa->args[1], sa->args[2]);
if (error == 0) {
STOPEVENT(p, S_SCE, sa->narg);
if (p->p_flag & P_TRACED && p->p_stops & S_PT_SCE) {
PROC_LOCK(p);
ptracestop((td), SIGTRAP);
PROC_UNLOCK(p);
}
if (td->td_dbgflags & TDB_USERWR) {
/*
* Reread syscall number and arguments if
* debugger modified registers or memory.
*/
error = (p->p_sysent->sv_fetch_syscall_args)(td, sa);
#ifdef KTRACE
if (KTRPOINT(td, KTR_SYSCALL))
ktrsyscall(sa->code, sa->narg, sa->args);
#endif
if (error != 0)
goto retval;
}
#ifdef CAPABILITY_MODE
/*
* In capability mode, we only allow access to system calls
* flagged with SYF_CAPENABLED.
*/
if (IN_CAPABILITY_MODE(td) &&
!(sa->callp->sy_flags & SYF_CAPENABLED)) {
error = ECAPMODE;
goto retval;
}
#endif
error = syscall_thread_enter(td, sa->callp);
if (error != 0)
goto retval;
#ifdef KDTRACE_HOOKS
/*
* If the systrace module has registered it's probe
* callback and if there is a probe active for the
* syscall 'entry', process the probe.
*/
if (systrace_probe_func != NULL && sa->callp->sy_entry != 0)
(*systrace_probe_func)(sa->callp->sy_entry, sa->code,
sa->callp, sa->args, 0);
#endif
AUDIT_SYSCALL_ENTER(sa->code, td);
error = (sa->callp->sy_call)(td, sa->args);
AUDIT_SYSCALL_EXIT(error, td);
/* Save the latest error return value. */
td->td_errno = error;
#ifdef KDTRACE_HOOKS
/*
* If the systrace module has registered it's probe
* callback and if there is a probe active for the
* syscall 'return', process the probe.
*/
if (systrace_probe_func != NULL && sa->callp->sy_return != 0)
(*systrace_probe_func)(sa->callp->sy_return, sa->code,
sa->callp, NULL, (error) ? -1 : td->td_retval[0]);
#endif
syscall_thread_exit(td, sa->callp);
CTR4(KTR_SYSC, "syscall: p=%p error=%d return %#lx %#lx",
p, error, td->td_retval[0], td->td_retval[1]);
}
retval:
if (traced) {
PROC_LOCK(p);
td->td_dbgflags &= ~TDB_SCE;
PROC_UNLOCK(p);
}
(p->p_sysent->sv_set_syscall_retval)(td, error);
return (error);
}
void
unix_syscall64(x86_saved_state_t *state)
{
thread_t thread;
void *vt;
unsigned int code;
struct sysent *callp;
int args_in_regs;
boolean_t args_start_at_rdi;
int error;
struct proc *p;
struct uthread *uthread;
x86_saved_state64_t *regs;
assert(is_saved_state64(state));
regs = saved_state64(state);
#if DEBUG
if (regs->rax == 0x2000800)
thread_exception_return();
#endif
thread = current_thread();
uthread = get_bsdthread_info(thread);
/* Get the approriate proc; may be different from task's for vfork() */
if (__probable(!(uthread->uu_flag & UT_VFORK)))
p = (struct proc *)get_bsdtask_info(current_task());
else
p = current_proc();
/* Verify that we are not being called from a task without a proc */
if (__improbable(p == NULL)) {
regs->rax = EPERM;
regs->isf.rflags |= EFL_CF;
task_terminate_internal(current_task());
thread_exception_return();
/* NOTREACHED */
}
code = regs->rax & SYSCALL_NUMBER_MASK;
DEBUG_KPRINT_SYSCALL_UNIX(
"unix_syscall64: code=%d(%s) rip=%llx\n",
code, syscallnames[code >= NUM_SYSENT ? 63 : code], regs->isf.rip);
callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
vt = (void *)uthread->uu_arg;
if (__improbable(callp == sysent)) {
/*
* indirect system call... system call number
* passed as 'arg0'
*/
code = regs->rdi;
callp = (code >= NUM_SYSENT) ? &sysent[63] : &sysent[code];
args_start_at_rdi = FALSE;
args_in_regs = 5;
} else {
args_start_at_rdi = TRUE;
args_in_regs = 6;
}
if (callp->sy_narg != 0) {
assert(callp->sy_narg <= 8); /* size of uu_arg */
args_in_regs = MIN(args_in_regs, callp->sy_narg);
memcpy(vt, args_start_at_rdi ? ®s->rdi : ®s->rsi, args_in_regs * sizeof(syscall_arg_t));
if (code != 180) {
uint64_t *ip = (uint64_t *)vt;
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
(int)(*ip), (int)(*(ip+1)), (int)(*(ip+2)), (int)(*(ip+3)), 0);
}
if (__improbable(callp->sy_narg > args_in_regs)) {
int copyin_count;
copyin_count = (callp->sy_narg - args_in_regs) * sizeof(syscall_arg_t);
error = copyin((user_addr_t)(regs->isf.rsp + sizeof(user_addr_t)), (char *)&uthread->uu_arg[args_in_regs], copyin_count);
if (error) {
regs->rax = error;
regs->isf.rflags |= EFL_CF;
thread_exception_return();
/* NOTREACHED */
}
}
} else
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_START,
0, 0, 0, 0, 0);
/*
* Delayed binding of thread credential to process credential, if we
* are not running with an explicitly set thread credential.
