static void
ttm_lock_send_sig(int signo)
{
struct proc *p;
p = curproc; /* XXXKIB curthread ? */
PROC_LOCK(p);
kern_psignal(p, signo);
PROC_UNLOCK(p);
}
/*
* Called by events that want to shut down.. e.g <CTL><ALT><DEL> on a PC
*/
void
shutdown_nice(int howto)
{
if (initproc != NULL) {
/* Send a signal to init(8) and have it shutdown the world. */
PROC_LOCK(initproc);
if (howto & RB_POWEROFF)
kern_psignal(initproc, SIGUSR2);
else if (howto & RB_HALT)
kern_psignal(initproc, SIGUSR1);
else
kern_psignal(initproc, SIGINT);
PROC_UNLOCK(initproc);
} else {
/* No init(8) running, so simply reboot. */
kern_reboot(howto | RB_NOSYNC);
}
}
/*
* This function is called by the hv_kvp_deinit -
* destroy character device
*/
static void
hv_kvp_dev_destroy(void)
{
if (daemon_task != NULL) {
PROC_LOCK(daemon_task);
kern_psignal(daemon_task, SIGKILL);
PROC_UNLOCK(daemon_task);
}
destroy_dev(hv_kvp_dev);
free(hv_kvp_dev_buf, M_HV_KVP_DEV_BUF);
return;
}
int
procfs_doprocctl(PFS_FILL_ARGS)
{
int error;
struct namemap *nm;
if (uio == NULL || uio->uio_rw != UIO_WRITE)
return (EOPNOTSUPP);
/*
* Map signal names into signal generation
* or debug control. Unknown commands and/or signals
* return EOPNOTSUPP.
*
* Sending a signal while the process is being debugged
* also has the side effect of letting the target continue
* to run. There is no way to single-step a signal delivery.
*/
error = EOPNOTSUPP;
sbuf_trim(sb);
sbuf_finish(sb);
nm = findname(ctlnames, sbuf_data(sb), sbuf_len(sb));
if (nm) {
printf("procfs: got a %s command\n", sbuf_data(sb));
error = procfs_control(td, p, nm->nm_val);
} else {
nm = findname(signames, sbuf_data(sb), sbuf_len(sb));
if (nm) {
printf("procfs: got a sig%s\n", sbuf_data(sb));
PROC_LOCK(p);
if (TRACE_WAIT_P(td->td_proc, p)) {
p->p_xstat = nm->nm_val;
#ifdef FIX_SSTEP
FIX_SSTEP(FIRST_THREAD_IN_PROC(p));
#endif
p->p_flag &= ~P_STOPPED_SIG;
PROC_SLOCK(p);
thread_unsuspend(p);
PROC_SUNLOCK(p);
} else
kern_psignal(p, nm->nm_val);
PROC_UNLOCK(p);
error = 0;
}
}
return (error);
}
static void
pmclog_stop_kthread(struct pmc_owner *po)
{
mtx_lock(&pmc_kthread_mtx);
po->po_flags &= ~PMC_PO_OWNS_LOGFILE;
if (po->po_kthread != NULL) {
PROC_LOCK(po->po_kthread);
kern_psignal(po->po_kthread, SIGHUP);
PROC_UNLOCK(po->po_kthread);
}
wakeup_one(po);
while (po->po_kthread)
msleep(po->po_kthread, &pmc_kthread_mtx, PPAUSE, "pmckstp", 0);
mtx_unlock(&pmc_kthread_mtx);
}
void
shutdown_nice(int howto)
{
shutdown_howto = howto;
/* Send a signal to init(8) and have it shutdown the world */
if (initproc != NULL) {
PROC_LOCK(initproc);
kern_psignal(initproc, SIGINT);
PROC_UNLOCK(initproc);
} else {
/* No init(8) running, so simply reboot */
kern_reboot(RB_NOSYNC);
}
return;
}
void
trap(struct trapframe *frame)
{
struct thread *td = curthread;
struct proc *p = td->td_proc;
int i = 0, ucode = 0, code;
u_int type;
register_t addr = 0;
vm_offset_t eva;
ksiginfo_t ksi;
#ifdef POWERFAIL_NMI
static int lastalert = 0;
#endif
PCPU_INC(cnt.v_trap);
type = frame->tf_trapno;
#ifdef SMP
/* Handler for NMI IPIs used for stopping CPUs. */
if (type == T_NMI) {
if (ipi_nmi_handler() == 0)
goto out;
}
#endif /* SMP */
#ifdef KDB
if (kdb_active) {
kdb_reenter();
goto out;
}
#endif
if (type == T_RESERVED) {
trap_fatal(frame, 0);
goto out;
}
#ifdef HWPMC_HOOKS
/*
* CPU PMCs interrupt using an NMI so we check for that first.
* If the HWPMC module is active, 'pmc_hook' will point to
* the function to be called. A return value of '1' from the
* hook means that the NMI was handled by it and that we can
* return immediately.
*/
if (type == T_NMI && pmc_intr &&
(*pmc_intr)(PCPU_GET(cpuid), frame))
goto out;
#endif
if (type == T_MCHK) {
if (!mca_intr())
trap_fatal(frame, 0);
goto out;
}
#ifdef KDTRACE_HOOKS
/*
* A trap can occur while DTrace executes a probe. Before
* executing the probe, DTrace blocks re-scheduling and sets
* a flag in it's per-cpu flags to indicate that it doesn't
* want to fault. On returning from the probe, the no-fault
* flag is cleared and finally re-scheduling is enabled.
*
* If the DTrace kernel module has registered a trap handler,
* call it and if it returns non-zero, assume that it has
* handled the trap and modified the trap frame so that this
* function can return normally.
*/
if ((type == T_PROTFLT || type == T_PAGEFLT) &&
dtrace_trap_func != NULL)
if ((*dtrace_trap_func)(frame, type))
goto out;
if (type == T_DTRACE_PROBE || type == T_DTRACE_RET ||
type == T_BPTFLT) {
struct reg regs;
fill_frame_regs(frame, ®s);
if (type == T_DTRACE_PROBE &&
dtrace_fasttrap_probe_ptr != NULL &&
dtrace_fasttrap_probe_ptr(®s) == 0)
goto out;
if (type == T_BPTFLT &&
dtrace_pid_probe_ptr != NULL &&
dtrace_pid_probe_ptr(®s) == 0)
goto out;
if (type == T_DTRACE_RET &&
dtrace_return_probe_ptr != NULL &&
dtrace_return_probe_ptr(®s) == 0)
goto out;
}
#endif
if ((frame->tf_eflags & PSL_I) == 0) {
/*
* Buggy application or kernel code has disabled
* interrupts and then trapped. Enabling interrupts
* now is wrong, but it is better than running with
* interrupts disabled until they are accidentally
* enabled later.
*/
if (ISPL(frame->tf_cs) == SEL_UPL || (frame->tf_eflags & PSL_VM))
uprintf(
"pid %ld (%s): trap %d with interrupts disabled\n",
(long)curproc->p_pid, curthread->td_name, type);
else if (type != T_BPTFLT && type != T_TRCTRAP &&
frame->tf_eip != (int)cpu_switch_load_gs) {
/*
* XXX not quite right, since this may be for a
* multiple fault in user mode.
*/
printf("kernel trap %d with interrupts disabled\n",
type);
/*
* Page faults need interrupts disabled until later,
* and we shouldn't enable interrupts while holding
* a spin lock or if servicing an NMI.
*/
if (type != T_NMI && type != T_PAGEFLT &&
td->td_md.md_spinlock_count == 0)
enable_intr();
}
}
eva = 0;
code = frame->tf_err;
if (type == T_PAGEFLT) {
/*
* For some Cyrix CPUs, %cr2 is clobbered by
* interrupts. This problem is worked around by using
* an interrupt gate for the pagefault handler. We
* are finally ready to read %cr2 and then must
* reenable interrupts.
*
* If we get a page fault while in a critical section, then
* it is most likely a fatal kernel page fault. The kernel
* is already going to panic trying to get a sleep lock to
* do the VM lookup, so just consider it a fatal trap so the
* kernel can print out a useful trap message and even get
* to the debugger.
*
* If we get a page fault while holding a non-sleepable
* lock, then it is most likely a fatal kernel page fault.
* If WITNESS is enabled, then it's going to whine about
* bogus LORs with various VM locks, so just skip to the
* fatal trap handling directly.
*/
eva = rcr2();
if (td->td_critnest != 0 ||
WITNESS_CHECK(WARN_SLEEPOK | WARN_GIANTOK, NULL,
"Kernel page fault") != 0)
trap_fatal(frame, eva);
else
enable_intr();
}
if ((ISPL(frame->tf_cs) == SEL_UPL) ||
((frame->tf_eflags & PSL_VM) &&
!(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL))) {
/* user trap */
td->td_pticks = 0;
td->td_frame = frame;
addr = frame->tf_eip;
if (td->td_ucred != p->p_ucred)
cred_update_thread(td);
switch (type) {
case T_PRIVINFLT: /* privileged instruction fault */
i = SIGILL;
ucode = ILL_PRVOPC;
break;
case T_BPTFLT: /* bpt instruction fault */
case T_TRCTRAP: /* trace trap */
enable_intr();
frame->tf_eflags &= ~PSL_T;
i = SIGTRAP;
ucode = (type == T_TRCTRAP ? TRAP_TRACE : TRAP_BRKPT);
break;
case T_ARITHTRAP: /* arithmetic trap */
#ifdef DEV_NPX
ucode = npxtrap();
if (ucode == -1)
goto userout;
#else
ucode = 0;
#endif
i = SIGFPE;
break;
/*
* The following two traps can happen in
* vm86 mode, and, if so, we want to handle
* them specially.
