core: go on a killing spree when transitioning from initrd to main system

This commit is contained in:
Lennart Poettering 2012-07-13 14:41:57 +02:00
parent 32c4bef882
commit bd3fa1d243
5 changed files with 222 additions and 189 deletions

View File

@ -877,7 +877,9 @@ libsystemd_core_la_SOURCES = \
src/core/build.h \
src/core/sysfs-show.h \
src/core/switch-root.h \
src/core/switch-root.c
src/core/switch-root.c \
src/core/killall.h \
src/core/killall.c
nodist_libsystemd_core_la_SOURCES = \
src/core/load-fragment-gperf.c \
@ -1134,7 +1136,9 @@ systemd_shutdown_SOURCES = \
src/core/umount.h \
src/core/shutdown.c \
src/core/mount-setup.c \
src/core/mount-setup.h
src/core/mount-setup.h \
src/core/killall.h \
src/core/killall.c
systemd_shutdown_LDADD = \
libsystemd-label.la \

176
src/core/killall.c Normal file
View File

@ -0,0 +1,176 @@
/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
/***
This file is part of systemd.
Copyright 2010 ProFUSION embedded systems
systemd 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.
systemd 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 systemd; If not, see <http://www.gnu.org/licenses/>.
***/
#include <sys/wait.h>
#include <signal.h>
#include <errno.h>
#include "util.h"
#include "def.h"
#include "killall.h"
#define TIMEOUT_USEC (5 * USEC_PER_SEC)
static bool ignore_proc(pid_t pid) {
char buf[PATH_MAX];
FILE *f;
char c;
size_t count;
uid_t uid;
int r;
/* We are PID 1, let's not commit suicide */
if (pid == 1)
return true;
r = get_process_uid(pid, &uid);
if (r < 0)
return true; /* not really, but better safe than sorry */
/* Non-root processes otherwise are always subject to be killed */
if (uid != 0)
return false;
snprintf(buf, sizeof(buf), "/proc/%lu/cmdline", (unsigned long) pid);
char_array_0(buf);
f = fopen(buf, "re");
if (!f)
return true; /* not really, but has the desired effect */
count = fread(&c, 1, 1, f);
fclose(f);
/* Kernel threads have an empty cmdline */
if (count <= 0)
return true;
/* Processes with argv[0][0] = '@' we ignore from the killing
* spree.
*
* http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons */
if (count == 1 && c == '@')
return true;
return false;
}
static void wait_for_children(int n_processes, sigset_t *mask) {
usec_t until;
assert(mask);
until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC;
for (;;) {
struct timespec ts;
int k;
usec_t n;
for (;;) {
pid_t pid = waitpid(-1, NULL, WNOHANG);
if (pid == 0)
break;
if (pid < 0 && errno == ECHILD)
return;
if (n_processes > 0)
if (--n_processes == 0)
return;
}
n = now(CLOCK_MONOTONIC);
if (n >= until)
return;
timespec_store(&ts, until - n);
if ((k = sigtimedwait(mask, NULL, &ts)) != SIGCHLD) {
if (k < 0 && errno != EAGAIN) {
log_error("sigtimedwait() failed: %m");
return;
}
if (k >= 0)
log_warning("sigtimedwait() returned unexpected signal.");
}
}
}
static int killall(int sig) {
DIR *dir;
struct dirent *d;
unsigned int n_processes = 0;
dir = opendir("/proc");
if (!dir)
return -errno;
while ((d = readdir(dir))) {
pid_t pid;
if (d->d_type != DT_DIR &&
d->d_type != DT_UNKNOWN)
continue;
if (parse_pid(d->d_name, &pid) < 0)
continue;
if (ignore_proc(pid))
continue;
if (kill(pid, sig) >= 0)
n_processes++;
else if (errno != ENOENT)
log_warning("Could not kill %d: %m", pid);
}
closedir(dir);
return n_processes;
}
void broadcast_signal(int sig) {
sigset_t mask, oldmask;
int n_processes;
assert_se(sigemptyset(&mask) == 0);
assert_se(sigaddset(&mask, SIGCHLD) == 0);
assert_se(sigprocmask(SIG_BLOCK, &mask, &oldmask) == 0);
if (kill(-1, SIGSTOP) < 0 && errno != ESRCH)
log_warning("kill(-1, SIGSTOP) failed: %m");
n_processes = killall(sig);
if (kill(-1, SIGCONT) < 0 && errno != ESRCH)
log_warning("kill(-1, SIGCONT) failed: %m");
if (n_processes <= 0)
goto finish;
wait_for_children(n_processes, &mask);
finish:
sigprocmask(SIG_SETMASK, &oldmask, NULL);
}

