static int
linux_clone_proc(struct thread *td, struct linux_clone_args *args)
{
struct fork_req fr;
int error, ff = RFPROC | RFSTOPPED;
struct proc *p2;
struct thread *td2;
int exit_signal;
struct linux_emuldata *em;
#ifdef DEBUG
if (ldebug(clone)) {
printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, "
"child tid: %p"), (unsigned)args->flags,
args->stack, args->parent_tidptr, args->child_tidptr);
}
#endif
exit_signal = args->flags & 0x000000ff;
if (LINUX_SIG_VALID(exit_signal)) {
exit_signal = linux_to_bsd_signal(exit_signal);
} else if (exit_signal != 0)
return (EINVAL);
if (args->flags & LINUX_CLONE_VM)
ff |= RFMEM;
if (args->flags & LINUX_CLONE_SIGHAND)
ff |= RFSIGSHARE;
/*
* XXX: In Linux, sharing of fs info (chroot/cwd/umask)
* and open files is independent. In FreeBSD, its in one
* structure but in reality it does not cause any problems
* because both of these flags are usually set together.
*/
if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS)))
ff |= RFFDG;
if (args->flags & LINUX_CLONE_PARENT_SETTID)
if (args->parent_tidptr == NULL)
return (EINVAL);
if (args->flags & LINUX_CLONE_VFORK)
ff |= RFPPWAIT;
bzero(&fr, sizeof(fr));
fr.fr_flags = ff;
fr.fr_procp = &p2;
error = fork1(td, &fr);
if (error)
return (error);
td2 = FIRST_THREAD_IN_PROC(p2);
/* create the emuldata */
linux_proc_init(td, td2, args->flags);
em = em_find(td2);
KASSERT(em != NULL, ("clone_proc: emuldata not found.\n"));
if (args->flags & LINUX_CLONE_CHILD_SETTID)
em->child_set_tid = args->child_tidptr;
else
em->child_set_tid = NULL;
if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
em->child_clear_tid = args->child_tidptr;
else
em->child_clear_tid = NULL;
if (args->flags & LINUX_CLONE_PARENT_SETTID) {
error = copyout(&p2->p_pid, args->parent_tidptr,
sizeof(p2->p_pid));
if (error)
printf(LMSG("copyout failed!"));
}
PROC_LOCK(p2);
p2->p_sigparent = exit_signal;
PROC_UNLOCK(p2);
/*
* In a case of stack = NULL, we are supposed to COW calling process
* stack. This is what normal fork() does, so we just keep tf_rsp arg
* intact.
*/
linux_set_upcall_kse(td2, PTROUT(args->stack));
if (args->flags & LINUX_CLONE_SETTLS)
linux_set_cloned_tls(td2, args->tls);
/*
* If CLONE_PARENT is set, then the parent of the new process will be
* the same as that of the calling process.
*/
if (args->flags & LINUX_CLONE_PARENT) {
sx_xlock(&proctree_lock);
PROC_LOCK(p2);
proc_reparent(p2, td->td_proc->p_pptr);
PROC_UNLOCK(p2);
sx_xunlock(&proctree_lock);
}
#ifdef DEBUG
if (ldebug(clone))
printf(LMSG("clone: successful rfork to %d, "
"stack %p sig = %d"), (int)p2->p_pid, args->stack,
exit_signal);
#endif
/*
* Make this runnable after we are finished with it.
