static int
elf_linux_fixup(register_t **stack_base, struct image_params *imgp)
{
Elf32_Auxargs *args;
Elf32_Addr *base;
Elf32_Addr *pos;
struct linux32_ps_strings *arginfo;
arginfo = (struct linux32_ps_strings *)LINUX32_PS_STRINGS;
KASSERT(curthread->td_proc == imgp->proc,
("unsafe elf_linux_fixup(), should be curproc"));
base = (Elf32_Addr *)*stack_base;
args = (Elf32_Auxargs *)imgp->auxargs;
pos = base + (imgp->args->argc + imgp->args->envc + 2);
AUXARGS_ENTRY_32(pos, LINUX_AT_SYSINFO_EHDR,
imgp->proc->p_sysent->sv_shared_page_base);
AUXARGS_ENTRY_32(pos, LINUX_AT_SYSINFO, linux32_vsyscall);
AUXARGS_ENTRY_32(pos, LINUX_AT_HWCAP, cpu_feature);
/*
* Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0,
* as it has appeared in the 2.4.0-rc7 first time.
* Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK),
* glibc falls back to the hard-coded CLK_TCK value when aux entry
* is not present.
* Also see linux_times() implementation.
*/
if (linux_kernver(curthread) >= LINUX_KERNVER_2004000)
AUXARGS_ENTRY_32(pos, LINUX_AT_CLKTCK, stclohz);
AUXARGS_ENTRY_32(pos, AT_PHDR, args->phdr);
AUXARGS_ENTRY_32(pos, AT_PHENT, args->phent);
AUXARGS_ENTRY_32(pos, AT_PHNUM, args->phnum);
AUXARGS_ENTRY_32(pos, AT_PAGESZ, args->pagesz);
AUXARGS_ENTRY_32(pos, AT_FLAGS, args->flags);
AUXARGS_ENTRY_32(pos, AT_ENTRY, args->entry);
AUXARGS_ENTRY_32(pos, AT_BASE, args->base);
AUXARGS_ENTRY_32(pos, LINUX_AT_SECURE, 0);
AUXARGS_ENTRY_32(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
AUXARGS_ENTRY_32(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
AUXARGS_ENTRY_32(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
AUXARGS_ENTRY_32(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
AUXARGS_ENTRY_32(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform));
AUXARGS_ENTRY(pos, LINUX_AT_RANDOM, PTROUT(imgp->canary));
if (imgp->execpathp != 0)
AUXARGS_ENTRY(pos, LINUX_AT_EXECFN, PTROUT(imgp->execpathp));
if (args->execfd != -1)
AUXARGS_ENTRY_32(pos, AT_EXECFD, args->execfd);
AUXARGS_ENTRY_32(pos, AT_NULL, 0);
free(imgp->auxargs, M_TEMP);
imgp->auxargs = NULL;
base--;
suword32(base, (uint32_t)imgp->args->argc);
*stack_base = (register_t *)base;
return (0);
}
static int
bsd_to_linux_msghdr(const struct msghdr *bhdr, struct l_msghdr *lhdr)
{
lhdr->msg_name = PTROUT(bhdr->msg_name);
lhdr->msg_namelen = bhdr->msg_namelen;
lhdr->msg_iov = PTROUT(bhdr->msg_iov);
lhdr->msg_iovlen = bhdr->msg_iovlen;
lhdr->msg_control = PTROUT(bhdr->msg_control);
lhdr->msg_controllen = bhdr->msg_controllen;
/* msg_flags skipped */
return (0);
}
int
linux_sigaltstack(struct thread *td, struct linux_sigaltstack_args *uap)
{
stack_t ss, oss;
l_stack_t lss;
int error;
LINUX_CTR2(sigaltstack, "%p, %p", uap->uss, uap->uoss);
if (uap->uss != NULL) {
error = copyin(uap->uss, &lss, sizeof(l_stack_t));
if (error)
return (error);
ss.ss_sp = PTRIN(lss.ss_sp);
ss.ss_size = lss.ss_size;
ss.ss_flags = linux_to_bsd_sigaltstack(lss.ss_flags);
}
error = kern_sigaltstack(td, (uap->uss != NULL) ? &ss : NULL,
(uap->uoss != NULL) ? &oss : NULL);
if (!error && uap->uoss != NULL) {
lss.ss_sp = PTROUT(oss.ss_sp);
lss.ss_size = oss.ss_size;
lss.ss_flags = bsd_to_linux_sigaltstack(oss.ss_flags);
error = copyout(&lss, uap->uoss, sizeof(l_stack_t));
}
return (error);
}
static void
bsd_to_linux_semid_ds(struct semid_ds *bsp, struct l_semid_ds *lsp)
{
bsd_to_linux_ipc_perm(&bsp->sem_perm, &lsp->sem_perm);
lsp->sem_otime = bsp->sem_otime;
lsp->sem_ctime = bsp->sem_ctime;
