glibc/hurd/hurdexec.c
Jeremie Koenig 653d74f12a hurd: Global signal disposition
This adds _hurd_sigstate_set_global_rcv used by libpthread to enable
POSIX-confirming behavior of signals on a per-thread basis.

This also provides a sigstate destructor _hurd_sigstate_delete, and a
global process signal state, which needs to be locked and check when
global disposition is enabled, thus the addition of _hurd_sigstate_lock
_hurd_sigstate_actions _hurd_sigstate_pending _hurd_sigstate_unlock helpers.

This also updates all the glibc code accordingly.

This also drops support for get_int(INIT_SIGMASK), which did not make sense
any more since we do not have a single signal thread any more.

During fork/spawn, this also reinitializes the child global sigstate's
lock. That cures an issue that would very rarely cause a deadlock in the
child in fork, tries to unlock ss' critical section lock at the end of
fork.  This will typically (always?) be observed in /bin/sh, which is not
surprising as that is the foremost caller of fork.

To reproduce an intermediate state, add an endless loop if
_hurd_global_sigstate is locked after __proc_dostop (cast through
volatile); that is, while still being in the fork's parent process.

When that triggers (use the libtool testsuite), the signal thread has
already locked ss (which is _hurd_global_sigstate), and is stuck at
hurdsig.c:685 in post_signal, trying to lock _hurd_siglock (which the
main thread already has locked and keeps locked until after
__task_create).  This is the case that ss->thread == MACH_PORT_NULL, that
is, a global signal.  In the main thread, between __proc_dostop and
__task_create is the __thread_abort call on the signal thread which would
abort any current kernel operation (but leave ss locked).  Later in fork,
in the parent, when _hurd_siglock is unlocked in fork, the parent's
signal thread can proceed and will unlock eventually the global sigstate.
In the client, _hurd_siglock will likewise be unlocked, but the global
sigstate never will be, as the client's signal thread has been configured
to restart execution from _hurd_msgport_receive.  Thus, when the child
tries to unlock ss' critical section lock at the end of fork, it will
first lock the global sigstate, will spin trying to lock it, which can
never be successful, and we get our deadlock.

Options seem to be:

  * Move the locking of _hurd_siglock earlier in post_signal -- but that
    may generally impact performance, if this locking isn't generally
    needed anyway?

    On the other hand, would it actually make sense to wait here until we
    are not any longer in a critical section (which is meant to disable
    signal delivery anyway (but not for preempted signals?))?

  * Clear the global sigstate in the fork's child with the rationale that
    we're anyway restarting the signal thread from a clean state.  This
    has now been implemented.

Why has this problem not been observed before Jérémie's patches?  (Or has
it?  Perhaps even more rarely?)  In _S_msg_sig_post, the signal is now
posted to a *global receiver thread*, whereas previously it was posted to
the *designated signal-receiving thread*.  The latter one was in a
critical section in fork, so didn't try to handle the signal until after
leaving the critical section?  (Not completely analyzed and verified.)

Another question is what the signal is that is being received
during/around the time __proc_dostop executes.
2019-12-29 18:32:49 +01:00

