int
linux_arch_prctl(struct thread *td, struct linux_arch_prctl_args *args)
{
int error;
struct pcb *pcb;
LINUX_CTR2(arch_prctl, "0x%x, %p", args->code, args->addr);
error = ENOTSUP;
pcb = td->td_pcb;
switch (args->code) {
case LINUX_ARCH_GET_GS:
error = copyout(&pcb->pcb_gsbase, (unsigned long *)args->addr,
sizeof(args->addr));
break;
case LINUX_ARCH_SET_GS:
if (args->addr >= VM_MAXUSER_ADDRESS)
return(EPERM);
break;
case LINUX_ARCH_GET_FS:
error = copyout(&pcb->pcb_fsbase, (unsigned long *)args->addr,
sizeof(args->addr));
break;
case LINUX_ARCH_SET_FS:
error = linux_set_cloned_tls(td, (void *)args->addr);
break;
default:
error = EINVAL;
}
return (error);
}
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);
}
int
linux_exit(struct thread *td, struct linux_exit_args *args)
{
struct linux_emuldata *em;
em = em_find(td);
KASSERT(em != NULL, ("exit: emuldata not found.\n"));
LINUX_CTR2(exit, "thread(%d) (%d)", em->em_tid, args->rval);
linux_thread_detach(td);
/*
* XXX. When the last two threads of a process
* exit via pthread_exit() try thr_exit() first.
*/
kern_thr_exit(td);
exit1(td, W_EXITCODE(args->rval, 0));
/* NOTREACHED */
}
int
linux_rt_sigsuspend(struct thread *td, struct linux_rt_sigsuspend_args *uap)
{
l_sigset_t lmask;
sigset_t sigmask;
int error;
LINUX_CTR2(rt_sigsuspend, "%p, %ld",
uap->newset, uap->sigsetsize);
if (uap->sigsetsize != sizeof(l_sigset_t))
return (EINVAL);
error = copyin(uap->newset, &lmask, sizeof(l_sigset_t));
if (error)
return (error);
linux_to_bsd_sigset(&lmask, &sigmask);
return (kern_sigsuspend(td, sigmask));
}
void
linux_thread_detach(struct thread *td)
{
struct linux_sys_futex_args cup;
struct linux_emuldata *em;
int *child_clear_tid;
int error;
em = em_find(td);
KASSERT(em != NULL, ("thread_detach: emuldata not found.\n"));
LINUX_CTR1(thread_detach, "thread(%d)", em->em_tid);
release_futexes(td, em);
child_clear_tid = em->child_clear_tid;
if (child_clear_tid != NULL) {
LINUX_CTR2(thread_detach, "thread(%d) %p",
em->em_tid, child_clear_tid);
error = suword32(child_clear_tid, 0);
if (error != 0)
return;
cup.uaddr = child_clear_tid;
cup.op = LINUX_FUTEX_WAKE;
cup.val = 1; /* wake one */
cup.timeout = NULL;
cup.uaddr2 = NULL;
cup.val3 = 0;
error = linux_sys_futex(td, &cup);
/*
* this cannot happen at the moment and if this happens it
* probably means there is a user space bug
*/
if (error != 0)
linux_msg(td, "futex stuff in thread_detach failed.");
}
}
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);
}
int
linux_mmap2(struct thread *td, struct linux_mmap2_args *args)
{
struct proc *p = td->td_proc;
struct mmap_args /* {
caddr_t addr;
size_t len;
int prot;
int flags;
int fd;
long pad;
off_t pos;
} */ bsd_args;
int error;
struct file *fp;
cap_rights_t rights;
LINUX_CTR6(mmap2, "0x%lx, %ld, %ld, 0x%08lx, %ld, 0x%lx",
args->addr, args->len, args->prot,
args->flags, args->fd, args->pgoff);
error = 0;
bsd_args.flags = 0;
fp = NULL;
/*
* Linux mmap(2):
* You must specify exactly one of MAP_SHARED and MAP_PRIVATE
*/
if (! ((args->flags & LINUX_MAP_SHARED) ^
(args->flags & LINUX_MAP_PRIVATE)))
return (EINVAL);
if (args->flags & LINUX_MAP_SHARED)
bsd_args.flags |= MAP_SHARED;
if (args->flags & LINUX_MAP_PRIVATE)
bsd_args.flags |= MAP_PRIVATE;
if (args->flags & LINUX_MAP_FIXED)
bsd_args.flags |= MAP_FIXED;
if (args->flags & LINUX_MAP_ANON)
bsd_args.flags |= MAP_ANON;
else
bsd_args.flags |= MAP_NOSYNC;
if (args->flags & LINUX_MAP_GROWSDOWN)
bsd_args.flags |= MAP_STACK;
/*
* PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC
* on Linux/i386. We do this to ensure maximum compatibility.