*/
thread_lock(td2);
TD_SET_CAN_RUN(td2);
sched_add(td2, SRQ_BORING);
thread_unlock(td2);
td->td_retval[0] = p2->p_pid;
return (0);
}
static int
linux_clone_thread(struct thread *td, struct linux_clone_args *args)
{
struct linux_emuldata *em;
struct thread *newtd;
struct proc *p;
int error;
#ifdef DEBUG
if (ldebug(clone)) {
printf(ARGS(clone, "thread: flags %x, stack %p, parent tid: %p, "
"child tid: %p"), (unsigned)args->flags,
args->stack, args->parent_tidptr, args->child_tidptr);
}
#endif
LINUX_CTR4(clone_thread, "thread(%d) flags %x ptid %p ctid %p",
td->td_tid, (unsigned)args->flags,
args->parent_tidptr, args->child_tidptr);
if (args->flags & LINUX_CLONE_PARENT_SETTID)
if (args->parent_tidptr == NULL)
return (EINVAL);
/* Threads should be created with own stack */
if (args->stack == NULL)
return (EINVAL);
p = td->td_proc;
#ifdef RACCT
if (racct_enable) {
PROC_LOCK(p);
error = racct_add(p, RACCT_NTHR, 1);
PROC_UNLOCK(p);
if (error != 0)
return (EPROCLIM);
}
#endif
/* Initialize our td */
error = kern_thr_alloc(p, 0, &newtd);
if (error)
goto fail;
cpu_copy_thread(newtd, td);
bzero(&newtd->td_startzero,
__rangeof(struct thread, td_startzero, td_endzero));
bcopy(&td->td_startcopy, &newtd->td_startcopy,
__rangeof(struct thread, td_startcopy, td_endcopy));
newtd->td_proc = p;
thread_cow_get(newtd, td);
/* create the emuldata */
linux_proc_init(td, newtd, args->flags);
em = em_find(newtd);
KASSERT(em != NULL, ("clone_thread: emuldata not found.\n"));
if (args->flags & LINUX_CLONE_SETTLS)
linux_set_cloned_tls(newtd, args->tls);
if (args->flags & LINUX_CLONE_CHILD_SETTID)
em->child_set_tid = args->child_tidptr;
else
em->child_set_tid = NULL;
if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
em->child_clear_tid = args->child_tidptr;
else
em->child_clear_tid = NULL;
cpu_thread_clean(newtd);
linux_set_upcall_kse(newtd, PTROUT(args->stack));
PROC_LOCK(p);
p->p_flag |= P_HADTHREADS;
bcopy(p->p_comm, newtd->td_name, sizeof(newtd->td_name));
if (args->flags & LINUX_CLONE_PARENT)
thread_link(newtd, p->p_pptr);
else
thread_link(newtd, p);
thread_lock(td);
/* let the scheduler know about these things. */
sched_fork_thread(td, newtd);
thread_unlock(td);
if (P_SHOULDSTOP(p))
newtd->td_flags |= TDF_ASTPENDING | TDF_NEEDSUSPCHK;
PROC_UNLOCK(p);
tidhash_add(newtd);
#ifdef DEBUG
if (ldebug(clone))
printf(ARGS(clone, "successful clone to %d, stack %p"),
(int)newtd->td_tid, args->stack);
#endif
LINUX_CTR2(clone_thread, "thread(%d) successful clone to %d",
td->td_tid, newtd->td_tid);
if (args->flags & LINUX_CLONE_PARENT_SETTID) {
error = copyout(&newtd->td_tid, args->parent_tidptr,
sizeof(newtd->td_tid));
if (error)
printf(LMSG("clone_thread: copyout failed!"));
}
/*
* Make this runnable after we are finished with it.