*/
kauth_cred_uthread_update(uthread, p);
uthread->uu_rval[0] = 0;
uthread->uu_rval[1] = 0;
uthread->uu_flag |= UT_NOTCANCELPT;
uthread->syscall_code = code;
#ifdef JOE_DEBUG
uthread->uu_iocount = 0;
uthread->uu_vpindex = 0;
#endif
AUDIT_SYSCALL_ENTER(code, p, uthread);
error = (*(callp->sy_call))((void *) p, vt, &(uthread->uu_rval[0]));
AUDIT_SYSCALL_EXIT(code, p, uthread, error);
#ifdef JOE_DEBUG
if (uthread->uu_iocount)
printf("system call returned with uu_iocount != 0\n");
#endif
#if CONFIG_DTRACE
uthread->t_dtrace_errno = error;
#endif /* CONFIG_DTRACE */
if (__improbable(error == ERESTART)) {
/*
* all system calls come through via the syscall instruction
* in 64 bit mode... its 2 bytes in length
* move the user's pc back to repeat the syscall:
*/
pal_syscall_restart( thread, state );
}
else if (error != EJUSTRETURN) {
if (__improbable(error)) {
regs->rax = error;
regs->isf.rflags |= EFL_CF; /* carry bit */
} else { /* (not error) */
switch (callp->sy_return_type) {
case _SYSCALL_RET_INT_T:
regs->rax = uthread->uu_rval[0];
regs->rdx = uthread->uu_rval[1];
break;
case _SYSCALL_RET_UINT_T:
regs->rax = ((u_int)uthread->uu_rval[0]);
regs->rdx = ((u_int)uthread->uu_rval[1]);
break;
case _SYSCALL_RET_OFF_T:
case _SYSCALL_RET_ADDR_T:
case _SYSCALL_RET_SIZE_T:
case _SYSCALL_RET_SSIZE_T:
case _SYSCALL_RET_UINT64_T:
regs->rax = *((uint64_t *)(&uthread->uu_rval[0]));
regs->rdx = 0;
break;
case _SYSCALL_RET_NONE:
break;
default:
panic("unix_syscall: unknown return type");
break;
}
regs->isf.rflags &= ~EFL_CF;
}
}
DEBUG_KPRINT_SYSCALL_UNIX(
"unix_syscall64: error=%d retval=(%llu,%llu)\n",
error, regs->rax, regs->rdx);
uthread->uu_flag &= ~UT_NOTCANCELPT;
if (__improbable(uthread->uu_lowpri_window)) {
/*
* task is marked as a low priority I/O type
* and the I/O we issued while in this system call
* collided with normal I/O operations... we'll
* delay in order to mitigate the impact of this
* task on the normal operation of the system
*/
throttle_lowpri_io(1);
}
if (__probable(code != 180))
KERNEL_DEBUG_CONSTANT_IST(KDEBUG_TRACE,
BSDDBG_CODE(DBG_BSD_EXCP_SC, code) | DBG_FUNC_END,
error, uthread->uu_rval[0], uthread->uu_rval[1], p->p_pid, 0);
thread_exception_return();
/* NOTREACHED */
}
void
ia32_syscall(struct trapframe *frame)
{
caddr_t params;
int i;
struct sysent *callp;
struct thread *td = curthread;
struct proc *p = td->td_proc;
register_t orig_tf_rflags;
int error;
int narg;
u_int32_t args[8];
u_int64_t args64[8];
u_int code;
ksiginfo_t ksi;
PCPU_INC(cnt.v_syscall);
td->td_pticks = 0;
td->td_frame = frame;
if (td->td_ucred != p->p_ucred)
cred_update_thread(td);
params = (caddr_t)frame->tf_rsp + sizeof(u_int32_t);
code = frame->tf_rax;
orig_tf_rflags = frame->tf_rflags;
if (p->p_sysent->sv_prepsyscall) {
/*
* The prep code is MP aware.