*/
case T_PROTFLT: /* general protection fault */
case T_STKFLT: /* stack fault */
if (frame->tf_eflags & PSL_VM) {
i = vm86_emulate((struct vm86frame *)frame);
if (i == 0)
goto user;
break;
}
i = SIGBUS;
ucode = (type == T_PROTFLT) ? BUS_OBJERR : BUS_ADRERR;
break;
case T_SEGNPFLT: /* segment not present fault */
i = SIGBUS;
ucode = BUS_ADRERR;
break;
case T_TSSFLT: /* invalid TSS fault */
i = SIGBUS;
ucode = BUS_OBJERR;
break;
case T_DOUBLEFLT: /* double fault */
default:
i = SIGBUS;
ucode = BUS_OBJERR;
break;
case T_PAGEFLT: /* page fault */
i = trap_pfault(frame, TRUE, eva);
#if defined(I586_CPU) && !defined(NO_F00F_HACK)
if (i == -2) {
/*
* The f00f hack workaround has triggered, so
* treat the fault as an illegal instruction
* (T_PRIVINFLT) instead of a page fault.
*/
type = frame->tf_trapno = T_PRIVINFLT;
/* Proceed as in that case. */
ucode = ILL_PRVOPC;
i = SIGILL;
break;
}
#endif
if (i == -1)
goto userout;
if (i == 0)
goto user;
if (i == SIGSEGV)
ucode = SEGV_MAPERR;
else {
if (prot_fault_translation == 0) {
/*
* Autodetect.
* This check also covers the images
* without the ABI-tag ELF note.
*/
if (SV_CURPROC_ABI() == SV_ABI_FREEBSD
&& p->p_osrel >= P_OSREL_SIGSEGV) {
i = SIGSEGV;
ucode = SEGV_ACCERR;
} else {
i = SIGBUS;
ucode = BUS_PAGE_FAULT;
}
} else if (prot_fault_translation == 1) {
/*
* Always compat mode.
*/
i = SIGBUS;
ucode = BUS_PAGE_FAULT;
} else {
/*
* Always SIGSEGV mode.
*/
i = SIGSEGV;
ucode = SEGV_ACCERR;
}
}
addr = eva;
break;
case T_DIVIDE: /* integer divide fault */
ucode = FPE_INTDIV;
i = SIGFPE;
break;
#ifdef DEV_ISA
case T_NMI:
#ifdef POWERFAIL_NMI
#ifndef TIMER_FREQ
# define TIMER_FREQ 1193182
#endif
if (time_second - lastalert > 10) {
log(LOG_WARNING, "NMI: power fail\n");
sysbeep(880, hz);
lastalert = time_second;
}
goto userout;
#else /* !POWERFAIL_NMI */
/* machine/parity/power fail/"kitchen sink" faults */
if (isa_nmi(code) == 0) {
#ifdef KDB
/*
* NMI can be hooked up to a pushbutton
* for debugging.
*/
if (kdb_on_nmi) {
printf ("NMI ... going to debugger\n");
kdb_trap(type, 0, frame);
}
#endif /* KDB */
goto userout;
} else if (panic_on_nmi)
panic("NMI indicates hardware failure");
break;
#endif /* POWERFAIL_NMI */
#endif /* DEV_ISA */
case T_OFLOW: /* integer overflow fault */
ucode = FPE_INTOVF;
i = SIGFPE;
break;
case T_BOUND: /* bounds check fault */
ucode = FPE_FLTSUB;
i = SIGFPE;
break;
case T_DNA:
#ifdef DEV_NPX
KASSERT(PCB_USER_FPU(td->td_pcb),
("kernel FPU ctx has leaked"));
/* transparent fault (due to context switch "late") */
if (npxdna())
goto userout;
#endif
uprintf("pid %d killed due to lack of floating point\n",
p->p_pid);
i = SIGKILL;
ucode = 0;
break;
case T_FPOPFLT: /* FPU operand fetch fault */
ucode = ILL_COPROC;
i = SIGILL;
break;
case T_XMMFLT: /* SIMD floating-point exception */
ucode = 0; /* XXX */
i = SIGFPE;
break;
}
} else {
/* kernel trap */
KASSERT(cold || td->td_ucred != NULL,
("kernel trap doesn't have ucred"));
switch (type) {
case T_PAGEFLT: /* page fault */
(void) trap_pfault(frame, FALSE, eva);
goto out;
case T_DNA:
#ifdef DEV_NPX
KASSERT(!PCB_USER_FPU(td->td_pcb),
("Unregistered use of FPU in kernel"));
if (npxdna())
goto out;
#endif
break;
case T_ARITHTRAP: /* arithmetic trap */
case T_XMMFLT: /* SIMD floating-point exception */
case T_FPOPFLT: /* FPU operand fetch fault */
/*
* XXXKIB for now disable any FPU traps in kernel
* handler registration seems to be overkill
*/
trap_fatal(frame, 0);
goto out;
/*
* The following two traps can happen in
* vm86 mode, and, if so, we want to handle
* them specially.
*/
case T_PROTFLT: /* general protection fault */
case T_STKFLT: /* stack fault */
if (frame->tf_eflags & PSL_VM) {
i = vm86_emulate((struct vm86frame *)frame);
if (i != 0)
/*
* returns to original process
*/
vm86_trap((struct vm86frame *)frame);
goto out;
}
if (type == T_STKFLT)
break;
/* FALL THROUGH */
case T_SEGNPFLT: /* segment not present fault */
if (PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)
break;
/*
* Invalid %fs's and %gs's can be created using
* procfs or PT_SETREGS or by invalidating the
* underlying LDT entry. This causes a fault
* in kernel mode when the kernel attempts to
* switch contexts. Lose the bad context
* (XXX) so that we can continue, and generate
* a signal.
*/
if (frame->tf_eip == (int)cpu_switch_load_gs) {
PCPU_GET(curpcb)->pcb_gs = 0;
#if 0
PROC_LOCK(p);
kern_psignal(p, SIGBUS);
PROC_UNLOCK(p);
#endif
goto out;
}
if (td->td_intr_nesting_level != 0)
break;
/*
* Invalid segment selectors and out of bounds
* %eip's and %esp's can be set up in user mode.
* This causes a fault in kernel mode when the
* kernel tries to return to user mode. We want
* to get this fault so that we can fix the
* problem here and not have to check all the
* selectors and pointers when the user changes
* them.
*/
if (frame->tf_eip == (int)doreti_iret) {
frame->tf_eip = (int)doreti_iret_fault;
goto out;
}
if (frame->tf_eip == (int)doreti_popl_ds) {
frame->tf_eip = (int)doreti_popl_ds_fault;
goto out;
}
if (frame->tf_eip == (int)doreti_popl_es) {
frame->tf_eip = (int)doreti_popl_es_fault;
goto out;
}
if (frame->tf_eip == (int)doreti_popl_fs) {
frame->tf_eip = (int)doreti_popl_fs_fault;
goto out;
}
if (PCPU_GET(curpcb)->pcb_onfault != NULL) {
frame->tf_eip =
(int)PCPU_GET(curpcb)->pcb_onfault;
goto out;
}
break;
case T_TSSFLT:
/*
* PSL_NT can be set in user mode and isn't cleared
* automatically when the kernel is entered. This
* causes a TSS fault when the kernel attempts to
* `iret' because the TSS link is uninitialized. We
* want to get this fault so that we can fix the
* problem here and not every time the kernel is
* entered.
*/
if (frame->tf_eflags & PSL_NT) {
frame->tf_eflags &= ~PSL_NT;
goto out;
}
break;
case T_TRCTRAP: /* trace trap */
if (frame->tf_eip == (int)IDTVEC(lcall_syscall)) {
/*
* We've just entered system mode via the
* syscall lcall. Continue single stepping
* silently until the syscall handler has
* saved the flags.
*/
goto out;
}
if (frame->tf_eip == (int)IDTVEC(lcall_syscall) + 1) {
/*
* The syscall handler has now saved the
* flags. Stop single stepping it.
*/
frame->tf_eflags &= ~PSL_T;
goto out;
}
/*
* Ignore debug register trace traps due to
* accesses in the user's address space, which
* can happen under several conditions such as
* if a user sets a watchpoint on a buffer and
* then passes that buffer to a system call.
* We still want to get TRCTRAPS for addresses
* in kernel space because that is useful when
* debugging the kernel.
*/
if (user_dbreg_trap() &&
!(PCPU_GET(curpcb)->pcb_flags & PCB_VM86CALL)) {
/*
* Reset breakpoint bits because the
* processor doesn't
*/
load_dr6(rdr6() & 0xfffffff0);
goto out;
}
/*
* FALLTHROUGH (TRCTRAP kernel mode, kernel address)
*/
case T_BPTFLT:
/*
* If KDB is enabled, let it handle the debugger trap.
* Otherwise, debugger traps "can't happen".
*/
#ifdef KDB
if (kdb_trap(type, 0, frame))
goto out;
#endif
break;
#ifdef DEV_ISA
case T_NMI:
#ifdef POWERFAIL_NMI
if (time_second - lastalert > 10) {
log(LOG_WARNING, "NMI: power fail\n");
sysbeep(880, hz);
lastalert = time_second;
}
goto out;
#else /* !POWERFAIL_NMI */
/* machine/parity/power fail/"kitchen sink" faults */
if (isa_nmi(code) == 0) {
#ifdef KDB
/*
* NMI can be hooked up to a pushbutton
* for debugging.