*/
thread_lock(newtd);
TD_SET_CAN_RUN(newtd);
sched_add(newtd, SRQ_BORING);
thread_unlock(newtd);
td->td_retval[0] = newtd->td_tid;
return (0);
fail:
#ifdef RACCT
if (racct_enable) {
PROC_LOCK(p);
racct_sub(p, RACCT_NTHR, 1);
PROC_UNLOCK(p);
}
#endif
return (error);
}
static int
linux_common_open(struct thread *td, int dirfd, char *path, int l_flags, int mode)
{
cap_rights_t rights;
struct proc *p = td->td_proc;
struct file *fp;
int fd;
int bsd_flags, error;
bsd_flags = 0;
switch (l_flags & LINUX_O_ACCMODE) {
case LINUX_O_WRONLY:
bsd_flags |= O_WRONLY;
break;
case LINUX_O_RDWR:
bsd_flags |= O_RDWR;
break;
default:
bsd_flags |= O_RDONLY;
}
if (l_flags & LINUX_O_NDELAY)
bsd_flags |= O_NONBLOCK;
if (l_flags & LINUX_O_APPEND)
bsd_flags |= O_APPEND;
if (l_flags & LINUX_O_SYNC)
bsd_flags |= O_FSYNC;
if (l_flags & LINUX_O_NONBLOCK)
bsd_flags |= O_NONBLOCK;
if (l_flags & LINUX_FASYNC)
bsd_flags |= O_ASYNC;
if (l_flags & LINUX_O_CREAT)
bsd_flags |= O_CREAT;
if (l_flags & LINUX_O_TRUNC)
bsd_flags |= O_TRUNC;
if (l_flags & LINUX_O_EXCL)
bsd_flags |= O_EXCL;
if (l_flags & LINUX_O_NOCTTY)
bsd_flags |= O_NOCTTY;
if (l_flags & LINUX_O_DIRECT)
bsd_flags |= O_DIRECT;
if (l_flags & LINUX_O_NOFOLLOW)
bsd_flags |= O_NOFOLLOW;
if (l_flags & LINUX_O_DIRECTORY)
bsd_flags |= O_DIRECTORY;
/* XXX LINUX_O_NOATIME: unable to be easily implemented. */
error = kern_openat(td, dirfd, path, UIO_SYSSPACE, bsd_flags, mode);
if (error != 0)
goto done;
if (bsd_flags & O_NOCTTY)
goto done;
/*
* XXX In between kern_open() and fget(), another process
* having the same filedesc could use that fd without
* checking below.
*/
fd = td->td_retval[0];
if (fget(td, fd, cap_rights_init(&rights, CAP_IOCTL), &fp) == 0) {
if (fp->f_type != DTYPE_VNODE) {
fdrop(fp, td);
goto done;
}
sx_slock(&proctree_lock);
PROC_LOCK(p);
if (SESS_LEADER(p) && !(p->p_flag & P_CONTROLT)) {
PROC_UNLOCK(p);
sx_sunlock(&proctree_lock);
/* XXXPJD: Verify if TIOCSCTTY is allowed. */
(void) fo_ioctl(fp, TIOCSCTTY, (caddr_t) 0,
td->td_ucred, td);
} else {
PROC_UNLOCK(p);
sx_sunlock(&proctree_lock);
}
fdrop(fp, td);
}
done:
#ifdef DEBUG
if (ldebug(open))
printf(LMSG("open returns error %d"), error);
#endif
LFREEPATH(path);
return (error);
}
/*
* System call to cleanup state after a signal
* has been taken. Reset signal mask and
* stack state from context left by rt_sendsig (above).
* Return to previous pc and psl as specified by
* context left by sendsig. Check carefully to
* make sure that the user has not modified the
* psl to gain improper privileges or to cause
* a machine fault.
*/
int
linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args)
{
struct l_ucontext uc;
struct l_sigcontext *context;
sigset_t bmask;
l_stack_t *lss;
stack_t ss;
struct trapframe *regs;
int eflags;
ksiginfo_t ksi;
regs = td->td_frame;
#ifdef DEBUG
if (ldebug(rt_sigreturn))
printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp);
#endif
/*
* The trampoline code hands us the ucontext.
* It is unsafe to keep track of it ourselves, in the event that a
* program jumps out of a signal handler.
*/
if (copyin(args->ucp, &uc, sizeof(uc)) != 0)
return (EFAULT);
context = &uc.uc_mcontext;
/*
* Check for security violations.
*/
#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
eflags = context->sc_eflags;
if (!EFLAGS_SECURE(eflags, regs->tf_rflags))
return(EINVAL);
/*
* Don't allow users to load a valid privileged %cs. Let the
* hardware check for invalid selectors, excess privilege in
* other selectors, invalid %eip's and invalid %esp's.