*/
(*p->p_sysent->sv_prepsyscall)(frame, args, &code, ¶ms);
} else {
/*
* Need to check if this is a 32 bit or 64 bit syscall.
* fuword is MP aware.
*/
if (code == SYS_syscall) {
/*
* Code is first argument, followed by actual args.
*/
code = fuword32(params);
params += sizeof(int);
} else if (code == SYS___syscall) {
/*
* Like syscall, but code is a quad, so as to maintain
* quad alignment for the rest of the arguments.
* We use a 32-bit fetch in case params is not
* aligned.
*/
code = fuword32(params);
params += sizeof(quad_t);
}
}
if (p->p_sysent->sv_mask)
code &= p->p_sysent->sv_mask;
if (code >= p->p_sysent->sv_size)
callp = &p->p_sysent->sv_table[0];
else
callp = &p->p_sysent->sv_table[code];
narg = callp->sy_narg;
/*
* copyin and the ktrsyscall()/ktrsysret() code is MP-aware
*/
if (params != NULL && narg != 0)
error = copyin(params, (caddr_t)args,
(u_int)(narg * sizeof(int)));
else
error = 0;
for (i = 0; i < narg; i++)
args64[i] = args[i];
#ifdef KTRACE
if (KTRPOINT(td, KTR_SYSCALL))
ktrsyscall(code, narg, args64);
#endif
CTR4(KTR_SYSC, "syscall enter thread %p pid %d proc %s code %d", td,
td->td_proc->p_pid, td->td_proc->p_comm, code);
if (error == 0) {
td->td_retval[0] = 0;
td->td_retval[1] = frame->tf_rdx;
STOPEVENT(p, S_SCE, narg);
PTRACESTOP_SC(p, td, S_PT_SCE);
AUDIT_SYSCALL_ENTER(code, td);
error = (*callp->sy_call)(td, args64);
AUDIT_SYSCALL_EXIT(error, td);
}
switch (error) {
case 0:
frame->tf_rax = td->td_retval[0];
frame->tf_rdx = td->td_retval[1];
frame->tf_rflags &= ~PSL_C;
break;
case ERESTART:
/*
* Reconstruct pc, assuming lcall $X,y is 7 bytes,
* int 0x80 is 2 bytes. We saved this in tf_err.
*/
frame->tf_rip -= frame->tf_err;
break;
case EJUSTRETURN:
break;
default:
if (p->p_sysent->sv_errsize) {
if (error >= p->p_sysent->sv_errsize)
error = -1; /* XXX */
else
error = p->p_sysent->sv_errtbl[error];
}
frame->tf_rax = error;
frame->tf_rflags |= PSL_C;
break;
}
/*
* Traced syscall.
*/
if (orig_tf_rflags & PSL_T) {
frame->tf_rflags &= ~PSL_T;
ksiginfo_init_trap(&ksi);
ksi.ksi_signo = SIGTRAP;
ksi.ksi_code = TRAP_TRACE;
ksi.ksi_addr = (void *)frame->tf_rip;
trapsignal(td, &ksi);
}
/*
* Check for misbehavior.
*/
WITNESS_WARN(WARN_PANIC, NULL, "System call %s returning",
(code >= 0 && code < SYS_MAXSYSCALL) ? freebsd32_syscallnames[code] : "???");
KASSERT(td->td_critnest == 0,
("System call %s returning in a critical section",
(code >= 0 && code < SYS_MAXSYSCALL) ? freebsd32_syscallnames[code] : "???"));
KASSERT(td->td_locks == 0,
("System call %s returning with %d locks held",
(code >= 0 && code < SYS_MAXSYSCALL) ? freebsd32_syscallnames[code] : "???",
td->td_locks));
/*
* Handle reschedule and other end-of-syscall issues
*/
userret(td, frame);
CTR4(KTR_SYSC, "syscall exit thread %p pid %d proc %s code %d", td,
td->td_proc->p_pid, td->td_proc->p_comm, code);
#ifdef KTRACE
if (KTRPOINT(td, KTR_SYSRET))
ktrsysret(code, error, td->td_retval[0]);
#endif
/*
* This works because errno is findable through the
* register set. If we ever support an emulation where this
* is not the case, this code will need to be revisited.
*/
STOPEVENT(p, S_SCX, code);
PTRACESTOP_SC(p, td, S_PT_SCX);
}