*/
if (kdb_on_nmi) {
printf ("NMI ... going to debugger\n");
kdb_trap(type, 0, frame);
}
#endif /* KDB */
goto out;
} else if (panic_on_nmi == 0)
goto out;
/* FALLTHROUGH */
#endif /* POWERFAIL_NMI */
#endif /* DEV_ISA */
}
trap_fatal(frame, eva);
goto out;
}
/* Translate fault for emulators (e.g. Linux) */
if (*p->p_sysent->sv_transtrap)
i = (*p->p_sysent->sv_transtrap)(i, type);
ksiginfo_init_trap(&ksi);
ksi.ksi_signo = i;
ksi.ksi_code = ucode;
ksi.ksi_addr = (void *)addr;
ksi.ksi_trapno = type;
trapsignal(td, &ksi);
#ifdef DEBUG
if (type <= MAX_TRAP_MSG) {
uprintf("fatal process exception: %s",
trap_msg[type]);
if ((type == T_PAGEFLT) || (type == T_PROTFLT))
uprintf(", fault VA = 0x%lx", (u_long)eva);
uprintf("\n");
}
#endif
user:
userret(td, frame);
mtx_assert(&Giant, MA_NOTOWNED);
KASSERT(PCB_USER_FPU(td->td_pcb),
("Return from trap with kernel FPU ctx leaked"));
userout:
out:
return;
}
static void
soaio_process_job(struct socket *so, struct sockbuf *sb, struct kaiocb *job)
{
struct ucred *td_savedcred;
struct thread *td;
struct file *fp;
struct uio uio;
struct iovec iov;
size_t cnt;
int error, flags;
SOCKBUF_UNLOCK(sb);
aio_switch_vmspace(job);
td = curthread;
fp = job->fd_file;
retry:
td_savedcred = td->td_ucred;
td->td_ucred = job->cred;
cnt = job->uaiocb.aio_nbytes;
iov.iov_base = (void *)(uintptr_t)job->uaiocb.aio_buf;
iov.iov_len = cnt;
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = 0;
uio.uio_resid = cnt;
uio.uio_segflg = UIO_USERSPACE;
uio.uio_td = td;
flags = MSG_NBIO;
/* TODO: Charge ru_msg* to job. */
if (sb == &so->so_rcv) {
uio.uio_rw = UIO_READ;
#ifdef MAC
error = mac_socket_check_receive(fp->f_cred, so);
if (error == 0)
#endif
error = soreceive(so, NULL, &uio, NULL, NULL, &flags);
} else {
uio.uio_rw = UIO_WRITE;
#ifdef MAC
error = mac_socket_check_send(fp->f_cred, so);
if (error == 0)
#endif
error = sosend(so, NULL, &uio, NULL, NULL, flags, td);
if (error == EPIPE && (so->so_options & SO_NOSIGPIPE) == 0) {
PROC_LOCK(job->userproc);
kern_psignal(job->userproc, SIGPIPE);
PROC_UNLOCK(job->userproc);
}
}
cnt -= uio.uio_resid;
td->td_ucred = td_savedcred;
/* XXX: Not sure if this is needed? */
if (cnt != 0 && (error == ERESTART || error == EINTR ||
error == EWOULDBLOCK))
error = 0;
if (error == EWOULDBLOCK) {
/*
* A read() or write() on the socket raced with this
* request. If the socket is now ready, try again.
* If it is not, place this request at the head of the
* queue to try again when the socket is ready.
*/
SOCKBUF_LOCK(sb);
empty_results++;
if (soaio_ready(so, sb)) {
empty_retries++;
SOCKBUF_UNLOCK(sb);
goto retry;
}
if (!aio_set_cancel_function(job, soo_aio_cancel)) {
MPASS(cnt == 0);
SOCKBUF_UNLOCK(sb);
aio_cancel(job);
SOCKBUF_LOCK(sb);
} else {
TAILQ_INSERT_HEAD(&sb->sb_aiojobq, job, list);
}
} else {
aio_complete(job, cnt, error);
SOCKBUF_LOCK(sb);
}
}
int
sc_mouse_ioctl(struct tty *tp, u_long cmd, caddr_t data, struct thread *td)
{
mouse_info_t *mouse;
mouse_info_t buf;
scr_stat *cur_scp;
scr_stat *scp;
struct proc *p1;
int s;
int f;
scp = SC_STAT(tp);
switch (cmd) {
case CONS_MOUSECTL: /* control mouse arrow */
case OLD_CONS_MOUSECTL:
mouse = (mouse_info_t*)data;
random_harvest_queue(mouse, sizeof(mouse_info_t), 2, RANDOM_MOUSE);
if (cmd == OLD_CONS_MOUSECTL) {
static u_char swapb[] = { 0, 4, 2, 6, 1, 5, 3, 7 };
old_mouse_info_t *old_mouse = (old_mouse_info_t *)data;
mouse = &buf;
mouse->operation = old_mouse->operation;
switch (mouse->operation) {
case MOUSE_MODE:
mouse->u.mode = old_mouse->u.mode;
break;
case MOUSE_SHOW:
case MOUSE_HIDE:
break;
case MOUSE_MOVEABS:
case MOUSE_MOVEREL:
case MOUSE_ACTION:
mouse->u.data.x = old_mouse->u.data.x;
mouse->u.data.y = old_mouse->u.data.y;
mouse->u.data.z = 0;
mouse->u.data.buttons = swapb[old_mouse->u.data.buttons & 0x7];
break;
case MOUSE_GETINFO:
old_mouse->u.data.x = scp->mouse_xpos;
old_mouse->u.data.y = scp->mouse_ypos;
old_mouse->u.data.buttons = swapb[scp->mouse_buttons & 0x7];
return 0;
default:
return EINVAL;
}
}
cur_scp = scp->sc->cur_scp;
switch (mouse->operation) {
case MOUSE_MODE:
if (ISSIGVALID(mouse->u.mode.signal)) {
scp->mouse_signal = mouse->u.mode.signal;
scp->mouse_proc = td->td_proc;
scp->mouse_pid = td->td_proc->p_pid;
}
else {
scp->mouse_signal = 0;
scp->mouse_proc = NULL;
scp->mouse_pid = 0;
}
return 0;
case MOUSE_SHOW:
s = spltty();
if (!(scp->sc->flags & SC_MOUSE_ENABLED)) {
scp->sc->flags |= SC_MOUSE_ENABLED;
cur_scp->status &= ~MOUSE_HIDDEN;
if (!ISGRAPHSC(cur_scp))
mark_all(cur_scp);
}
splx(s);
return 0;
/* NOTREACHED */
case MOUSE_HIDE:
s = spltty();
if (scp->sc->flags & SC_MOUSE_ENABLED) {
scp->sc->flags &= ~SC_MOUSE_ENABLED;
sc_remove_all_mouse(scp->sc);
}
splx(s);
return 0;
/* NOTREACHED */
case MOUSE_MOVEABS:
s = spltty();
scp->mouse_xpos = mouse->u.data.x;
scp->mouse_ypos = mouse->u.data.y;
set_mouse_pos(scp);
splx(s);
break;
case MOUSE_MOVEREL:
s = spltty();
scp->mouse_xpos += mouse->u.data.x;
scp->mouse_ypos += mouse->u.data.y;
set_mouse_pos(scp);
splx(s);
break;
case MOUSE_GETINFO:
mouse->u.data.x = scp->mouse_xpos;
mouse->u.data.y = scp->mouse_ypos;
mouse->u.data.z = 0;
mouse->u.data.buttons = scp->mouse_buttons;
return 0;
case MOUSE_ACTION:
case MOUSE_MOTION_EVENT:
/* send out mouse event on /dev/sysmouse */
#if 0
/* this should maybe only be settable from /dev/consolectl SOS */
if (SC_VTY(tp->t_dev) != SC_CONSOLECTL)
return ENOTTY;
#endif
s = spltty();
if (mouse->u.data.x != 0 || mouse->u.data.y != 0) {
cur_scp->mouse_xpos += mouse->u.data.x;
cur_scp->mouse_ypos += mouse->u.data.y;
set_mouse_pos(cur_scp);
}
f = 0;
if (mouse->operation == MOUSE_ACTION) {
f = cur_scp->mouse_buttons ^ mouse->u.data.buttons;
cur_scp->mouse_buttons = mouse->u.data.buttons;
}
splx(s);
if (sysmouse_event(mouse) == 0)
return 0;
/*
* If any buttons are down or the mouse has moved a lot,
* stop the screen saver.
*/
if (((mouse->operation == MOUSE_ACTION) && mouse->u.data.buttons)
|| (mouse->u.data.x*mouse->u.data.x
+ mouse->u.data.y*mouse->u.data.y
>= SC_WAKEUP_DELTA*SC_WAKEUP_DELTA)) {
sc_touch_scrn_saver();
}
cur_scp->status &= ~MOUSE_HIDDEN;
if (cur_scp->mouse_level > 0) {
sc_mouse_input(scp, mouse);
break;
}
if (cur_scp->mouse_signal && cur_scp->mouse_proc) {
/* has controlling process died? */
if (cur_scp->mouse_proc != (p1 = pfind(cur_scp->mouse_pid))) {
cur_scp->mouse_signal = 0;
cur_scp->mouse_proc = NULL;
cur_scp->mouse_pid = 0;
if (p1)
PROC_UNLOCK(p1);
} else {
kern_psignal(cur_scp->mouse_proc, cur_scp->mouse_signal);
PROC_UNLOCK(cur_scp->mouse_proc);
break;
}
}
#ifndef SC_NO_CUTPASTE
if (ISGRAPHSC(cur_scp) || (cut_buffer == NULL))
break;
if ((mouse->operation == MOUSE_ACTION) && f) {
/* process button presses */
if (cur_scp->mouse_buttons & MOUSE_BUTTON1DOWN)
mouse_cut_start(cur_scp);
else
mouse_cut_end(cur_scp);
if (cur_scp->mouse_buttons & MOUSE_BUTTON2DOWN ||
cur_scp->mouse_buttons & MOUSE_BUTTON3DOWN)
sc_mouse_paste(cur_scp);
}
#endif /* SC_NO_CUTPASTE */
break;
case MOUSE_BUTTON_EVENT:
if ((mouse->u.event.id & MOUSE_BUTTONS) == 0)
return EINVAL;
if (mouse->u.event.value < 0)
return EINVAL;
#if 0
/* this should maybe only be settable from /dev/consolectl SOS */
if (SC_VTY(tp->t_dev) != SC_CONSOLECTL)
return ENOTTY;
#endif
if (mouse->u.event.value > 0)
cur_scp->mouse_buttons |= mouse->u.event.id;
else
cur_scp->mouse_buttons &= ~mouse->u.event.id;
if (sysmouse_event(mouse) == 0)
return 0;
/* if a button is held down, stop the screen saver */
if (mouse->u.event.value > 0)
sc_touch_scrn_saver();
cur_scp->status &= ~MOUSE_HIDDEN;
if (cur_scp->mouse_level > 0) {
sc_mouse_input(scp, mouse);
break;
}
if (cur_scp->mouse_signal && cur_scp->mouse_proc) {
if (cur_scp->mouse_proc != (p1 = pfind(cur_scp->mouse_pid))){
cur_scp->mouse_signal = 0;
cur_scp->mouse_proc = NULL;
cur_scp->mouse_pid = 0;
if (p1)
PROC_UNLOCK(p1);
} else {
kern_psignal(cur_scp->mouse_proc, cur_scp->mouse_signal);
PROC_UNLOCK(cur_scp->mouse_proc);
break;
}
}
#ifndef SC_NO_CUTPASTE
if (ISGRAPHSC(cur_scp) || (cut_buffer == NULL))
break;
switch (mouse->u.event.id) {
case MOUSE_BUTTON1DOWN:
switch (mouse->u.event.value % 4) {
case 0: /* up */
mouse_cut_end(cur_scp);
break;
case 1: /* single click: start cut operation */
mouse_cut_start(cur_scp);
break;
case 2: /* double click: cut a word */
mouse_cut_word(cur_scp);
mouse_cut_end(cur_scp);
break;
case 3: /* triple click: cut a line */
mouse_cut_line(cur_scp);
mouse_cut_end(cur_scp);
break;
}
break;
case SC_MOUSE_PASTEBUTTON:
switch (mouse->u.event.value) {
case 0: /* up */
break;
default:
sc_mouse_paste(cur_scp);
break;
}
break;
case SC_MOUSE_EXTENDBUTTON:
switch (mouse->u.event.value) {
case 0: /* up */
if (!(cur_scp->mouse_buttons & MOUSE_BUTTON1DOWN))
mouse_cut_end(cur_scp);
break;
default:
mouse_cut_extend(cur_scp);
break;
}
break;
}
#endif /* SC_NO_CUTPASTE */
break;
case MOUSE_MOUSECHAR:
if (mouse->u.mouse_char < 0) {
mouse->u.mouse_char = scp->sc->mouse_char;
} else {
if (mouse->u.mouse_char > UCHAR_MAX - 3)
return EINVAL;
s = spltty();
sc_remove_all_mouse(scp->sc);
#ifndef SC_NO_FONT_LOADING
if (ISTEXTSC(cur_scp) && (cur_scp->font != NULL))
sc_load_font(cur_scp, 0, cur_scp->font_size,
cur_scp->font_width,
cur_scp->font + cur_scp->font_size
* cur_scp->sc->mouse_char,
cur_scp->sc->mouse_char, 4);
#endif
scp->sc->mouse_char = mouse->u.mouse_char;
splx(s);
}
break;
default:
return EINVAL;
}
return 0;
}
return ENOIOCTL;
}
/*
* Process an asynchronous software trap.