*/
#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
if (!CS_SECURE(context->sc_cs)) {
ksiginfo_init_trap(&ksi);
ksi.ksi_signo = SIGBUS;
ksi.ksi_code = BUS_OBJERR;
ksi.ksi_trapno = T_PROTFLT;
ksi.ksi_addr = (void *)regs->tf_rip;
trapsignal(td, &ksi);
return(EINVAL);
}
linux_to_bsd_sigset(&uc.uc_sigmask, &bmask);
kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);
/*
* Restore signal context
*/
regs->tf_gs = context->sc_gs;
regs->tf_fs = context->sc_fs;
regs->tf_es = context->sc_es;
regs->tf_ds = context->sc_ds;
regs->tf_rdi = context->sc_edi;
regs->tf_rsi = context->sc_esi;
regs->tf_rbp = context->sc_ebp;
regs->tf_rbx = context->sc_ebx;
regs->tf_rdx = context->sc_edx;
regs->tf_rcx = context->sc_ecx;
regs->tf_rax = context->sc_eax;
regs->tf_rip = context->sc_eip;
regs->tf_cs = context->sc_cs;
regs->tf_rflags = eflags;
regs->tf_rsp = context->sc_esp_at_signal;
regs->tf_ss = context->sc_ss;
set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
/*
* call sigaltstack & ignore results..
*/
lss = &uc.uc_stack;
ss.ss_sp = PTRIN(lss->ss_sp);
ss.ss_size = lss->ss_size;
ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags);
#ifdef DEBUG
if (ldebug(rt_sigreturn))
printf(LMSG("rt_sigret flags: 0x%x, sp: %p, ss: 0x%lx, mask: 0x%x"),
ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask);
#endif
(void)kern_sigaltstack(td, &ss, NULL);
return (EJUSTRETURN);
}
static void
linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
{
struct thread *td = curthread;
struct proc *p = td->td_proc;
struct sigacts *psp;
struct trapframe *regs;
struct l_rt_sigframe *fp, frame;
int oonstack;
int sig;
int code;
sig = ksi->ksi_signo;
code = ksi->ksi_code;
PROC_LOCK_ASSERT(p, MA_OWNED);
psp = p->p_sigacts;
mtx_assert(&psp->ps_mtx, MA_OWNED);
regs = td->td_frame;
oonstack = sigonstack(regs->tf_rsp);
#ifdef DEBUG
if (ldebug(rt_sendsig))
printf(ARGS(rt_sendsig, "%p, %d, %p, %u"),
catcher, sig, (void*)mask, code);
#endif
/*
* Allocate space for the signal handler context.
*/
if ((td->td_pflags & TDP_ALTSTACK) && !oonstack &&
SIGISMEMBER(psp->ps_sigonstack, sig)) {
fp = (struct l_rt_sigframe *)(td->td_sigstk.ss_sp +
td->td_sigstk.ss_size - sizeof(struct l_rt_sigframe));
} else
fp = (struct l_rt_sigframe *)regs->tf_rsp - 1;
mtx_unlock(&psp->ps_mtx);
/*
* Build the argument list for the signal handler.
*/
if (p->p_sysent->sv_sigtbl)
if (sig <= p->p_sysent->sv_sigsize)
sig = p->p_sysent->sv_sigtbl[_SIG_IDX(sig)];
bzero(&frame, sizeof(frame));
frame.sf_handler = PTROUT(catcher);
frame.sf_sig = sig;
frame.sf_siginfo = PTROUT(&fp->sf_si);
frame.sf_ucontext = PTROUT(&fp->sf_sc);
/* Fill in POSIX parts */
ksiginfo_to_lsiginfo(ksi, &frame.sf_si, sig);
/*
* Build the signal context to be used by sigreturn.