lsp->sem_nsems = bsp->sem_nsems;
lsp->sem_base = PTROUT(bsp->sem_base);
}
static int
elf_linux_fixup(register_t **stack_base, struct image_params *imgp)
{
struct proc *p;
Elf32_Auxargs *args;
Elf32_Addr *uplatform;
struct ps_strings *arginfo;
register_t *pos;
KASSERT(curthread->td_proc == imgp->proc,
("unsafe elf_linux_fixup(), should be curproc"));
p = imgp->proc;
arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
uplatform = (Elf32_Addr *)((caddr_t)arginfo - linux_szplatform);
args = (Elf32_Auxargs *)imgp->auxargs;
pos = *stack_base + (imgp->args->argc + imgp->args->envc + 2);
AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature);
/*
* Do not export AT_CLKTCK when emulating Linux kernel prior to 2.4.0,
* as it has appeared in the 2.4.0-rc7 first time.
* Being exported, AT_CLKTCK is returned by sysconf(_SC_CLK_TCK),
* glibc falls back to the hard-coded CLK_TCK value when aux entry
* is not present.
* Also see linux_times() implementation.
*/
if (linux_kernver(curthread) >= LINUX_KERNVER_2004000)
AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz);
AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
AUXARGS_ENTRY(pos, AT_BASE, args->base);
AUXARGS_ENTRY(pos, LINUX_AT_SECURE, 0);
AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(uplatform));
if (args->execfd != -1)
AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
AUXARGS_ENTRY(pos, AT_NULL, 0);
free(imgp->auxargs, M_TEMP);
imgp->auxargs = NULL;
(*stack_base)--;
suword(*stack_base, (register_t)imgp->args->argc);
return (0);
}
static int
elf_linux_fixup(register_t **stack_base, struct image_params *imgp)
{
Elf_Auxargs *args;
Elf_Addr *base;
Elf_Addr *pos;
struct ps_strings *arginfo;
struct proc *p;
int issetugid;
p = imgp->proc;
arginfo = (struct ps_strings *)p->p_sysent->sv_psstrings;
KASSERT(curthread->td_proc == imgp->proc,
("unsafe elf_linux_fixup(), should be curproc"));
base = (Elf64_Addr *)*stack_base;
args = (Elf64_Auxargs *)imgp->auxargs;
pos = base + (imgp->args->argc + imgp->args->envc + 2);
issetugid = p->p_flag & P_SUGID ? 1 : 0;
AUXARGS_ENTRY(pos, LINUX_AT_SYSINFO_EHDR,
imgp->proc->p_sysent->sv_shared_page_base);
AUXARGS_ENTRY(pos, LINUX_AT_HWCAP, cpu_feature);
AUXARGS_ENTRY(pos, LINUX_AT_CLKTCK, stclohz);
AUXARGS_ENTRY(pos, AT_PHDR, args->phdr);
AUXARGS_ENTRY(pos, AT_PHENT, args->phent);
AUXARGS_ENTRY(pos, AT_PHNUM, args->phnum);
AUXARGS_ENTRY(pos, AT_PAGESZ, args->pagesz);
AUXARGS_ENTRY(pos, AT_BASE, args->base);
AUXARGS_ENTRY(pos, AT_FLAGS, args->flags);
AUXARGS_ENTRY(pos, AT_ENTRY, args->entry);
AUXARGS_ENTRY(pos, AT_UID, imgp->proc->p_ucred->cr_ruid);
AUXARGS_ENTRY(pos, AT_EUID, imgp->proc->p_ucred->cr_svuid);
AUXARGS_ENTRY(pos, AT_GID, imgp->proc->p_ucred->cr_rgid);
AUXARGS_ENTRY(pos, AT_EGID, imgp->proc->p_ucred->cr_svgid);
AUXARGS_ENTRY(pos, LINUX_AT_SECURE, issetugid);
AUXARGS_ENTRY(pos, LINUX_AT_PLATFORM, PTROUT(linux_platform));
AUXARGS_ENTRY(pos, LINUX_AT_RANDOM, imgp->canary);
if (imgp->execpathp != 0)
AUXARGS_ENTRY(pos, LINUX_AT_EXECFN, imgp->execpathp);
if (args->execfd != -1)
AUXARGS_ENTRY(pos, AT_EXECFD, args->execfd);
AUXARGS_ENTRY(pos, AT_NULL, 0);
free(imgp->auxargs, M_TEMP);
imgp->auxargs = NULL;
base--;
suword(base, (uint64_t)imgp->args->argc);
*stack_base = (register_t *)base;
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_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);
}
/*
* copied from amd64/amd64/machdep.c
*
* Send an interrupt to process.