* This is relatively easy.
* This function will return with preemption disabled.
*/
void
ast(struct trapframe *framep)
{
struct thread *td;
struct proc *p;
int flags;
int sig;
td = curthread;
p = td->td_proc;
CTR3(KTR_SYSC, "ast: thread %p (pid %d, %s)", td, p->p_pid,
p->p_comm);
KASSERT(TRAPF_USERMODE(framep), ("ast in kernel mode"));
WITNESS_WARN(WARN_PANIC, NULL, "Returning to user mode");
mtx_assert(&Giant, MA_NOTOWNED);
THREAD_LOCK_ASSERT(td, MA_NOTOWNED);
td->td_frame = framep;
td->td_pticks = 0;
/*
* This updates the td_flag's for the checks below in one
* "atomic" operation with turning off the astpending flag.
* If another AST is triggered while we are handling the
* AST's saved in flags, the astpending flag will be set and
* ast() will be called again.
*/
thread_lock(td);
flags = td->td_flags;
td->td_flags &= ~(TDF_ASTPENDING | TDF_NEEDSIGCHK | TDF_NEEDSUSPCHK |
TDF_NEEDRESCHED | TDF_ALRMPEND | TDF_PROFPEND | TDF_MACPEND);
thread_unlock(td);
PCPU_INC(cnt.v_trap);
if (td->td_ucred != p->p_ucred)
cred_update_thread(td);
if (td->td_pflags & TDP_OWEUPC && p->p_flag & P_PROFIL) {
addupc_task(td, td->td_profil_addr, td->td_profil_ticks);
td->td_profil_ticks = 0;
td->td_pflags &= ~TDP_OWEUPC;
}
#ifdef HWPMC_HOOKS
/* Handle Software PMC callchain capture. */
if (PMC_IS_PENDING_CALLCHAIN(td))
PMC_CALL_HOOK_UNLOCKED(td, PMC_FN_USER_CALLCHAIN_SOFT, (void *) framep);
#endif
if (flags & TDF_ALRMPEND) {
PROC_LOCK(p);
kern_psignal(p, SIGVTALRM);
PROC_UNLOCK(p);
}
if (flags & TDF_PROFPEND) {
PROC_LOCK(p);
kern_psignal(p, SIGPROF);
PROC_UNLOCK(p);
}
#ifdef MAC
if (flags & TDF_MACPEND)
mac_thread_userret(td);
#endif
if (flags & TDF_NEEDRESCHED) {
#ifdef KTRACE
if (KTRPOINT(td, KTR_CSW))
ktrcsw(1, 1, __func__);
#endif
thread_lock(td);
sched_prio(td, td->td_user_pri);
mi_switch(SW_INVOL | SWT_NEEDRESCHED, NULL);
thread_unlock(td);
#ifdef KTRACE
if (KTRPOINT(td, KTR_CSW))
ktrcsw(0, 1, __func__);
#endif
}
/*
* Check for signals. Unlocked reads of p_pendingcnt or
* p_siglist might cause process-directed signal to be handled
* later.
*/
if (flags & TDF_NEEDSIGCHK || p->p_pendingcnt > 0 ||
!SIGISEMPTY(p->p_siglist)) {
PROC_LOCK(p);
mtx_lock(&p->p_sigacts->ps_mtx);
while ((sig = cursig(td)) != 0)
postsig(sig);
mtx_unlock(&p->p_sigacts->ps_mtx);
PROC_UNLOCK(p);
}
/*
* We need to check to see if we have to exit or wait due to a
* single threading requirement or some other STOP condition.
*/
if (flags & TDF_NEEDSUSPCHK) {
PROC_LOCK(p);
thread_suspend_check(0);
PROC_UNLOCK(p);
}
if (td->td_pflags & TDP_OLDMASK) {
td->td_pflags &= ~TDP_OLDMASK;
kern_sigprocmask(td, SIG_SETMASK, &td->td_oldsigmask, NULL, 0);
}
userret(td, framep);
}
static void
pmclog_loop(void *arg)
{
int error;
struct pmc_owner *po;
struct pmclog_buffer *lb;
struct proc *p;
struct ucred *ownercred;
struct ucred *mycred;
struct thread *td;
struct uio auio;
struct iovec aiov;
size_t nbytes;
po = (struct pmc_owner *) arg;
p = po->po_owner;
td = curthread;
mycred = td->td_ucred;
PROC_LOCK(p);
ownercred = crhold(p->p_ucred);
PROC_UNLOCK(p);
PMCDBG(LOG,INI,1, "po=%p kt=%p", po, po->po_kthread);
KASSERT(po->po_kthread == curthread->td_proc,
("[pmclog,%d] proc mismatch po=%p po/kt=%p curproc=%p", __LINE__,
po, po->po_kthread, curthread->td_proc));
lb = NULL;
/*
* Loop waiting for I/O requests to be added to the owner
* struct's queue. The loop is exited when the log file
* is deconfigured.
*/
mtx_lock(&pmc_kthread_mtx);
for (;;) {
/* check if we've been asked to exit */
if ((po->po_flags & PMC_PO_OWNS_LOGFILE) == 0)
break;
if (lb == NULL) { /* look for a fresh buffer to write */
mtx_lock_spin(&po->po_mtx);
if ((lb = TAILQ_FIRST(&po->po_logbuffers)) == NULL) {
mtx_unlock_spin(&po->po_mtx);
/* No more buffers and shutdown required. */
if (po->po_flags & PMC_PO_SHUTDOWN) {
mtx_unlock(&pmc_kthread_mtx);
/*
* Close the file to get PMCLOG_EOF
* error in pmclog(3).
*/
fo_close(po->po_file, curthread);
mtx_lock(&pmc_kthread_mtx);
break;
}
(void) msleep(po, &pmc_kthread_mtx, PWAIT,
"pmcloop", 0);
continue;
}
TAILQ_REMOVE(&po->po_logbuffers, lb, plb_next);
mtx_unlock_spin(&po->po_mtx);
}
mtx_unlock(&pmc_kthread_mtx);
/* process the request */
PMCDBG(LOG,WRI,2, "po=%p base=%p ptr=%p", po,
lb->plb_base, lb->plb_ptr);
/* change our thread's credentials before issuing the I/O */
aiov.iov_base = lb->plb_base;
aiov.iov_len = nbytes = lb->plb_ptr - lb->plb_base;
auio.uio_iov = &aiov;
auio.uio_iovcnt = 1;
auio.uio_offset = -1;
auio.uio_resid = nbytes;
auio.uio_rw = UIO_WRITE;
auio.uio_segflg = UIO_SYSSPACE;
auio.uio_td = td;
/* switch thread credentials -- see kern_ktrace.c */
td->td_ucred = ownercred;
error = fo_write(po->po_file, &auio, ownercred, 0, td);
td->td_ucred = mycred;
if (error) {
/* XXX some errors are recoverable */
/* send a SIGIO to the owner and exit */
PROC_LOCK(p);
kern_psignal(p, SIGIO);
PROC_UNLOCK(p);
mtx_lock(&pmc_kthread_mtx);
po->po_error = error; /* save for flush log */
PMCDBG(LOG,WRI,2, "po=%p error=%d", po, error);
break;
}
mtx_lock(&pmc_kthread_mtx);
/* put the used buffer back into the global pool */
PMCLOG_INIT_BUFFER_DESCRIPTOR(lb);
mtx_lock_spin(&pmc_bufferlist_mtx);
TAILQ_INSERT_HEAD(&pmc_bufferlist, lb, plb_next);
mtx_unlock_spin(&pmc_bufferlist_mtx);
lb = NULL;
}
wakeup_one(po->po_kthread);
po->po_kthread = NULL;
mtx_unlock(&pmc_kthread_mtx);
/* return the current I/O buffer to the global pool */
if (lb) {
PMCLOG_INIT_BUFFER_DESCRIPTOR(lb);
mtx_lock_spin(&pmc_bufferlist_mtx);
TAILQ_INSERT_HEAD(&pmc_bufferlist, lb, plb_next);
mtx_unlock_spin(&pmc_bufferlist_mtx);
}
/*
* Exit this thread, signalling the waiter
*/
crfree(ownercred);
kproc_exit(0);
}
/*
* Exit: deallocate address space and other resources, change proc state to
* zombie, and unlink proc from allproc and parent's lists. Save exit status
* and rusage for wait(). Check for child processes and orphan them.