*/
frame.sf_sc.uc_flags = 0; /* XXX ??? */
frame.sf_sc.uc_link = 0; /* XXX ??? */
frame.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp);
frame.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size;
frame.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK)
? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE;
PROC_UNLOCK(p);
bsd_to_linux_sigset(mask, &frame.sf_sc.uc_sigmask);
frame.sf_sc.uc_mcontext.sc_mask = frame.sf_sc.uc_sigmask.__bits[0];
frame.sf_sc.uc_mcontext.sc_edi = regs->tf_rdi;
frame.sf_sc.uc_mcontext.sc_esi = regs->tf_rsi;
frame.sf_sc.uc_mcontext.sc_ebp = regs->tf_rbp;
frame.sf_sc.uc_mcontext.sc_ebx = regs->tf_rbx;
frame.sf_sc.uc_mcontext.sc_edx = regs->tf_rdx;
frame.sf_sc.uc_mcontext.sc_ecx = regs->tf_rcx;
frame.sf_sc.uc_mcontext.sc_eax = regs->tf_rax;
frame.sf_sc.uc_mcontext.sc_eip = regs->tf_rip;
frame.sf_sc.uc_mcontext.sc_cs = regs->tf_cs;
frame.sf_sc.uc_mcontext.sc_gs = regs->tf_gs;
frame.sf_sc.uc_mcontext.sc_fs = regs->tf_fs;
frame.sf_sc.uc_mcontext.sc_es = regs->tf_es;
frame.sf_sc.uc_mcontext.sc_ds = regs->tf_ds;
frame.sf_sc.uc_mcontext.sc_eflags = regs->tf_rflags;
frame.sf_sc.uc_mcontext.sc_esp_at_signal = regs->tf_rsp;
frame.sf_sc.uc_mcontext.sc_ss = regs->tf_ss;
frame.sf_sc.uc_mcontext.sc_err = regs->tf_err;
frame.sf_sc.uc_mcontext.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr;
frame.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code);
#ifdef DEBUG
if (ldebug(rt_sendsig))
printf(LMSG("rt_sendsig flags: 0x%x, sp: %p, ss: 0x%lx, mask: 0x%x"),
frame.sf_sc.uc_stack.ss_flags, td->td_sigstk.ss_sp,
td->td_sigstk.ss_size, frame.sf_sc.uc_mcontext.sc_mask);
#endif
if (copyout(&frame, fp, sizeof(frame)) != 0) {
/*
* Process has trashed its stack; give it an illegal
* instruction to halt it in its tracks.
*/
#ifdef DEBUG
if (ldebug(rt_sendsig))
printf(LMSG("rt_sendsig: bad stack %p, oonstack=%x"),
fp, oonstack);
#endif
PROC_LOCK(p);
sigexit(td, SIGILL);
}
/*
* Build context to run handler in.
*/
regs->tf_rsp = PTROUT(fp);
regs->tf_rip = p->p_sysent->sv_sigcode_base + linux_sznonrtsigcode;
regs->tf_rflags &= ~(PSL_T | PSL_D);
regs->tf_cs = _ucode32sel;
regs->tf_ss = _udatasel;
regs->tf_ds = _udatasel;
regs->tf_es = _udatasel;
regs->tf_fs = _ufssel;
regs->tf_gs = _ugssel;
regs->tf_flags = TF_HASSEGS;
set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
PROC_LOCK(p);
mtx_lock(&psp->ps_mtx);
}
int
linux_clone(struct thread *td, struct linux_clone_args *args)
{
int error, ff = RFPROC | RFSTOPPED;
struct proc *p2;
struct thread *td2;
int exit_signal;
struct linux_emuldata *em;
#ifdef DEBUG
if (ldebug(clone)) {
printf(ARGS(clone, "flags %x, stack %p, parent tid: %p, "
"child tid: %p"), (unsigned)args->flags,
args->stack, args->parent_tidptr, args->child_tidptr);
}
#endif
exit_signal = args->flags & 0x000000ff;
if (LINUX_SIG_VALID(exit_signal)) {
if (exit_signal <= LINUX_SIGTBLSZ)
exit_signal =
linux_to_bsd_signal[_SIG_IDX(exit_signal)];
} else if (exit_signal != 0)
return (EINVAL);
if (args->flags & LINUX_CLONE_VM)
ff |= RFMEM;
if (args->flags & LINUX_CLONE_SIGHAND)
ff |= RFSIGSHARE;
/*
* XXX: In Linux, sharing of fs info (chroot/cwd/umask)
* and open files is independant. In FreeBSD, its in one
* structure but in reality it does not cause any problems
* because both of these flags are usually set together.