440 lines
14 KiB
C

/* Copyright (C) 1991-2019 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<https://www.gnu.org/licenses/>. */
#include <errno.h>
#include <unistd.h>
#include <fcntl.h>
#include <limits.h>
#include <stdlib.h>
#include <string.h>
#include <hurd.h>
#include <hurd/fd.h>
#include <hurd/signal.h>
#include <hurd/id.h>
#include <assert.h>
#include <argz.h>
/* Overlay TASK, executing FILE with arguments ARGV and environment ENVP.
If TASK == mach_task_self (), some ports are dealloc'd by the exec server.
ARGV and ENVP are terminated by NULL pointers.
Deprecated: use _hurd_exec_paths instead. */
error_t
_hurd_exec (task_t task, file_t file,
char *const argv[], char *const envp[])
{
return _hurd_exec_paths (task, file, NULL, NULL, argv, envp);
}
link_warning (_hurd_exec,
"_hurd_exec is deprecated, use _hurd_exec_paths instead");
/* Overlay TASK, executing FILE with arguments ARGV and environment ENVP.
If TASK == mach_task_self (), some ports are dealloc'd by the exec server.
ARGV and ENVP are terminated by NULL pointers. PATH is the relative path to
FILE and ABSPATH is the absolute path to FILE. Passing NULL, though possible,
should be avoided, since then the exec server may not know the path to
FILE if FILE is a script, and will then pass /dev/fd/N to the
interpreter. */
error_t
_hurd_exec_paths (task_t task, file_t file,
const char *path, const char *abspath,
char *const argv[], char *const envp[])
{
error_t err;
char *args, *env;
size_t argslen, envlen;
int ints[INIT_INT_MAX];
mach_port_t ports[_hurd_nports];
struct hurd_userlink ulink_ports[_hurd_nports];
inline void free_port (unsigned int i)
{
_hurd_port_free (&_hurd_ports[i], &ulink_ports[i], ports[i]);
}
file_t *dtable;
unsigned int dtablesize, i;
struct hurd_port **dtable_cells;
struct hurd_userlink *ulink_dtable;
struct hurd_sigstate *ss;
mach_port_t *please_dealloc, *pdp;
int reauth = 0;
/* XXX needs to be hurdmalloc XXX */
if (argv == NULL)
args = NULL, argslen = 0;
else if (err = __argz_create (argv, &args, &argslen))
return err;
if (envp == NULL)
env = NULL, envlen = 0;
else if (err = __argz_create (envp, &env, &envlen))
goto outargs;
/* Load up the ports to give to the new program. */
for (i = 0; i < _hurd_nports; ++i)
if (i == INIT_PORT_PROC && task != __mach_task_self ())
{
/* This is another task, so we need to ask the proc server
for the right proc server port for it. */
if (err = __USEPORT (PROC, __proc_task2proc (port, task, &ports[i])))
{
while (--i > 0)
free_port (i);
goto outenv;
}
}
else
ports[i] = _hurd_port_get (&_hurd_ports[i], &ulink_ports[i]);
/* Load up the ints to give the new program. */
for (i = 0; i < INIT_INT_MAX; ++i)
switch (i)
{
case INIT_UMASK:
ints[i] = _hurd_umask;
break;
case INIT_SIGMASK:
case INIT_SIGIGN:
case INIT_SIGPENDING:
/* We will set these all below. */
break;
case INIT_TRACEMASK:
ints[i] = _hurdsig_traced;
break;
default:
ints[i] = 0;
}
ss = _hurd_self_sigstate ();
assert (! __spin_lock_locked (&ss->critical_section_lock));
__spin_lock (&ss->critical_section_lock);
_hurd_sigstate_lock (ss);
struct sigaction *actions = _hurd_sigstate_actions (ss);
ints[INIT_SIGMASK] = ss->blocked;
ints[INIT_SIGPENDING] = _hurd_sigstate_pending (ss);
ints[INIT_SIGIGN] = 0;
for (i = 1; i < NSIG; ++i)
if (actions[i].sa_handler == SIG_IGN)
ints[INIT_SIGIGN] |= __sigmask (i);
/* We hold the sigstate lock until the exec has failed so that no signal
can arrive between when we pack the blocked and ignored signals, and
when the exec actually happens. A signal handler could change what
signals are blocked and ignored. Either the change will be reflected
in the exec, or the signal will never be delivered. Setting the
critical section flag avoids anything we call trying to acquire the
sigstate lock. */
_hurd_sigstate_unlock (ss);
/* Pack up the descriptor table to give the new program. */
__mutex_lock (&_hurd_dtable_lock);
dtablesize = _hurd_dtable ? _hurd_dtablesize : _hurd_init_dtablesize;
if (task == __mach_task_self ())