*/
static void
linux_rt_sendsig(sig_t catcher, ksiginfo_t *ksi, sigset_t *mask)
{
struct l_rt_sigframe sf, *sfp;
struct proc *p;
struct thread *td;
struct sigacts *psp;
caddr_t sp;
struct trapframe *regs;
int sig, code;
int oonstack;
td = curthread;
p = td->td_proc;
PROC_LOCK_ASSERT(p, MA_OWNED);
sig = ksi->ksi_signo;
psp = p->p_sigacts;
code = ksi->ksi_code;
mtx_assert(&psp->ps_mtx, MA_OWNED);
regs = td->td_frame;
oonstack = sigonstack(regs->tf_rsp);
LINUX_CTR4(rt_sendsig, "%p, %d, %p, %u",
catcher, sig, mask, code);
/* Allocate space for the signal handler context. */
if ((td->td_pflags & TDP_ALTSTACK) != 0 && !oonstack &&
SIGISMEMBER(psp->ps_sigonstack, sig)) {
sp = (caddr_t)td->td_sigstk.ss_sp + td->td_sigstk.ss_size -
sizeof(struct l_rt_sigframe);
} else
sp = (caddr_t)regs->tf_rsp - sizeof(struct l_rt_sigframe) - 128;
/* Align to 16 bytes. */
sfp = (struct l_rt_sigframe *)((unsigned long)sp & ~0xFul);
mtx_unlock(&psp->ps_mtx);
/* Translate the signal. */
sig = bsd_to_linux_signal(sig);
/* Save user context. */
bzero(&sf, sizeof(sf));
bsd_to_linux_sigset(mask, &sf.sf_sc.uc_sigmask);
bsd_to_linux_sigset(mask, &sf.sf_sc.uc_mcontext.sc_mask);
sf.sf_sc.uc_stack.ss_sp = PTROUT(td->td_sigstk.ss_sp);
sf.sf_sc.uc_stack.ss_size = td->td_sigstk.ss_size;
sf.sf_sc.uc_stack.ss_flags = (td->td_pflags & TDP_ALTSTACK)
? ((oonstack) ? LINUX_SS_ONSTACK : 0) : LINUX_SS_DISABLE;
PROC_UNLOCK(p);
sf.sf_sc.uc_mcontext.sc_rdi = regs->tf_rdi;
sf.sf_sc.uc_mcontext.sc_rsi = regs->tf_rsi;
sf.sf_sc.uc_mcontext.sc_rdx = regs->tf_rdx;
sf.sf_sc.uc_mcontext.sc_rbp = regs->tf_rbp;
sf.sf_sc.uc_mcontext.sc_rbx = regs->tf_rbx;
sf.sf_sc.uc_mcontext.sc_rcx = regs->tf_rcx;
sf.sf_sc.uc_mcontext.sc_rax = regs->tf_rax;
sf.sf_sc.uc_mcontext.sc_rip = regs->tf_rip;
sf.sf_sc.uc_mcontext.sc_rsp = regs->tf_rsp;
sf.sf_sc.uc_mcontext.sc_r8 = regs->tf_r8;
sf.sf_sc.uc_mcontext.sc_r9 = regs->tf_r9;
sf.sf_sc.uc_mcontext.sc_r10 = regs->tf_r10;
sf.sf_sc.uc_mcontext.sc_r11 = regs->tf_r11;
sf.sf_sc.uc_mcontext.sc_r12 = regs->tf_r12;
sf.sf_sc.uc_mcontext.sc_r13 = regs->tf_r13;
sf.sf_sc.uc_mcontext.sc_r14 = regs->tf_r14;
sf.sf_sc.uc_mcontext.sc_r15 = regs->tf_r15;
sf.sf_sc.uc_mcontext.sc_cs = regs->tf_cs;
sf.sf_sc.uc_mcontext.sc_rflags = regs->tf_rflags;
sf.sf_sc.uc_mcontext.sc_err = regs->tf_err;
sf.sf_sc.uc_mcontext.sc_trapno = bsd_to_linux_trapcode(code);
sf.sf_sc.uc_mcontext.sc_cr2 = (register_t)ksi->ksi_addr;
/* Build the argument list for the signal handler. */
regs->tf_rdi = sig; /* arg 1 in %rdi */
regs->tf_rax = 0;
regs->tf_rsi = (register_t)&sfp->sf_si; /* arg 2 in %rsi */
regs->tf_rdx = (register_t)&sfp->sf_sc; /* arg 3 in %rdx */
sf.sf_handler = catcher;
/* Fill in POSIX parts */
ksiginfo_to_lsiginfo(ksi, &sf.sf_si, sig);
/*
* Copy the sigframe out to the user's stack.