27
src/core/killall.h Normal file
View File

@ -0,0 +1,27 @@
/*-*- Mode: C; c-basic-offset: 8; indent-tabs-mode: nil -*-*/
#ifndef fookillallhfoo
#define fookillallhfoo
/***
This file is part of systemd.
Copyright 2012 Lennart Poettering
systemd 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.
systemd 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 systemd; If not, see <http://www.gnu.org/licenses/>.
***/
void broadcast_signal(int sig);
#endif

View File

@ -51,6 +51,7 @@
#include "path-util.h"
#include "switch-root.h"
#include "capability.h"
#include "killall.h"
#include "mount-setup.h"
#include "loopback-setup.h"
@ -1676,6 +1677,11 @@ finish:
make_console_stdio();
if (switch_root_dir) {
/* Kill all remaining processes from the initrd */
broadcast_signal(SIGTERM);
broadcast_signal(SIGKILL);
/* And switch root */
r = switch_root(switch_root_dir);
if (r < 0)
log_error("Failed to switch root, ignoring: %s", strerror(-r));

View File

@ -45,178 +45,8 @@
#include "virt.h"
#include "watchdog.h"
#define TIMEOUT_USEC (5 * USEC_PER_SEC)
#define FINALIZE_ATTEMPTS 50
static bool ignore_proc(pid_t pid) {
char buf[PATH_MAX];
FILE *f;
char c;
size_t count;
uid_t uid;
int r;
/* We are PID 1, let's not commit suicide */
if (pid == 1)
return true;
r = get_process_uid(pid, &uid);
if (r < 0)
return true; /* not really, but better safe than sorry */
/* Non-root processes otherwise are always subject to be killed */
if (uid != 0)
return false;
snprintf(buf, sizeof(buf), "/proc/%lu/cmdline", (unsigned long) pid);
char_array_0(buf);
f = fopen(buf, "re");
if (!f)
return true; /* not really, but has the desired effect */
count = fread(&c, 1, 1, f);
fclose(f);
/* Kernel threads have an empty cmdline */
if (count <= 0)
return true;
/* Processes with argv[0][0] = '@' we ignore from the killing
* spree.
*
* http://www.freedesktop.org/wiki/Software/systemd/RootStorageDaemons */
if (count == 1 && c == '@')
return true;
return false;
}
static int killall(int sign) {
DIR *dir;
struct dirent *d;
unsigned int n_processes = 0;
dir = opendir("/proc");
if (!dir)
return -errno;
while ((d = readdir(dir))) {
pid_t pid;
if (parse_pid(d->d_name, &pid) < 0)
continue;
if (ignore_proc(pid))
continue;
if (kill(pid, sign) == 0)
n_processes++;
else
log_warning("Could not kill %d: %m", pid);
}
closedir(dir);
return n_processes;
}
static void wait_for_children(int n_processes, sigset_t *mask) {
usec_t until;
assert(mask);
until = now(CLOCK_MONOTONIC) + TIMEOUT_USEC;
for (;;) {
struct timespec ts;
int k;
usec_t n;
for (;;) {
pid_t pid = waitpid(-1, NULL, WNOHANG);
if (pid == 0)
break;
if (pid < 0 && errno == ECHILD)
return;
if (n_processes > 0)
if (--n_processes == 0)
return;
}
n = now(CLOCK_MONOTONIC);
if (n >= until)
return;
timespec_store(&ts, until - n);
if ((k = sigtimedwait(mask, NULL, &ts)) != SIGCHLD) {
if (k < 0 && errno != EAGAIN) {
log_error("sigtimedwait() failed: %m");
return;
}
if (k >= 0)
log_warning("sigtimedwait() returned unexpected signal.");