/* Request the exec server to deallocate some ports from us if the exec
succeeds. The init ports and descriptor ports will arrive in the
new program's exec_startup message. If we failed to deallocate
them, the new program would have duplicate user references for them.
But we cannot deallocate them ourselves, because we must still have
them after a failed exec call. */
please_dealloc = __alloca ((_hurd_nports + 3 + (3 * dtablesize))
* sizeof (mach_port_t));
else
please_dealloc = NULL;
pdp = please_dealloc;
if (_hurd_dtable != NULL)
{
dtable = __alloca (dtablesize * sizeof (dtable[0]));
ulink_dtable = __alloca (dtablesize * sizeof (ulink_dtable[0]));
dtable_cells = __alloca (dtablesize * sizeof (dtable_cells[0]));
for (i = 0; i < dtablesize; ++i)
{
struct hurd_fd *const d = _hurd_dtable[i];
if (d == NULL)
{
dtable[i] = MACH_PORT_NULL;
continue;
}
__spin_lock (&d->port.lock);
if (d->flags & FD_CLOEXEC)
{
/* This descriptor is marked to be closed on exec.
So don't pass it to the new program. */
dtable[i] = MACH_PORT_NULL;
if (pdp && d->port.port != MACH_PORT_NULL)
{
/* We still need to deallocate the ports. */
*pdp++ = d->port.port;
if (d->ctty.port != MACH_PORT_NULL)
*pdp++ = d->ctty.port;
}
__spin_unlock (&d->port.lock);
}
else
{
if (pdp && d->ctty.port != MACH_PORT_NULL)
/* All the elements of DTABLE are added to PLEASE_DEALLOC
below, so we needn't add the port itself.
But we must deallocate the ctty port as well as
the normal port that got installed in DTABLE[I]. */
*pdp++ = d->ctty.port;
dtable[i] = _hurd_port_locked_get (&d->port, &ulink_dtable[i]);
dtable_cells[i] = &d->port;
}
}
}
else
{
dtable = _hurd_init_dtable;
ulink_dtable = NULL;
dtable_cells = NULL;
}
/* Prune trailing null ports from the descriptor table. */
while (dtablesize > 0 && dtable[dtablesize - 1] == MACH_PORT_NULL)
--dtablesize;
/* See if we need to diddle the auth port of the new program.
The purpose of this is to get the effect setting the saved-set UID and
GID to the respective effective IDs after the exec, as POSIX.1 requires.
Note that we don't reauthenticate with the proc server; that would be a
no-op since it only keeps track of the effective UIDs, and if it did
keep track of the available IDs we would have the problem that we'd be
changing the IDs before the exec and have to change them back after a
failure. Arguably we could skip all the reauthentications because the
available IDs have no bearing on any filesystem. But the conservative
approach is to reauthenticate all the io ports so that no state anywhere
reflects that our whole ID set differs from what we've set it to. */
__mutex_lock (&_hurd_id.lock);
err = _hurd_check_ids ();
if (err == 0 && ((_hurd_id.aux.nuids >= 2 && _hurd_id.gen.nuids >= 1
&& _hurd_id.aux.uids[1] != _hurd_id.gen.uids[0])
|| (_hurd_id.aux.ngids >= 2 && _hurd_id.gen.ngids >= 1
&& _hurd_id.aux.gids[1] != _hurd_id.gen.gids[0])))
{
/* We have euid != svuid or egid != svgid. POSIX.1 says that exec
sets svuid = euid and svgid = egid. So we must get a new auth
port and reauthenticate everything with it. We'll pass the new
ports in file_exec_paths instead of our own ports. */
auth_t newauth;
_hurd_id.aux.uids[1] = _hurd_id.gen.uids[0];
_hurd_id.aux.gids[1] = _hurd_id.gen.gids[0];
_hurd_id.valid = 0;
if (_hurd_id.rid_auth != MACH_PORT_NULL)
{
__mach_port_deallocate (__mach_task_self (), _hurd_id.rid_auth);
_hurd_id.rid_auth = MACH_PORT_NULL;
}
err = __auth_makeauth (ports[INIT_PORT_AUTH],
NULL, MACH_MSG_TYPE_COPY_SEND, 0,
_hurd_id.gen.uids, _hurd_id.gen.nuids,
_hurd_id.aux.uids, _hurd_id.aux.nuids,
_hurd_id.gen.gids, _hurd_id.gen.ngids,
_hurd_id.aux.gids, _hurd_id.aux.ngids,
&newauth);
if (err == 0)
{
/* Now we have to reauthenticate the ports with this new ID.
*/
inline error_t reauth_io (io_t port, io_t *newport)
{
mach_port_t ref = __mach_reply_port ();
*newport = MACH_PORT_NULL;
error_t err = __io_reauthenticate (port,
ref, MACH_MSG_TYPE_MAKE_SEND);
if (!err)
err = __auth_user_authenticate (newauth,
ref, MACH_MSG_TYPE_MAKE_SEND,
newport);