*/
if (!(args->flags & (LINUX_CLONE_FILES | LINUX_CLONE_FS)))
ff |= RFFDG;
/*
* Attempt to detect when linux_clone(2) is used for creating
* kernel threads. Unfortunately despite the existence of the
* CLONE_THREAD flag, version of linuxthreads package used in
* most popular distros as of beginning of 2005 doesn't make
* any use of it. Therefore, this detection relies on
* empirical observation that linuxthreads sets certain
* combination of flags, so that we can make more or less
* precise detection and notify the FreeBSD kernel that several
* processes are in fact part of the same threading group, so
* that special treatment is necessary for signal delivery
* between those processes and fd locking.
*/
if ((args->flags & 0xffffff00) == LINUX_THREADING_FLAGS)
ff |= RFTHREAD;
if (args->flags & LINUX_CLONE_PARENT_SETTID)
if (args->parent_tidptr == NULL)
return (EINVAL);
error = fork1(td, ff, 0, &p2, NULL, 0);
if (error)
return (error);
if (args->flags & (LINUX_CLONE_PARENT | LINUX_CLONE_THREAD)) {
sx_xlock(&proctree_lock);
PROC_LOCK(p2);
proc_reparent(p2, td->td_proc->p_pptr);
PROC_UNLOCK(p2);
sx_xunlock(&proctree_lock);
}
/* create the emuldata */
error = linux_proc_init(td, p2->p_pid, args->flags);
/* reference it - no need to check this */
em = em_find(p2, EMUL_DOLOCK);
KASSERT(em != NULL, ("clone: emuldata not found."));
/* and adjust it */
if (args->flags & LINUX_CLONE_THREAD) {
#ifdef notyet
PROC_LOCK(p2);
p2->p_pgrp = td->td_proc->p_pgrp;
PROC_UNLOCK(p2);
#endif
exit_signal = 0;
}
if (args->flags & LINUX_CLONE_CHILD_SETTID)
em->child_set_tid = args->child_tidptr;
else
em->child_set_tid = NULL;
if (args->flags & LINUX_CLONE_CHILD_CLEARTID)
em->child_clear_tid = args->child_tidptr;
else
em->child_clear_tid = NULL;
EMUL_UNLOCK(&emul_lock);
if (args->flags & LINUX_CLONE_PARENT_SETTID) {
error = copyout(&p2->p_pid, args->parent_tidptr,
sizeof(p2->p_pid));
if (error)
printf(LMSG("copyout failed!"));
}
PROC_LOCK(p2);
p2->p_sigparent = exit_signal;
PROC_UNLOCK(p2);
td2 = FIRST_THREAD_IN_PROC(p2);
/*
* In a case of stack = NULL, we are supposed to COW calling process
* stack. This is what normal fork() does, so we just keep tf_rsp arg
* intact.
*/
if (args->stack)
linux_set_upcall_kse(td2, PTROUT(args->stack));
if (args->flags & LINUX_CLONE_SETTLS)
linux_set_cloned_tls(td2, args->tls);
#ifdef DEBUG
if (ldebug(clone))
printf(LMSG("clone: successful rfork to %d, "
"stack %p sig = %d"), (int)p2->p_pid, args->stack,
exit_signal);
#endif
if (args->flags & LINUX_CLONE_VFORK) {
PROC_LOCK(p2);
p2->p_flag |= P_PPWAIT;
PROC_UNLOCK(p2);
}
/*
* Make this runnable after we are finished with it.
*/
thread_lock(td2);
TD_SET_CAN_RUN(td2);
sched_add(td2, SRQ_BORING);
thread_unlock(td2);
td->td_retval[0] = p2->p_pid;
td->td_retval[1] = 0;
if (args->flags & LINUX_CLONE_VFORK) {
/* wait for the children to exit, ie. emulate vfork */
PROC_LOCK(p2);
while (p2->p_flag & P_PPWAIT)
cv_wait(&p2->p_pwait, &p2->p_mtx);
PROC_UNLOCK(p2);
}
return (0);
}
/*
* System call to cleanup state after a signal
* has been taken. Reset signal mask and
* stack state from context left by rt_sendsig (above).