}
}
}
static void send_signal(int sign) {
sigset_t mask, oldmask;
int n_processes;
assert_se(sigemptyset(&mask) == 0);
assert_se(sigaddset(&mask, SIGCHLD) == 0);
assert_se(sigprocmask(SIG_BLOCK, &mask, &oldmask) == 0);
if (kill(-1, SIGSTOP) < 0 && errno != ESRCH)
log_warning("kill(-1, SIGSTOP) failed: %m");
n_processes = killall(sign);
if (kill(-1, SIGCONT) < 0 && errno != ESRCH)
log_warning("kill(-1, SIGCONT) failed: %m");
if (n_processes <= 0)
goto finish;
wait_for_children(n_processes, &mask);
finish:
sigprocmask(SIG_SETMASK, &oldmask, NULL);
}
static void ultimate_send_signal(int sign) {
sigset_t mask, oldmask;
int r;
assert_se(sigemptyset(&mask) == 0);
assert_se(sigaddset(&mask, SIGCHLD) == 0);
assert_se(sigprocmask(SIG_BLOCK, &mask, &oldmask) == 0);
if (kill(-1, SIGSTOP) < 0 && errno != ESRCH)
log_warning("kill(-1, SIGSTOP) failed: %m");
r = kill(-1, sign);
if (r < 0 && errno != ESRCH)
log_warning("kill(-1, %s) failed: %m", signal_to_string(sign));
if (kill(-1, SIGCONT) < 0 && errno != ESRCH)
log_warning("kill(-1, SIGCONT) failed: %m");
if (r < 0)
goto finish;
wait_for_children(0, &mask);
finish:
sigprocmask(SIG_SETMASK, &oldmask, NULL);
}
static int prepare_new_root(void) {
static const char dirs[] =
"/run/initramfs/oldroot\0"
@ -267,6 +97,7 @@ static int prepare_new_root(void) {
}
static int pivot_to_new_root(void) {
if (chdir("/run/initramfs") < 0) {
log_error("Failed to change directory to /run/initramfs: %m");
return -errno;
@ -302,7 +133,7 @@ int main(int argc, char *argv[]) {
int cmd, r;
unsigned retries;
bool need_umount = true, need_swapoff = true, need_loop_detach = true, need_dm_detach = true;
bool killed_everbody = false, in_container, use_watchdog = false;
bool in_container, use_watchdog = false;
char *arguments[3];
log_parse_environment();
@ -345,10 +176,10 @@ int main(int argc, char *argv[]) {
mlockall(MCL_CURRENT|MCL_FUTURE);
log_info("Sending SIGTERM to remaining processes...");
send_signal(SIGTERM);
broadcast_signal(SIGTERM);
log_info("Sending SIGKILL to remaining processes...");
send_signal(SIGKILL);
broadcast_signal(SIGKILL);
if (in_container) {
need_swapoff = false;
@ -414,21 +245,10 @@ int main(int argc, char *argv[]) {
}
/* If in this iteration we didn't manage to
* unmount/deactivate anything, we either kill more
* processes, or simply give up */
* unmount/deactivate anything, we simply give up */
if (!changed) {
if (killed_everbody) {
/* Hmm, we already killed everybody,
* let's just give up */
log_error("Cannot finalize remaining file systems and devices, giving up.");
break;
}
log_warning("Cannot finalize remaining file systems and devices, trying to kill remaining processes.");
ultimate_send_signal(SIGTERM);
ultimate_send_signal(SIGKILL);
killed_everbody = true;
log_error("Cannot finalize remaining file systems and devices, giving up.");
break;
}
log_debug("Couldn't finalize remaining file systems and devices after %u retries, trying again.", retries+1);