*/
void
exit1(struct thread *td, int rv)
{
struct proc *p, *nq, *q;
struct vnode *vtmp;
struct vnode *ttyvp = NULL;
struct plimit *plim;
mtx_assert(&Giant, MA_NOTOWNED);
p = td->td_proc;
/*
* XXX in case we're rebooting we just let init die in order to
* work around an unsolved stack overflow seen very late during
* shutdown on sparc64 when the gmirror worker process exists.
*/
if (p == initproc && rebooting == 0) {
printf("init died (signal %d, exit %d)\n",
WTERMSIG(rv), WEXITSTATUS(rv));
panic("Going nowhere without my init!");
}
/*
* MUST abort all other threads before proceeding past here.
*/
PROC_LOCK(p);
while (p->p_flag & P_HADTHREADS) {
/*
* First check if some other thread got here before us.
* If so, act appropriately: exit or suspend.
*/
thread_suspend_check(0);
/*
* Kill off the other threads. This requires
* some co-operation from other parts of the kernel
* so it may not be instantaneous. With this state set
* any thread entering the kernel from userspace will
* thread_exit() in trap(). Any thread attempting to
* sleep will return immediately with EINTR or EWOULDBLOCK
* which will hopefully force them to back out to userland
* freeing resources as they go. Any thread attempting
* to return to userland will thread_exit() from userret().
* thread_exit() will unsuspend us when the last of the
* other threads exits.
* If there is already a thread singler after resumption,
* calling thread_single will fail; in that case, we just
* re-check all suspension request, the thread should
* either be suspended there or exit.
*/
if (!thread_single(SINGLE_EXIT))
break;
/*
* All other activity in this process is now stopped.
* Threading support has been turned off.
*/
}
KASSERT(p->p_numthreads == 1,
("exit1: proc %p exiting with %d threads", p, p->p_numthreads));
racct_sub(p, RACCT_NTHR, 1);
/*
* Wakeup anyone in procfs' PIOCWAIT. They should have a hold
* on our vmspace, so we should block below until they have
* released their reference to us. Note that if they have
* requested S_EXIT stops we will block here until they ack
* via PIOCCONT.
*/
_STOPEVENT(p, S_EXIT, rv);
/*
* Ignore any pending request to stop due to a stop signal.
* Once P_WEXIT is set, future requests will be ignored as
* well.
*/
p->p_flag &= ~P_STOPPED_SIG;
KASSERT(!P_SHOULDSTOP(p), ("exiting process is stopped"));
/*
* Note that we are exiting and do another wakeup of anyone in
* PIOCWAIT in case they aren't listening for S_EXIT stops or
* decided to wait again after we told them we are exiting.
*/
p->p_flag |= P_WEXIT;
wakeup(&p->p_stype);
/*
* Wait for any processes that have a hold on our vmspace to
* release their reference.
*/
while (p->p_lock > 0)
msleep(&p->p_lock, &p->p_mtx, PWAIT, "exithold", 0);
p->p_xstat = rv; /* Let event handler change exit status */
PROC_UNLOCK(p);
/* Drain the limit callout while we don't have the proc locked */
callout_drain(&p->p_limco);
#ifdef AUDIT
/*
* The Sun BSM exit token contains two components: an exit status as
* passed to exit(), and a return value to indicate what sort of exit
* it was. The exit status is WEXITSTATUS(rv), but it's not clear
* what the return value is.
*/
AUDIT_ARG_EXIT(WEXITSTATUS(rv), 0);
AUDIT_SYSCALL_EXIT(0, td);
#endif
/* Are we a task leader? */
if (p == p->p_leader) {
mtx_lock(&ppeers_lock);
q = p->p_peers;
while (q != NULL) {
PROC_LOCK(q);
kern_psignal(q, SIGKILL);
PROC_UNLOCK(q);
q = q->p_peers;
}
while (p->p_peers != NULL)
msleep(p, &ppeers_lock, PWAIT, "exit1", 0);
mtx_unlock(&ppeers_lock);
}
/*
* Check if any loadable modules need anything done at process exit.
* E.g. SYSV IPC stuff
* XXX what if one of these generates an error?
*/
EVENTHANDLER_INVOKE(process_exit, p);
/*
* If parent is waiting for us to exit or exec,
* P_PPWAIT is set; we will wakeup the parent below.
*/
PROC_LOCK(p);
rv = p->p_xstat; /* Event handler could change exit status */
stopprofclock(p);
p->p_flag &= ~(P_TRACED | P_PPWAIT | P_PPTRACE);
/*
* Stop the real interval timer. If the handler is currently
* executing, prevent it from rearming itself and let it finish.
*/
if (timevalisset(&p->p_realtimer.it_value) &&
callout_stop(&p->p_itcallout) == 0) {
timevalclear(&p->p_realtimer.it_interval);
msleep(&p->p_itcallout, &p->p_mtx, PWAIT, "ritwait", 0);
KASSERT(!timevalisset(&p->p_realtimer.it_value),
("realtime timer is still armed"));
}
PROC_UNLOCK(p);
/*
* Reset any sigio structures pointing to us as a result of
* F_SETOWN with our pid.
*/
funsetownlst(&p->p_sigiolst);
/*
* If this process has an nlminfo data area (for lockd), release it
*/
if (nlminfo_release_p != NULL && p->p_nlminfo != NULL)
(*nlminfo_release_p)(p);
/*
* Close open files and release open-file table.
* This may block!
*/
fdescfree(td);
/*
* If this thread tickled GEOM, we need to wait for the giggling to
* stop before we return to userland
*/
if (td->td_pflags & TDP_GEOM)
g_waitidle();
/*
* Remove ourself from our leader's peer list and wake our leader.
*/
mtx_lock(&ppeers_lock);
if (p->p_leader->p_peers) {
q = p->p_leader;
while (q->p_peers != p)
q = q->p_peers;
q->p_peers = p->p_peers;
wakeup(p->p_leader);
}
mtx_unlock(&ppeers_lock);
vmspace_exit(td);
sx_xlock(&proctree_lock);
if (SESS_LEADER(p)) {
struct session *sp = p->p_session;
struct tty *tp;
/*
* s_ttyp is not zero'd; we use this to indicate that
* the session once had a controlling terminal. (for
* logging and informational purposes)
*/
SESS_LOCK(sp);
ttyvp = sp->s_ttyvp;
tp = sp->s_ttyp;
sp->s_ttyvp = NULL;
sp->s_ttydp = NULL;
sp->s_leader = NULL;
SESS_UNLOCK(sp);
/*
* Signal foreground pgrp and revoke access to
* controlling terminal if it has not been revoked
* already.
*
* Because the TTY may have been revoked in the mean
* time and could already have a new session associated
* with it, make sure we don't send a SIGHUP to a
* foreground process group that does not belong to this
* session.
*/
if (tp != NULL) {
tty_lock(tp);
if (tp->t_session == sp)
tty_signal_pgrp(tp, SIGHUP);
tty_unlock(tp);
}
if (ttyvp != NULL) {
sx_xunlock(&proctree_lock);
if (vn_lock(ttyvp, LK_EXCLUSIVE) == 0) {
VOP_REVOKE(ttyvp, REVOKEALL);
VOP_UNLOCK(ttyvp, 0);
}
sx_xlock(&proctree_lock);
}
}
fixjobc(p, p->p_pgrp, 0);
sx_xunlock(&proctree_lock);
(void)acct_process(td);
/* Release the TTY now we've unlocked everything. */
if (ttyvp != NULL)
vrele(ttyvp);
#ifdef KTRACE
ktrprocexit(td);
#endif
/*
* Release reference to text vnode
*/
if ((vtmp = p->p_textvp) != NULL) {
p->p_textvp = NULL;
vrele(vtmp);
}
/*
* Release our limits structure.
*/
plim = p->p_limit;
p->p_limit = NULL;
lim_free(plim);
tidhash_remove(td);
/*
* Remove proc from allproc queue and pidhash chain.
* Place onto zombproc. Unlink from parent's child list.
*/
sx_xlock(&allproc_lock);
LIST_REMOVE(p, p_list);
LIST_INSERT_HEAD(&zombproc, p, p_list);
LIST_REMOVE(p, p_hash);
sx_xunlock(&allproc_lock);
/*
* Call machine-dependent code to release any
* machine-dependent resources other than the address space.
* The address space is released by "vmspace_exitfree(p)" in
* vm_waitproc().
*/
cpu_exit(td);
WITNESS_WARN(WARN_PANIC, NULL, "process (pid %d) exiting", p->p_pid);
/*
* Reparent all of our children to init.
*/
sx_xlock(&proctree_lock);
q = LIST_FIRST(&p->p_children);
if (q != NULL) /* only need this if any child is S_ZOMB */
wakeup(initproc);
for (; q != NULL; q = nq) {
nq = LIST_NEXT(q, p_sibling);
PROC_LOCK(q);
proc_reparent(q, initproc);
q->p_sigparent = SIGCHLD;
/*
* Traced processes are killed
* since their existence means someone is screwing up.
*/
if (q->p_flag & P_TRACED) {
struct thread *temp;
/*
* Since q was found on our children list, the
* proc_reparent() call moved q to the orphan
* list due to present P_TRACED flag. Clear
* orphan link for q now while q is locked.
*/
clear_orphan(q);
q->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
FOREACH_THREAD_IN_PROC(q, temp)
temp->td_dbgflags &= ~TDB_SUSPEND;
kern_psignal(q, SIGKILL);
}
PROC_UNLOCK(q);
}
/*
* Also get rid of our orphans.
*/
while ((q = LIST_FIRST(&p->p_orphans)) != NULL) {
PROC_LOCK(q);
clear_orphan(q);
PROC_UNLOCK(q);
}
/* Save exit status. */
PROC_LOCK(p);
p->p_xthread = td;
/* Tell the prison that we are gone. */
prison_proc_free(p->p_ucred->cr_prison);
#ifdef KDTRACE_HOOKS
/*
* Tell the DTrace fasttrap provider about the exit if it
* has declared an interest.
