chiark / gitweb /
build: basic autoconfization
[elogind.git] / execute.c
index 1ca91fddd78f8b1fdea5e431b4b00e968afa62c0..7bdf1b13b8ec789d259dfdf4caebbd760b5e7ce6 100644 (file)
--- a/execute.c
+++ b/execute.c
@@ -1,5 +1,26 @@
 /*-*- Mode: C; c-basic-offset: 8 -*-*/
 
+/***
+  This file is part of systemd.
+
+  Copyright 2010 Lennart Poettering
+
+  systemd is free software; you can redistribute it and/or modify it
+  under the terms of the GNU General Public License as published by
+  the Free Software Foundation; either version 2 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
+  General Public License for more details.
+
+  You should have received a copy of the GNU General Public License
+  along with systemd; If not, see <http://www.gnu.org/licenses/>.
+***/
+
+#pragma GCC diagnostic ignored "-Wattributes"
+
 #include <assert.h>
 #include <dirent.h>
 #include <errno.h>
 #include <unistd.h>
 #include <string.h>
 #include <signal.h>
+#include <sys/socket.h>
+#include <sys/un.h>
+#include <sys/prctl.h>
+#include <linux/sched.h>
 
 #include "execute.h"
 #include "strv.h"
 #include "macro.h"
 #include "util.h"
 #include "log.h"
+#include "ioprio.h"
+#include "securebits.h"
 
 static int close_fds(int except[], unsigned n_except) {
         DIR *d;
@@ -139,6 +166,130 @@ static int flags_fds(int fds[], unsigned n_fds) {
         return 0;
 }
 
+static int replace_null_fd(int fd, int flags) {
+        int nfd;
+        assert(fd >= 0);
+
+        close_nointr(fd);
+
+        if ((nfd = open("/dev/null", flags|O_NOCTTY)) < 0)
+                return -errno;
+
+        if (nfd != fd) {
+                close_nointr_nofail(nfd);
+                return -EIO;
+        }
+
+        return 0;
+}
+
+static int setup_output(const ExecContext *context, const char *ident) {
+        int r;
+
+        assert(context);
+
+        switch (context->output) {
+
+        case EXEC_OUTPUT_CONSOLE:
+                return 0;
+
+        case EXEC_OUTPUT_NULL:
+
+                if ((r = replace_null_fd(STDOUT_FILENO, O_WRONLY)) < 0 ||
+                    (r = replace_null_fd(STDERR_FILENO, O_WRONLY)) < 0)
+                        return r;
+
+                return 0;
+
+        case EXEC_OUTPUT_KERNEL:
+        case EXEC_OUTPUT_SYSLOG: {
+
+                int fd;
+                union {
+                        struct sockaddr sa;
+                        struct sockaddr_un un;
+                } sa;
+
+                close_nointr(STDOUT_FILENO);
+                close_nointr(STDERR_FILENO);
+
+                if ((fd = socket(AF_UNIX, SOCK_STREAM, 0)) < 0)
+                        return -errno;
+
+                if (fd != STDOUT_FILENO) {
+                        close_nointr_nofail(fd);
+                        return -EIO;
+                }
+
+                zero(sa);
+                sa.sa.sa_family = AF_UNIX;
+                strncpy(sa.un.sun_path+1, LOGGER_SOCKET, sizeof(sa.un.sun_path)-1);
+
+                if (connect(fd, &sa.sa, sizeof(sa)) < 0) {
+                        close_nointr_nofail(fd);
+                        return -errno;
+                }
+
+                if (shutdown(fd, SHUT_RD) < 0) {
+                        close_nointr_nofail(fd);
+                        return -errno;
+                }
+
+                if ((fd = dup(fd)) < 0) {
+                        close_nointr_nofail(fd);
+                        return -errno;
+                }
+
+                if (fd != STDERR_FILENO) {
+                        close_nointr_nofail(fd);
+                        return -EIO;
+                }
+
+                /* We speak a very simple protocol between log server
+                 * and client: one line for the log destination (kmsg
+                 * or syslog), followed by the priority field,
+                 * followed by the process name. Since we replaced
+                 * stdin/stderr we simple use stdio to write to
+                 * it. Note that we use stderr, to minimize buffer
+                 * flushing issues. */
+
+                fprintf(stderr,
+                        "%s\n"
+                        "%i\n"
+                        "%s\n",
+                        context->output == EXEC_OUTPUT_KERNEL ? "kmsg" : "syslog",
+                        context->syslog_priority,
+                        context->syslog_identifier ? context->syslog_identifier : ident);
+
+                return 0;
+        }
+
+        default:
+                assert_not_reached("Unknown output type");
+        }
+}
+
+static int setup_input(const ExecContext *context) {
+        int r;
+
+        assert(context);
+
+        switch (context->input) {
+
+        case EXEC_INPUT_CONSOLE:
+                return 0;
+
+        case EXEC_INPUT_NULL:
+                if ((r = replace_null_fd(STDIN_FILENO, O_RDONLY)) < 0)
+                        return r;
+
+                return 0;
+
+        default:
+                assert_not_reached("Unknown input type");
+        }
+}
+
 int exec_spawn(const ExecCommand *command, const ExecContext *context, int *fds, unsigned n_fds, pid_t *ret) {
         pid_t pid;
 