*/
if (copyout(&sf, sfp, sizeof(*sfp)) != 0) {
#ifdef DEBUG
printf("process %ld has trashed its stack\n", (long)p->p_pid);
#endif
PROC_LOCK(p);
sigexit(td, SIGILL);
}
regs->tf_rsp = (long)sfp;
regs->tf_rip = linux_rt_sigcode;
regs->tf_rflags &= ~(PSL_T | PSL_D);
regs->tf_cs = _ucodesel;
set_pcb_flags(td->td_pcb, PCB_FULL_IRET);
PROC_LOCK(p);
mtx_lock(&psp->ps_mtx);
}
/*
* Send an interrupt to process.
*
* Stack is set up to allow sigcode stored
* in u. to call routine, followed by kcall
* to sigreturn routine below. After sigreturn
* resets the signal mask, the stack, and the
* frame pointer, it returns to the user
* specified pc, psl.
*/
static void
linux_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_sigframe *fp, frame;
l_sigset_t lmask;
int oonstack, i;
int sig, 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);
if (SIGISMEMBER(psp->ps_siginfo, sig)) {
/* Signal handler installed with SA_SIGINFO. */
linux_rt_sendsig(catcher, ksi, mask);
return;
}
regs = td->td_frame;
oonstack = sigonstack(regs->tf_rsp);
#ifdef DEBUG
if (ldebug(sendsig))
printf(ARGS(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_sigframe *)(td->td_sigstk.ss_sp +
td->td_sigstk.ss_size - sizeof(struct l_sigframe));
} else
fp = (struct l_sigframe *)regs->tf_rsp - 1;
mtx_unlock(&psp->ps_mtx);
PROC_UNLOCK(p);
/*
* 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;
bsd_to_linux_sigset(mask, &lmask);
/*
* Build the signal context to be used by sigreturn.
*/
frame.sf_sc.sc_mask = lmask.__bits[0];
frame.sf_sc.sc_gs = regs->tf_gs;
frame.sf_sc.sc_fs = regs->tf_fs;
frame.sf_sc.sc_es = regs->tf_es;
frame.sf_sc.sc_ds = regs->tf_ds;
frame.sf_sc.sc_edi = regs->tf_rdi;
frame.sf_sc.sc_esi = regs->tf_rsi;
frame.sf_sc.sc_ebp = regs->tf_rbp;
frame.sf_sc.sc_ebx = regs->tf_rbx;
frame.sf_sc.sc_edx = regs->tf_rdx;
frame.sf_sc.sc_ecx = regs->tf_rcx;
frame.sf_sc.sc_eax = regs->tf_rax;
frame.sf_sc.sc_eip = regs->tf_rip;
frame.sf_sc.sc_cs = regs->tf_cs;
frame.sf_sc.sc_eflags = regs->tf_rflags;
frame.sf_sc.sc_esp_at_signal = regs->tf_rsp;
frame.sf_sc.sc_ss = regs->tf_ss;
frame.sf_sc.sc_err = regs->tf_err;
frame.sf_sc.sc_cr2 = (u_int32_t)(uintptr_t)ksi->ksi_addr;
frame.sf_sc.sc_trapno = bsd_to_linux_trapcode(code);
for (i = 0; i < (LINUX_NSIG_WORDS-1); i++)
frame.sf_extramask[i] = lmask.__bits[i+1];
if (copyout(&frame, fp, sizeof(frame)) != 0) {
/*
* Process has trashed its stack; give it an illegal
* instruction to halt it in its tracks.
*/
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;
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);
}
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);
}