*/
if (dtrace_fasttrap_exit)
dtrace_fasttrap_exit(p);
#endif
/*
* Notify interested parties of our demise.
*/
KNOTE_LOCKED(&p->p_klist, NOTE_EXIT);
#ifdef KDTRACE_HOOKS
int reason = CLD_EXITED;
if (WCOREDUMP(rv))
reason = CLD_DUMPED;
else if (WIFSIGNALED(rv))
reason = CLD_KILLED;
SDT_PROBE(proc, kernel, , exit, reason, 0, 0, 0, 0);
#endif
/*
* Just delete all entries in the p_klist. At this point we won't
* report any more events, and there are nasty race conditions that
* can beat us if we don't.
*/
knlist_clear(&p->p_klist, 1);
/*
* If this is a process with a descriptor, we may not need to deliver
* a signal to the parent. proctree_lock is held over
* procdesc_exit() to serialize concurrent calls to close() and
* exit().
*/
if (p->p_procdesc == NULL || procdesc_exit(p)) {
/*
* Notify parent that we're gone. If parent has the
* PS_NOCLDWAIT flag set, or if the handler is set to SIG_IGN,
* notify process 1 instead (and hope it will handle this
* situation).
*/
PROC_LOCK(p->p_pptr);
mtx_lock(&p->p_pptr->p_sigacts->ps_mtx);
if (p->p_pptr->p_sigacts->ps_flag &
(PS_NOCLDWAIT | PS_CLDSIGIGN)) {
struct proc *pp;
mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
pp = p->p_pptr;
PROC_UNLOCK(pp);
proc_reparent(p, initproc);
p->p_sigparent = SIGCHLD;
PROC_LOCK(p->p_pptr);
/*
* Notify parent, so in case he was wait(2)ing or
* executing waitpid(2) with our pid, he will
* continue.
*/
wakeup(pp);
} else
mtx_unlock(&p->p_pptr->p_sigacts->ps_mtx);
if (p->p_pptr == initproc)
kern_psignal(p->p_pptr, SIGCHLD);
else if (p->p_sigparent != 0) {
if (p->p_sigparent == SIGCHLD)
childproc_exited(p);
else /* LINUX thread */
kern_psignal(p->p_pptr, p->p_sigparent);
}
} else
PROC_LOCK(p->p_pptr);
sx_xunlock(&proctree_lock);
/*
* The state PRS_ZOMBIE prevents other proesses from sending
* signal to the process, to avoid memory leak, we free memory
* for signal queue at the time when the state is set.
*/
sigqueue_flush(&p->p_sigqueue);
sigqueue_flush(&td->td_sigqueue);
/*
* We have to wait until after acquiring all locks before
* changing p_state. We need to avoid all possible context
* switches (including ones from blocking on a mutex) while
* marked as a zombie. We also have to set the zombie state
* before we release the parent process' proc lock to avoid
* a lost wakeup. So, we first call wakeup, then we grab the
* sched lock, update the state, and release the parent process'
* proc lock.
*/
wakeup(p->p_pptr);
cv_broadcast(&p->p_pwait);
sched_exit(p->p_pptr, td);
PROC_SLOCK(p);
p->p_state = PRS_ZOMBIE;
PROC_UNLOCK(p->p_pptr);
/*
* Hopefully no one will try to deliver a signal to the process this
* late in the game.
*/
knlist_destroy(&p->p_klist);
/*
* Save our children's rusage information in our exit rusage.
*/
ruadd(&p->p_ru, &p->p_rux, &p->p_stats->p_cru, &p->p_crux);
/*
* Make sure the scheduler takes this thread out of its tables etc.
* This will also release this thread's reference to the ucred.
* Other thread parts to release include pcb bits and such.
*/
thread_exit();
}
static int
procfs_control(struct thread *td, struct proc *p, int op)
{
int error = 0;
struct thread *temp;
/*
* Attach - attaches the target process for debugging
* by the calling process.
*/
if (op == PROCFS_CTL_ATTACH) {
sx_xlock(&proctree_lock);
PROC_LOCK(p);
if ((error = p_candebug(td, p)) != 0)
goto out;
if (p->p_flag & P_TRACED) {
error = EBUSY;
goto out;
}
/* Can't trace yourself! */
if (p->p_pid == td->td_proc->p_pid) {
error = EINVAL;
goto out;
}
/*
* Go ahead and set the trace flag.
* Save the old parent (it's reset in
* _DETACH, and also in kern_exit.c:wait4()
* Reparent the process so that the tracing
* proc gets to see all the action.
* Stop the target.
*/
p->p_flag |= P_TRACED;
faultin(p);
p->p_xstat = 0; /* XXX ? */
if (p->p_pptr != td->td_proc) {
p->p_oppid = p->p_pptr->p_pid;
proc_reparent(p, td->td_proc);
}
kern_psignal(p, SIGSTOP);
out:
PROC_UNLOCK(p);
sx_xunlock(&proctree_lock);
return (error);
}
/*
* Authorization check: rely on normal debugging protection, except
* allow processes to disengage debugging on a process onto which
* they have previously attached, but no longer have permission to
* debug.
*/
PROC_LOCK(p);
if (op != PROCFS_CTL_DETACH &&
((error = p_candebug(td, p)))) {
PROC_UNLOCK(p);
return (error);
}
/*
* Target process must be stopped, owned by (td) and
* be set up for tracing (P_TRACED flag set).
* Allow DETACH to take place at any time for sanity.
* Allow WAIT any time, of course.
*/
switch (op) {
case PROCFS_CTL_DETACH:
case PROCFS_CTL_WAIT:
break;
default:
if (!TRACE_WAIT_P(td->td_proc, p)) {
PROC_UNLOCK(p);
return (EBUSY);
}
}
#ifdef FIX_SSTEP
/*
* do single-step fixup if needed
*/
FIX_SSTEP(FIRST_THREAD_IN_PROC(p));
#endif
/*
* Don't deliver any signal by default.
* To continue with a signal, just send
* the signal name to the ctl file
*/
p->p_xstat = 0;
switch (op) {
/*
* Detach. Cleans up the target process, reparent it if possible
* and set it running once more.
*/
case PROCFS_CTL_DETACH:
/* if not being traced, then this is a painless no-op */
if ((p->p_flag & P_TRACED) == 0) {
PROC_UNLOCK(p);
return (0);
}
/* not being traced any more */
p->p_flag &= ~(P_TRACED | P_STOPPED_TRACE);
/* remove pending SIGTRAP, else the process will die */
sigqueue_delete_proc(p, SIGTRAP);
FOREACH_THREAD_IN_PROC(p, temp)
temp->td_dbgflags &= ~TDB_SUSPEND;
PROC_UNLOCK(p);
/* give process back to original parent */
sx_xlock(&proctree_lock);
if (p->p_oppid != p->p_pptr->p_pid) {
struct proc *pp;
pp = pfind(p->p_oppid);
PROC_LOCK(p);
if (pp) {
PROC_UNLOCK(pp);
proc_reparent(p, pp);
}
} else
PROC_LOCK(p);
p->p_oppid = 0;
p->p_stops = 0;
p->p_flag &= ~P_WAITED; /* XXX ? */
sx_xunlock(&proctree_lock);
wakeup(td->td_proc); /* XXX for CTL_WAIT below ? */
break;
/*
* Step. Let the target process execute a single instruction.
* What does it mean to single step a threaded program?
*/
case PROCFS_CTL_STEP:
error = proc_sstep(FIRST_THREAD_IN_PROC(p));
if (error) {
PROC_UNLOCK(p);
return (error);
}
break;
/*
* Run. Let the target process continue running until a breakpoint
* or some other trap.
*/
case PROCFS_CTL_RUN:
p->p_flag &= ~P_STOPPED_SIG; /* this uses SIGSTOP */
break;
/*
* Wait for the target process to stop.
* If the target is not being traced then just wait
* to enter
*/
case PROCFS_CTL_WAIT:
if (p->p_flag & P_TRACED) {
while (error == 0 &&
(P_SHOULDSTOP(p)) &&
(p->p_flag & P_TRACED) &&
(p->p_pptr == td->td_proc))
error = msleep(p, &p->p_mtx,
PWAIT|PCATCH, "procfsx", 0);
if (error == 0 && !TRACE_WAIT_P(td->td_proc, p))
error = EBUSY;
} else {
while (error == 0 && P_SHOULDSTOP(p))
error = msleep(p, &p->p_mtx,
PWAIT|PCATCH, "procfs", 0);
}
PROC_UNLOCK(p);
return (error);
default:
panic("procfs_control");
}
PROC_SLOCK(p);
thread_unsuspend(p); /* If it can run, let it do so. */
PROC_SUNLOCK(p);
PROC_UNLOCK(p);
return (0);
}
static void
do_fork(struct thread *td, struct fork_req *fr, struct proc *p2, struct thread *td2,
struct vmspace *vm2, struct file *fp_procdesc)
{
struct proc *p1, *pptr;
int trypid;
struct filedesc *fd;
struct filedesc_to_leader *fdtol;
struct sigacts *newsigacts;
sx_assert(&proctree_lock, SX_SLOCKED);
sx_assert(&allproc_lock, SX_XLOCKED);
p1 = td->td_proc;
trypid = fork_findpid(fr->fr_flags);
sx_sunlock(&proctree_lock);
p2->p_state = PRS_NEW; /* protect against others */
p2->p_pid = trypid;
AUDIT_ARG_PID(p2->p_pid);
LIST_INSERT_HEAD(&allproc, p2, p_list);
allproc_gen++;
LIST_INSERT_HEAD(PIDHASH(p2->p_pid), p2, p_hash);
tidhash_add(td2);
PROC_LOCK(p2);
PROC_LOCK(p1);
sx_xunlock(&allproc_lock);
bcopy(&p1->p_startcopy, &p2->p_startcopy,
__rangeof(struct proc, p_startcopy, p_endcopy));
pargs_hold(p2->p_args);
PROC_UNLOCK(p1);
bzero(&p2->p_startzero,
__rangeof(struct proc, p_startzero, p_endzero));
/* Tell the prison that we exist. */
prison_proc_hold(p2->p_ucred->cr_prison);
PROC_UNLOCK(p2);
/*
* Malloc things while we don't hold any locks.