@@ -155,7 +306,6 @@ int exec_spawn(const ExecCommand *command, const ExecContext *context, int *fds,
         if (pid == 0) {
                 char **e, **f = NULL;
                 int i, r;
-                char t[16];
                 sigset_t ss;
 
                 /* child */
@@ -166,26 +316,83 @@ int exec_spawn(const ExecCommand *command, const ExecContext *context, int *fds,
                         goto fail;
                 }
 
+                if (setpgid(0, 0) < 0) {
+                        r = EXIT_PGID;
+                        goto fail;
+                }
+
                 umask(context->umask);
 
-                if (chdir(context->directory ? context->directory : "/") < 0) {
-                        r = EXIT_CHDIR;
+                if (setup_input(context) < 0) {
+                        r = EXIT_INPUT;
                         goto fail;
                 }
 
-                snprintf(t, sizeof(t), "%i", context->oom_adjust);
-                char_array_0(t);
-
-                if (write_one_line_file("/proc/self/oom_adj", t) < 0) {
-                        r = EXIT_OOM_ADJUST;
+                if (setup_output(context, file_name_from_path(command->path)) < 0) {
+                        r = EXIT_OUTPUT;
                         goto fail;
                 }
 
-                if (setpriority(PRIO_PROCESS, 0, context->nice) < 0) {
-                        r = EXIT_NICE;
+                if (context->oom_adjust_set) {
+                        char t[16];
+
+                        snprintf(t, sizeof(t), "%i", context->oom_adjust);
+                        char_array_0(t);
+
+                        if (write_one_line_file("/proc/self/oom_adj", t) < 0) {
+                                r = EXIT_OOM_ADJUST;
+                                goto fail;
+                        }
+                }
+
+                if (context->root_directory)
+                        if (chroot(context->root_directory) < 0) {
+                                r = EXIT_CHROOT;
+                                goto fail;
+                        }
+
+                if (chdir(context->working_directory ? context->working_directory : "/") < 0) {
+                        r = EXIT_CHDIR;
                         goto fail;
                 }
 