* Linux/ia64 does the same in i386 emulation mode.
*/
bsd_args.prot = args->prot;
if (bsd_args.prot & (PROT_READ | PROT_WRITE | PROT_EXEC))
bsd_args.prot |= PROT_READ | PROT_EXEC;
/* Linux does not check file descriptor when MAP_ANONYMOUS is set. */
bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : args->fd;
if (bsd_args.fd != -1) {
/*
* Linux follows Solaris mmap(2) description:
* The file descriptor fildes is opened with
* read permission, regardless of the
* protection options specified.
*/
error = fget(td, bsd_args.fd,
cap_rights_init(&rights, CAP_MMAP), &fp);
if (error != 0 )
return (error);
if (fp->f_type != DTYPE_VNODE) {
fdrop(fp, td);
return (EINVAL);
}
/* Linux mmap() just fails for O_WRONLY files */
if (!(fp->f_flag & FREAD)) {
fdrop(fp, td);
return (EACCES);
}
fdrop(fp, td);
}
if (args->flags & LINUX_MAP_GROWSDOWN) {
/*
* The Linux MAP_GROWSDOWN option does not limit auto
* growth of the region. Linux mmap with this option
* takes as addr the inital BOS, and as len, the initial
* region size. It can then grow down from addr without
* limit. However, Linux threads has an implicit internal
* limit to stack size of STACK_SIZE. Its just not
* enforced explicitly in Linux. But, here we impose
* a limit of (STACK_SIZE - GUARD_SIZE) on the stack
* region, since we can do this with our mmap.
*
* Our mmap with MAP_STACK takes addr as the maximum
* downsize limit on BOS, and as len the max size of
* the region. It then maps the top SGROWSIZ bytes,
* and auto grows the region down, up to the limit
* in addr.
*
* If we don't use the MAP_STACK option, the effect
* of this code is to allocate a stack region of a
* fixed size of (STACK_SIZE - GUARD_SIZE).
*/
if ((caddr_t)PTRIN(args->addr) + args->len >
p->p_vmspace->vm_maxsaddr) {
/*
* Some Linux apps will attempt to mmap
* thread stacks near the top of their
* address space. If their TOS is greater
* than vm_maxsaddr, vm_map_growstack()
* will confuse the thread stack with the
* process stack and deliver a SEGV if they
* attempt to grow the thread stack past their
* current stacksize rlimit. To avoid this,
* adjust vm_maxsaddr upwards to reflect
* the current stacksize rlimit rather
* than the maximum possible stacksize.
* It would be better to adjust the
* mmap'ed region, but some apps do not check
* mmap's return value.
*/
PROC_LOCK(p);
p->p_vmspace->vm_maxsaddr = (char *)USRSTACK -
lim_cur_proc(p, RLIMIT_STACK);
PROC_UNLOCK(p);
}
/*
* This gives us our maximum stack size and a new BOS.
* If we're using VM_STACK, then mmap will just map
* the top SGROWSIZ bytes, and let the stack grow down
* to the limit at BOS. If we're not using VM_STACK
* we map the full stack, since we don't have a way
* to autogrow it.
*/
if (args->len > STACK_SIZE - GUARD_SIZE) {
bsd_args.addr = (caddr_t)PTRIN(args->addr);
bsd_args.len = args->len;
} else {
bsd_args.addr = (caddr_t)PTRIN(args->addr) -
(STACK_SIZE - GUARD_SIZE - args->len);
bsd_args.len = STACK_SIZE - GUARD_SIZE;
}
} else {
bsd_args.addr = (caddr_t)PTRIN(args->addr);
bsd_args.len = args->len;
}
bsd_args.pos = (off_t)args->pgoff;
error = sys_mmap(td, &bsd_args);
LINUX_CTR2(mmap2, "return: %d (%p)",
error, td->td_retval[0]);
return (error);
}
int
linux_mmap_common(struct thread *td, uintptr_t addr, size_t len, int prot,
int flags, int fd, off_t pos)
{
struct proc *p = td->td_proc;
struct vmspace *vms = td->td_proc->p_vmspace;
struct mmap_args /* {
caddr_t addr;
size_t len;
int prot;
int flags;
int fd;
off_t pos;
} */ bsd_args;
int error;
struct file *fp;
cap_rights_t rights;
LINUX_CTR6(mmap2, "0x%lx, %ld, %ld, 0x%08lx, %ld, 0x%lx",
addr, len, prot, flags, fd, pos);
error = 0;
bsd_args.flags = 0;
fp = NULL;
/*
* Linux mmap(2):
* You must specify exactly one of MAP_SHARED and MAP_PRIVATE
*/
if (!((flags & LINUX_MAP_SHARED) ^ (flags & LINUX_MAP_PRIVATE)))
return (EINVAL);
if (flags & LINUX_MAP_SHARED)
bsd_args.flags |= MAP_SHARED;
if (flags & LINUX_MAP_PRIVATE)
bsd_args.flags |= MAP_PRIVATE;
if (flags & LINUX_MAP_FIXED)
bsd_args.flags |= MAP_FIXED;
if (flags & LINUX_MAP_ANON) {
/* Enforce pos to be on page boundary, then ignore. */
if ((pos & PAGE_MASK) != 0)
return (EINVAL);
pos = 0;
bsd_args.flags |= MAP_ANON;
} else
bsd_args.flags |= MAP_NOSYNC;
if (flags & LINUX_MAP_GROWSDOWN)
bsd_args.flags |= MAP_STACK;
/*
* PROT_READ, PROT_WRITE, or PROT_EXEC implies PROT_READ and PROT_EXEC
* on Linux/i386 if the binary requires executable stack.