*/
if (fr->fr_flags & RFSIGSHARE)
newsigacts = NULL;
else
newsigacts = sigacts_alloc();
/*
* Copy filedesc.
*/
if (fr->fr_flags & RFCFDG) {
fd = fdinit(p1->p_fd, false);
fdtol = NULL;
} else if (fr->fr_flags & RFFDG) {
fd = fdcopy(p1->p_fd);
fdtol = NULL;
} else {
fd = fdshare(p1->p_fd);
if (p1->p_fdtol == NULL)
p1->p_fdtol = filedesc_to_leader_alloc(NULL, NULL,
p1->p_leader);
if ((fr->fr_flags & RFTHREAD) != 0) {
/*
* Shared file descriptor table, and shared
* process leaders.
*/
fdtol = p1->p_fdtol;
FILEDESC_XLOCK(p1->p_fd);
fdtol->fdl_refcount++;
FILEDESC_XUNLOCK(p1->p_fd);
} else {
/*
* Shared file descriptor table, and different
* process leaders.
*/
fdtol = filedesc_to_leader_alloc(p1->p_fdtol,
p1->p_fd, p2);
}
}
/*
* Make a proc table entry for the new process.
* Start by zeroing the section of proc that is zero-initialized,
* then copy the section that is copied directly from the parent.
*/
PROC_LOCK(p2);
PROC_LOCK(p1);
bzero(&td2->td_startzero,
__rangeof(struct thread, td_startzero, td_endzero));
bcopy(&td->td_startcopy, &td2->td_startcopy,
__rangeof(struct thread, td_startcopy, td_endcopy));
bcopy(&p2->p_comm, &td2->td_name, sizeof(td2->td_name));
td2->td_sigstk = td->td_sigstk;
td2->td_flags = TDF_INMEM;
td2->td_lend_user_pri = PRI_MAX;
#ifdef VIMAGE
td2->td_vnet = NULL;
td2->td_vnet_lpush = NULL;
#endif
/*
* Allow the scheduler to initialize the child.
*/
thread_lock(td);
sched_fork(td, td2);
thread_unlock(td);
/*
* Duplicate sub-structures as needed.
* Increase reference counts on shared objects.
*/
p2->p_flag = P_INMEM;
p2->p_flag2 = p1->p_flag2 & (P2_NOTRACE | P2_NOTRACE_EXEC | P2_TRAPCAP);
p2->p_swtick = ticks;
if (p1->p_flag & P_PROFIL)
startprofclock(p2);
/*
* Whilst the proc lock is held, copy the VM domain data out
* using the VM domain method.
*/
vm_domain_policy_init(&p2->p_vm_dom_policy);
vm_domain_policy_localcopy(&p2->p_vm_dom_policy,
&p1->p_vm_dom_policy);
if (fr->fr_flags & RFSIGSHARE) {
p2->p_sigacts = sigacts_hold(p1->p_sigacts);
} else {
sigacts_copy(newsigacts, p1->p_sigacts);
p2->p_sigacts = newsigacts;
}
if (fr->fr_flags & RFTSIGZMB)
p2->p_sigparent = RFTSIGNUM(fr->fr_flags);
else if (fr->fr_flags & RFLINUXTHPN)
p2->p_sigparent = SIGUSR1;
else
p2->p_sigparent = SIGCHLD;
p2->p_textvp = p1->p_textvp;
p2->p_fd = fd;
p2->p_fdtol = fdtol;
if (p1->p_flag2 & P2_INHERIT_PROTECTED) {
p2->p_flag |= P_PROTECTED;
p2->p_flag2 |= P2_INHERIT_PROTECTED;
}
/*
* p_limit is copy-on-write. Bump its refcount.
*/
lim_fork(p1, p2);
thread_cow_get_proc(td2, p2);
pstats_fork(p1->p_stats, p2->p_stats);
PROC_UNLOCK(p1);
PROC_UNLOCK(p2);
/* Bump references to the text vnode (for procfs). */
if (p2->p_textvp)
vrefact(p2->p_textvp);
/*
* Set up linkage for kernel based threading.
*/
if ((fr->fr_flags & RFTHREAD) != 0) {
mtx_lock(&ppeers_lock);
p2->p_peers = p1->p_peers;
p1->p_peers = p2;
p2->p_leader = p1->p_leader;
mtx_unlock(&ppeers_lock);
PROC_LOCK(p1->p_leader);
if ((p1->p_leader->p_flag & P_WEXIT) != 0) {
PROC_UNLOCK(p1->p_leader);
/*
* The task leader is exiting, so process p1 is
* going to be killed shortly. Since p1 obviously
* isn't dead yet, we know that the leader is either
* sending SIGKILL's to all the processes in this
* task or is sleeping waiting for all the peers to
* exit. We let p1 complete the fork, but we need
* to go ahead and kill the new process p2 since
* the task leader may not get a chance to send
* SIGKILL to it. We leave it on the list so that
* the task leader will wait for this new process
* to commit suicide.
*/
PROC_LOCK(p2);
kern_psignal(p2, SIGKILL);
PROC_UNLOCK(p2);
} else
PROC_UNLOCK(p1->p_leader);
} else {
p2->p_peers = NULL;
p2->p_leader = p2;
}
sx_xlock(&proctree_lock);
PGRP_LOCK(p1->p_pgrp);
PROC_LOCK(p2);
PROC_LOCK(p1);
/*
* Preserve some more flags in subprocess. P_PROFIL has already
* been preserved.
*/
p2->p_flag |= p1->p_flag & P_SUGID;
td2->td_pflags |= (td->td_pflags & TDP_ALTSTACK) | TDP_FORKING;
SESS_LOCK(p1->p_session);
if (p1->p_session->s_ttyvp != NULL && p1->p_flag & P_CONTROLT)
p2->p_flag |= P_CONTROLT;
SESS_UNLOCK(p1->p_session);
if (fr->fr_flags & RFPPWAIT)
p2->p_flag |= P_PPWAIT;
p2->p_pgrp = p1->p_pgrp;
LIST_INSERT_AFTER(p1, p2, p_pglist);
PGRP_UNLOCK(p1->p_pgrp);
LIST_INIT(&p2->p_children);
LIST_INIT(&p2->p_orphans);
callout_init_mtx(&p2->p_itcallout, &p2->p_mtx, 0);
/*
* If PF_FORK is set, the child process inherits the
* procfs ioctl flags from its parent.
*/
if (p1->p_pfsflags & PF_FORK) {
p2->p_stops = p1->p_stops;
p2->p_pfsflags = p1->p_pfsflags;
}
/*
* This begins the section where we must prevent the parent
* from being swapped.
*/
_PHOLD(p1);
PROC_UNLOCK(p1);
/*
* Attach the new process to its parent.
*
* If RFNOWAIT is set, the newly created process becomes a child
* of init. This effectively disassociates the child from the
* parent.
*/
if ((fr->fr_flags & RFNOWAIT) != 0) {
pptr = p1->p_reaper;
p2->p_reaper = pptr;
} else {
p2->p_reaper = (p1->p_treeflag & P_TREE_REAPER) != 0 ?
p1 : p1->p_reaper;
pptr = p1;
}
p2->p_pptr = pptr;
LIST_INSERT_HEAD(&pptr->p_children, p2, p_sibling);
LIST_INIT(&p2->p_reaplist);
LIST_INSERT_HEAD(&p2->p_reaper->p_reaplist, p2, p_reapsibling);
if (p2->p_reaper == p1)
p2->p_reapsubtree = p2->p_pid;
sx_xunlock(&proctree_lock);
/* Inform accounting that we have forked. */
p2->p_acflag = AFORK;
PROC_UNLOCK(p2);
#ifdef KTRACE
ktrprocfork(p1, p2);
#endif
/*
* Finish creating the child process. It will return via a different
* execution path later. (ie: directly into user mode)
*/
vm_forkproc(td, p2, td2, vm2, fr->fr_flags);
if (fr->fr_flags == (RFFDG | RFPROC)) {
VM_CNT_INC(v_forks);
VM_CNT_ADD(v_forkpages, p2->p_vmspace->vm_dsize +
p2->p_vmspace->vm_ssize);
} else if (fr->fr_flags == (RFFDG | RFPROC | RFPPWAIT | RFMEM)) {
VM_CNT_INC(v_vforks);
VM_CNT_ADD(v_vforkpages, p2->p_vmspace->vm_dsize +
p2->p_vmspace->vm_ssize);
} else if (p1 == &proc0) {
VM_CNT_INC(v_kthreads);
VM_CNT_ADD(v_kthreadpages, p2->p_vmspace->vm_dsize +
p2->p_vmspace->vm_ssize);
} else {
VM_CNT_INC(v_rforks);
VM_CNT_ADD(v_rforkpages, p2->p_vmspace->vm_dsize +
p2->p_vmspace->vm_ssize);
}
/*
* Associate the process descriptor with the process before anything
* can happen that might cause that process to need the descriptor.
* However, don't do this until after fork(2) can no longer fail.
*/
if (fr->fr_flags & RFPROCDESC)
procdesc_new(p2, fr->fr_pd_flags);
/*
* Both processes are set up, now check if any loadable modules want
* to adjust anything.
*/
EVENTHANDLER_INVOKE(process_fork, p1, p2, fr->fr_flags);
/*
* Set the child start time and mark the process as being complete.
*/
PROC_LOCK(p2);
PROC_LOCK(p1);
microuptime(&p2->p_stats->p_start);
PROC_SLOCK(p2);
p2->p_state = PRS_NORMAL;
PROC_SUNLOCK(p2);
#ifdef KDTRACE_HOOKS
/*
* Tell the DTrace fasttrap provider about the new process so that any
* tracepoints inherited from the parent can be removed. We have to do
* this only after p_state is PRS_NORMAL since the fasttrap module will
* use pfind() later on.
*/
if ((fr->fr_flags & RFMEM) == 0 && dtrace_fasttrap_fork)
dtrace_fasttrap_fork(p1, p2);
#endif
/*
* Hold the process so that it cannot exit after we make it runnable,
* but before we wait for the debugger.