* Return to previous pc and psl as specified by
* context left by sendsig. Check carefully to
* make sure that the user has not modified the
* psl to gain improper privileges or to cause
* a machine fault.
*/
int
linux_rt_sigreturn(struct thread *td, struct linux_rt_sigreturn_args *args)
{
struct l_ucontext uc;
struct l_sigcontext *context;
sigset_t bmask;
l_stack_t *lss;
stack_t ss;
struct trapframe *regs;
int eflags;
ksiginfo_t ksi;
regs = td->td_frame;
#ifdef DEBUG
if (ldebug(rt_sigreturn))
printf(ARGS(rt_sigreturn, "%p"), (void *)args->ucp);
#endif
/*
* The trampoline code hands us the ucontext.
* It is unsafe to keep track of it ourselves, in the event that a
* program jumps out of a signal handler.
*/
if (copyin(args->ucp, &uc, sizeof(uc)) != 0)
return (EFAULT);
context = &uc.uc_mcontext;
/*
* Check for security violations.
*/
#define EFLAGS_SECURE(ef, oef) ((((ef) ^ (oef)) & ~PSL_USERCHANGE) == 0)
eflags = context->sc_eflags;
/*
* XXX do allow users to change the privileged flag PSL_RF. The
* cpu sets PSL_RF in tf_eflags for faults. Debuggers should
* sometimes set it there too. tf_eflags is kept in the signal
* context during signal handling and there is no other place
* to remember it, so the PSL_RF bit may be corrupted by the
* signal handler without us knowing. Corruption of the PSL_RF
* bit at worst causes one more or one less debugger trap, so
* allowing it is fairly harmless.
*/
if (!EFLAGS_SECURE(eflags & ~PSL_RF, regs->tf_eflags & ~PSL_RF))
return(EINVAL);
/*
* Don't allow users to load a valid privileged %cs. Let the
* hardware check for invalid selectors, excess privilege in
* other selectors, invalid %eip's and invalid %esp's.
*/
#define CS_SECURE(cs) (ISPL(cs) == SEL_UPL)
if (!CS_SECURE(context->sc_cs)) {
ksiginfo_init_trap(&ksi);
ksi.ksi_signo = SIGBUS;
ksi.ksi_code = BUS_OBJERR;
ksi.ksi_trapno = T_PROTFLT;
ksi.ksi_addr = (void *)regs->tf_eip;
trapsignal(td, &ksi);
return(EINVAL);
}
linux_to_bsd_sigset(&uc.uc_sigmask, &bmask);
kern_sigprocmask(td, SIG_SETMASK, &bmask, NULL, 0);
/*
* Restore signal context
*/
/* %gs was restored by the trampoline. */
regs->tf_fs = context->sc_fs;
regs->tf_es = context->sc_es;
regs->tf_ds = context->sc_ds;
regs->tf_edi = context->sc_edi;
regs->tf_esi = context->sc_esi;
regs->tf_ebp = context->sc_ebp;
regs->tf_ebx = context->sc_ebx;
regs->tf_edx = context->sc_edx;
regs->tf_ecx = context->sc_ecx;
regs->tf_eax = context->sc_eax;
regs->tf_eip = context->sc_eip;
regs->tf_cs = context->sc_cs;
regs->tf_eflags = eflags;
regs->tf_esp = context->sc_esp_at_signal;
regs->tf_ss = context->sc_ss;
/*
* call sigaltstack & ignore results..
*/
lss = &uc.uc_stack;
ss.ss_sp = lss->ss_sp;
ss.ss_size = lss->ss_size;
ss.ss_flags = linux_to_bsd_sigaltstack(lss->ss_flags);
#ifdef DEBUG
if (ldebug(rt_sigreturn))
printf(LMSG("rt_sigret flags: 0x%x, sp: %p, ss: 0x%x, mask: 0x%x"),
ss.ss_flags, ss.ss_sp, ss.ss_size, context->sc_mask);
#endif
(void)kern_sigaltstack(td, &ss, NULL);
return (EJUSTRETURN);
}