* We do this only for IA32 emulation as on native i386 this is does not
* make sense without PAE.
*
* XXX. Linux checks that the file system is not mounted with noexec.
*/
bsd_args.prot = prot;
#if defined(__amd64__)
linux_fixup_prot(td, &bsd_args.prot);
#endif
/* Linux does not check file descriptor when MAP_ANONYMOUS is set. */
bsd_args.fd = (bsd_args.flags & MAP_ANON) ? -1 : fd;
if (bsd_args.fd != -1) {
/*
* Linux follows Solaris mmap(2) description:
* The file descriptor fildes is opened with
* read permission, regardless of the
* protection options specified.
*/
error = fget(td, bsd_args.fd,
cap_rights_init(&rights, CAP_MMAP), &fp);
if (error != 0)
return (error);
if (fp->f_type != DTYPE_VNODE) {
fdrop(fp, td);
return (EINVAL);
}
/* Linux mmap() just fails for O_WRONLY files */
if (!(fp->f_flag & FREAD)) {
fdrop(fp, td);
return (EACCES);
}
fdrop(fp, td);
}
if (flags & LINUX_MAP_GROWSDOWN) {
/*
* The Linux MAP_GROWSDOWN option does not limit auto
* growth of the region. Linux mmap with this option
* takes as addr the initial BOS, and as len, the initial
* region size. It can then grow down from addr without
* limit. However, Linux threads has an implicit internal
* limit to stack size of STACK_SIZE. Its just not
* enforced explicitly in Linux. But, here we impose
* a limit of (STACK_SIZE - GUARD_SIZE) on the stack
* region, since we can do this with our mmap.
*
* Our mmap with MAP_STACK takes addr as the maximum
* downsize limit on BOS, and as len the max size of
* the region. It then maps the top SGROWSIZ bytes,
* and auto grows the region down, up to the limit
* in addr.
*
* If we don't use the MAP_STACK option, the effect
* of this code is to allocate a stack region of a
* fixed size of (STACK_SIZE - GUARD_SIZE).
*/
if ((caddr_t)addr + len > vms->vm_maxsaddr) {
/*
* Some Linux apps will attempt to mmap
* thread stacks near the top of their
* address space. If their TOS is greater
* than vm_maxsaddr, vm_map_growstack()
* will confuse the thread stack with the
* process stack and deliver a SEGV if they
* attempt to grow the thread stack past their
* current stacksize rlimit. To avoid this,
* adjust vm_maxsaddr upwards to reflect
* the current stacksize rlimit rather
* than the maximum possible stacksize.
* It would be better to adjust the
* mmap'ed region, but some apps do not check
* mmap's return value.
*/
PROC_LOCK(p);
vms->vm_maxsaddr = (char *)p->p_sysent->sv_usrstack -
lim_cur_proc(p, RLIMIT_STACK);
PROC_UNLOCK(p);
}
/*
* This gives us our maximum stack size and a new BOS.
* If we're using VM_STACK, then mmap will just map
* the top SGROWSIZ bytes, and let the stack grow down
* to the limit at BOS. If we're not using VM_STACK
* we map the full stack, since we don't have a way
* to autogrow it.
*/
if (len > STACK_SIZE - GUARD_SIZE) {
bsd_args.addr = (caddr_t)addr;
bsd_args.len = len;
} else {
bsd_args.addr = (caddr_t)addr -
(STACK_SIZE - GUARD_SIZE - len);
bsd_args.len = STACK_SIZE - GUARD_SIZE;
}
} else {
bsd_args.addr = (caddr_t)addr;
bsd_args.len = len;
}
bsd_args.pos = pos;
error = sys_mmap(td, &bsd_args);
LINUX_CTR2(mmap2, "return: %d (%p)", error, td->td_retval[0]);
return (error);
}