*/
_PHOLD(p2);
if (p1->p_ptevents & PTRACE_FORK) {
/*
* Arrange for debugger to receive the fork event.
*
* We can report PL_FLAG_FORKED regardless of
* P_FOLLOWFORK settings, but it does not make a sense
* for runaway child.
*/
td->td_dbgflags |= TDB_FORK;
td->td_dbg_forked = p2->p_pid;
td2->td_dbgflags |= TDB_STOPATFORK;
}
if (fr->fr_flags & RFPPWAIT) {
td->td_pflags |= TDP_RFPPWAIT;
td->td_rfppwait_p = p2;
td->td_dbgflags |= TDB_VFORK;
}
PROC_UNLOCK(p2);
/*
* Now can be swapped.
*/
_PRELE(p1);
PROC_UNLOCK(p1);
/*
* Tell any interested parties about the new process.
*/
knote_fork(p1->p_klist, p2->p_pid);
SDT_PROBE3(proc, , , create, p2, p1, fr->fr_flags);
if (fr->fr_flags & RFPROCDESC) {
procdesc_finit(p2->p_procdesc, fp_procdesc);
fdrop(fp_procdesc, td);
}
if ((fr->fr_flags & RFSTOPPED) == 0) {
/*
* If RFSTOPPED not requested, make child runnable and
* add to run queue.
*/
thread_lock(td2);
TD_SET_CAN_RUN(td2);
sched_add(td2, SRQ_BORING);
thread_unlock(td2);
if (fr->fr_pidp != NULL)
*fr->fr_pidp = p2->p_pid;
} else {
*fr->fr_procp = p2;
}
PROC_LOCK(p2);
/*
* Wait until debugger is attached to child.
*/
while (td2->td_proc == p2 && (td2->td_dbgflags & TDB_STOPATFORK) != 0)
cv_wait(&p2->p_dbgwait, &p2->p_mtx);
_PRELE(p2);
racct_proc_fork_done(p2);
PROC_UNLOCK(p2);
}
/*
* Process ioctls
*/
int
procfs_ioctl(PFS_IOCTL_ARGS)
{
struct procfs_status *ps;
#ifdef COMPAT_FREEBSD32
struct procfs_status32 *ps32;
#endif
int error, flags, sig;
#ifdef COMPAT_FREEBSD6
int ival;
#endif
KASSERT(p != NULL,
("%s() called without a process", __func__));
PROC_LOCK_ASSERT(p, MA_OWNED);
error = 0;
switch (cmd) {
#if defined(COMPAT_FREEBSD5) || defined(COMPAT_FREEBSD4) || defined(COMPAT_43)
case _IOC(IOC_IN, 'p', 1, 0):
#endif
#ifdef COMPAT_FREEBSD6
case _IO('p', 1):
ival = IOCPARM_IVAL(data);
data = &ival;
#endif
case PIOCBIS:
p->p_stops |= *(unsigned int *)data;
break;
#if defined(COMPAT_FREEBSD5) || defined(COMPAT_FREEBSD4) || defined(COMPAT_43)
case _IOC(IOC_IN, 'p', 2, 0):
#endif
#ifdef COMPAT_FREEBSD6
case _IO('p', 2):
ival = IOCPARM_IVAL(data);
data = &ival;
#endif
case PIOCBIC:
p->p_stops &= ~*(unsigned int *)data;
break;
#if defined(COMPAT_FREEBSD5) || defined(COMPAT_FREEBSD4) || defined(COMPAT_43)
case _IOC(IOC_IN, 'p', 3, 0):
#endif
#ifdef COMPAT_FREEBSD6
case _IO('p', 3):
ival = IOCPARM_IVAL(data);
data = &ival;
#endif
case PIOCSFL:
flags = *(unsigned int *)data;
if (flags & PF_ISUGID) {
/*
* XXXRW: Is this specific check required here, as
* p_candebug() should implement it, or other checks
* are missing.
*/
error = priv_check(td, PRIV_DEBUG_SUGID);
if (error)
break;
}
p->p_pfsflags = flags;
break;
case PIOCGFL:
*(unsigned int *)data = p->p_pfsflags;
break;
case PIOCWAIT:
while (p->p_step == 0 && (p->p_flag & P_WEXIT) == 0) {
/* sleep until p stops */
_PHOLD(p);
error = msleep(&p->p_stype, &p->p_mtx,
PWAIT|PCATCH, "pioctl", 0);
_PRELE(p);
if (error != 0)
break;
}
/* fall through to PIOCSTATUS */
case PIOCSTATUS:
ps = (struct procfs_status *)data;
ps->state = (p->p_step == 0);
ps->flags = 0; /* nope */
ps->events = p->p_stops;
ps->why = p->p_step ? p->p_stype : 0;
ps->val = p->p_step ? p->p_xstat : 0;
break;
#ifdef COMPAT_FREEBSD32
case PIOCWAIT32:
while (p->p_step == 0 && (p->p_flag & P_WEXIT) == 0) {
/* sleep until p stops */
_PHOLD(p);
error = msleep(&p->p_stype, &p->p_mtx,
PWAIT|PCATCH, "pioctl", 0);
_PRELE(p);
if (error != 0)
break;
}
/* fall through to PIOCSTATUS32 */
case PIOCSTATUS32:
ps32 = (struct procfs_status32 *)data;
ps32->state = (p->p_step == 0);
ps32->flags = 0; /* nope */
ps32->events = p->p_stops;
ps32->why = p->p_step ? p->p_stype : 0;
ps32->val = p->p_step ? p->p_xstat : 0;
break;
#endif
#if defined(COMPAT_FREEBSD5) || defined(COMPAT_FREEBSD4) || defined(COMPAT_43)
case _IOC(IOC_IN, 'p', 5, 0):
#endif
#ifdef COMPAT_FREEBSD6
case _IO('p', 5):
ival = IOCPARM_IVAL(data);
data = &ival;
#endif
case PIOCCONT:
if (p->p_step == 0)
break;
sig = *(unsigned int *)data;
if (sig != 0 && !_SIG_VALID(sig)) {
error = EINVAL;
break;
}
#if 0
p->p_step = 0;
if (P_SHOULDSTOP(p)) {
p->p_xstat = sig;
p->p_flag &= ~(P_STOPPED_TRACE|P_STOPPED_SIG);
PROC_SLOCK(p);
thread_unsuspend(p);
PROC_SUNLOCK(p);
} else if (sig)
kern_psignal(p, sig);
#else
if (sig)
kern_psignal(p, sig);
p->p_step = 0;
wakeup(&p->p_step);
#endif
break;
default:
error = (ENOTTY);
}
return (error);
}
static void
soaio_process_job(struct socket *so, struct sockbuf *sb, struct kaiocb *job)
{
struct ucred *td_savedcred;
struct thread *td;
struct file *fp;
struct uio uio;
struct iovec iov;
size_t cnt, done;
long ru_before;
int error, flags;
SOCKBUF_UNLOCK(sb);
aio_switch_vmspace(job);
td = curthread;
fp = job->fd_file;
retry:
td_savedcred = td->td_ucred;
td->td_ucred = job->cred;
done = job->aio_done;
cnt = job->uaiocb.aio_nbytes - done;
iov.iov_base = (void *)((uintptr_t)job->uaiocb.aio_buf + done);
iov.iov_len = cnt;
uio.uio_iov = &iov;
uio.uio_iovcnt = 1;
uio.uio_offset = 0;
uio.uio_resid = cnt;
uio.uio_segflg = UIO_USERSPACE;
uio.uio_td = td;
flags = MSG_NBIO;
/*
* For resource usage accounting, only count a completed request
* as a single message to avoid counting multiple calls to
* sosend/soreceive on a blocking socket.
*/
if (sb == &so->so_rcv) {
uio.uio_rw = UIO_READ;
ru_before = td->td_ru.ru_msgrcv;
#ifdef MAC
error = mac_socket_check_receive(fp->f_cred, so);
if (error == 0)
#endif
error = soreceive(so, NULL, &uio, NULL, NULL, &flags);
if (td->td_ru.ru_msgrcv != ru_before)
job->msgrcv = 1;
} else {
uio.uio_rw = UIO_WRITE;
ru_before = td->td_ru.ru_msgsnd;
#ifdef MAC
error = mac_socket_check_send(fp->f_cred, so);
if (error == 0)
#endif
error = sosend(so, NULL, &uio, NULL, NULL, flags, td);
if (td->td_ru.ru_msgsnd != ru_before)
job->msgsnd = 1;
if (error == EPIPE && (so->so_options & SO_NOSIGPIPE) == 0) {
PROC_LOCK(job->userproc);
kern_psignal(job->userproc, SIGPIPE);
PROC_UNLOCK(job->userproc);
}
}
done += cnt - uio.uio_resid;
job->aio_done = done;
td->td_ucred = td_savedcred;
if (error == EWOULDBLOCK) {
/*
* The request was either partially completed or not
* completed at all due to racing with a read() or
* write() on the socket. If the socket is
* non-blocking, return with any partial completion.
* If the socket is blocking or if no progress has
* been made, requeue this request at the head of the
* queue to try again when the socket is ready.
*/
MPASS(done != job->uaiocb.aio_nbytes);
SOCKBUF_LOCK(sb);
if (done == 0 || !(so->so_state & SS_NBIO)) {
empty_results++;
if (soaio_ready(so, sb)) {
empty_retries++;
SOCKBUF_UNLOCK(sb);
goto retry;
}
if (!aio_set_cancel_function(job, soo_aio_cancel)) {
SOCKBUF_UNLOCK(sb);
if (done != 0)
aio_complete(job, done, 0);
else
aio_cancel(job);
SOCKBUF_LOCK(sb);
} else {
TAILQ_INSERT_HEAD(&sb->sb_aiojobq, job, list);
}
return;
}
SOCKBUF_UNLOCK(sb);
}
if (done != 0 && (error == ERESTART || error == EINTR ||
error == EWOULDBLOCK))
error = 0;
if (error)
aio_complete(job, -1, error);
else
aio_complete(job, done, 0);
SOCKBUF_LOCK(sb);
}