__mach_port_destroy (__mach_task_self (), ref);
return err;
}
inline void reauth_port (unsigned int idx)
{
io_t newport;
err = reauth_io (ports[idx], &newport) ?: err;
if (pdp)
*pdp++ = ports[idx]; /* XXX presumed still in _hurd_ports */
free_port (idx);
ports[idx] = newport;
}
if (pdp)
*pdp++ = ports[INIT_PORT_AUTH];
free_port (INIT_PORT_AUTH);
ports[INIT_PORT_AUTH] = newauth;
reauth_port (INIT_PORT_CRDIR);
reauth_port (INIT_PORT_CWDIR);
if (!err)
{
/* Now we'll reauthenticate each file descriptor. */
if (ulink_dtable == NULL)
{
assert (dtable == _hurd_init_dtable);
dtable = __alloca (dtablesize * sizeof (dtable[0]));
for (i = 0; i < dtablesize; ++i)
if (_hurd_init_dtable[i] != MACH_PORT_NULL)
{
if (pdp)
*pdp++ = _hurd_init_dtable[i];
err = reauth_io (_hurd_init_dtable[i], &dtable[i]);
if (err)
{
while (++i < dtablesize)
dtable[i] = MACH_PORT_NULL;
break;
}
}
else
dtable[i] = MACH_PORT_NULL;
}
else
{
if (pdp)
{
/* Ask to deallocate all the old fd ports,
since we will have new ones in DTABLE. */
memcpy (pdp, dtable, dtablesize * sizeof pdp[0]);
pdp += dtablesize;
}
for (i = 0; i < dtablesize; ++i)
if (dtable[i] != MACH_PORT_NULL)
{
io_t newport;
err = reauth_io (dtable[i], &newport);
_hurd_port_free (dtable_cells[i], &ulink_dtable[i],
dtable[i]);
dtable[i] = newport;
if (err)
{
while (++i < dtablesize)
_hurd_port_free (dtable_cells[i],
&ulink_dtable[i], dtable[i]);
break;
}
}
ulink_dtable = NULL;
dtable_cells = NULL;
}
}
}
reauth = 1;
}
__mutex_unlock (&_hurd_id.lock);
/* The information is all set up now. Try to exec the file. */
if (!err)
{
int flags;
if (pdp)
{
/* Request the exec server to deallocate some ports from us if
the exec succeeds. The init ports and descriptor ports will
arrive in the new program's exec_startup message. If we
failed to deallocate them, the new program would have
duplicate user references for them. But we cannot deallocate
them ourselves, because we must still have them after a failed
exec call. */
for (i = 0; i < _hurd_nports; ++i)
*pdp++ = ports[i];
for (i = 0; i < dtablesize; ++i)
*pdp++ = dtable[i];
}
flags = 0;
#ifdef EXEC_SIGTRAP
/* PTRACE_TRACEME sets all bits in _hurdsig_traced, which is
propagated through exec by INIT_TRACEMASK, so this checks if
PTRACE_TRACEME has been called in this process in any of its
current or prior lives. */
if (__sigismember (&_hurdsig_traced, SIGKILL))
flags |= EXEC_SIGTRAP;
#endif
err = __file_exec_paths (file, task, flags,
path ? path : "",
abspath ? abspath : "",
args, argslen, env, envlen,
dtable, MACH_MSG_TYPE_COPY_SEND, dtablesize,
ports, MACH_MSG_TYPE_COPY_SEND,
_hurd_nports,
ints, INIT_INT_MAX,
please_dealloc, pdp - please_dealloc,
&_hurd_msgport,
task == __mach_task_self () ? 1 : 0);
/* Fall back for backwards compatibility. This can just be removed
when __file_exec goes away. */
if (err == MIG_BAD_ID)
err = __file_exec (file, task, flags,
args, argslen, env, envlen,
dtable, MACH_MSG_TYPE_COPY_SEND, dtablesize,
ports, MACH_MSG_TYPE_COPY_SEND, _hurd_nports,
ints, INIT_INT_MAX,
please_dealloc, pdp - please_dealloc,
&_hurd_msgport,
task == __mach_task_self () ? 1 : 0);
}
/* Release references to the standard ports. */
for (i = 0; i < _hurd_nports; ++i)
if ((i == INIT_PORT_PROC && task != __mach_task_self ())
|| (reauth && (i == INIT_PORT_AUTH
|| i == INIT_PORT_CRDIR || i == INIT_PORT_CWDIR)))
__mach_port_deallocate (__mach_task_self (), ports[i]);
else
free_port (i);
/* Release references to the file descriptor ports. */
if (ulink_dtable != NULL)
{
for (i = 0; i < dtablesize; ++i)
if (dtable[i] != MACH_PORT_NULL)
_hurd_port_free (dtable_cells[i], &ulink_dtable[i], dtable[i]);
}
else if (dtable && dtable != _hurd_init_dtable)
for (i = 0; i < dtablesize; ++i)
__mach_port_deallocate (__mach_task_self (), dtable[i]);
/* Release lock on the file descriptor table. */
__mutex_unlock (&_hurd_dtable_lock);
/* Safe to let signals happen now. */
_hurd_critical_section_unlock (ss);
outargs:
free (args);
outenv:
free (env);
return err;
}
libc_hidden_def (_hurd_exec_paths)