+                if (context->nice_set)
+                        if (setpriority(PRIO_PROCESS, 0, context->nice) < 0) {
+                                r = EXIT_NICE;
+                                goto fail;
+                        }
+
+                if (context->cpu_sched_set) {
+                        struct sched_param param;
+
+                        zero(param);
+                        param.sched_priority = context->cpu_sched_priority;
+
+                        if (sched_setscheduler(0, context->cpu_sched_policy |
+                                               (context->cpu_sched_reset_on_fork ? SCHED_RESET_ON_FORK : 0), &param) < 0) {
+                                r = EXIT_SETSCHEDULER;
+                                goto fail;
+                        }
+                }
+
+                if (context->cpu_affinity_set)
+                        if (sched_setaffinity(0, sizeof(context->cpu_affinity), &context->cpu_affinity) < 0) {
+                                r = EXIT_CPUAFFINITY;
+                                goto fail;
+                        }
+
+                if (context->ioprio_set)
+                        if (ioprio_set(IOPRIO_WHO_PROCESS, 0, context->ioprio) < 0) {
+                                r = EXIT_IOPRIO;
+                                goto fail;
+                        }
+
+                if (context->timer_slack_ns_set)
+                        if (prctl(PR_SET_TIMERSLACK, context->timer_slack_ns_set) < 0) {
+                                r = EXIT_TIMERSLACK;
+                                goto fail;
+                        }
+
                 if (close_fds(fds, n_fds) < 0 ||
                     shift_fds(fds, n_fds) < 0 ||
                     flags_fds(fds, n_fds) < 0) {
@@ -203,6 +410,13 @@ int exec_spawn(const ExecCommand *command, const ExecContext *context, int *fds,
                         }
                 }
 
+                if (context->secure_bits) {
+                        if (prctl(PR_SET_SECUREBITS, context->secure_bits) < 0) {
+                                r = EXIT_SECUREBITS;
+                                goto fail;
+                        }
+                }
+
                 if (n_fds > 0) {
                         char a[64], b[64];
                         char *listen_env[3] = {
@@ -248,9 +462,24 @@ void exec_context_init(ExecContext *c) {
         assert(c);
 
         c->umask = 0002;
-        cap_clear(c->capabilities);
         c->oom_adjust = 0;
+        c->oom_adjust_set = false;
         c->nice = 0;
+        c->nice_set = false;
+        c->ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 0);
+        c->ioprio_set = false;
+        c->cpu_sched_policy = SCHED_OTHER;
+        c->cpu_sched_priority = 0;
+        c->cpu_sched_set = false;
+        CPU_ZERO(&c->cpu_affinity);
+        c->cpu_affinity_set = false;
+
+        c->input = 0;
+        c->output = 0;
+        c->syslog_priority = LOG_DAEMON|LOG_INFO;
+
+        c->secure_bits = 0;
+        c->capability_bounding_set_drop = 0;
 }
 
 void exec_context_done(ExecContext *c) {
@@ -266,8 +495,13 @@ void exec_context_done(ExecContext *c) {
                 c->rlimit[l] = NULL;
         }
 
-        free(c->directory);
-        c->directory = NULL;
+        free(c->working_directory);
+        c->working_directory = NULL;
+        free(c->root_directory);
+        c->root_directory = NULL;
+
+        free(c->syslog_identifier);
+        c->syslog_identifier = NULL;
 
         free(c->user);
         c->user = NULL;
@@ -277,6 +511,11 @@ void exec_context_done(ExecContext *c) {
 
         strv_free(c->supplementary_groups);
         c->supplementary_groups = NULL;
+
+        if (c->capabilities) {
+                cap_free(c->capabilities);
+                c->capabilities = NULL;
+        }
 }
 
 void exec_command_free_list(ExecCommand *c) {
@@ -301,6 +540,9 @@ void exec_command_free_array(ExecCommand **c, unsigned n) {
 }
 
 void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) {
+        char ** e;
+        unsigned i;
+
         assert(c);
         assert(f);
 
@@ -308,14 +550,121 @@ void exec_context_dump(ExecContext *c, FILE* f, const char *prefix) {
                 prefix = "";
 
         fprintf(f,
-                "%sUmask: %04o\n"
-                "%sDirectory: %s\n"
-                "%sNice: %i\n"
-                "%sOOMAdjust: %i\n",
+                "%sUMask: %04o\n"
+                "%sWorkingDirectory: %s\n"
+                "%sRootDirectory: %s\n",
                 prefix, c->umask,
-                prefix, c->directory ? c->directory : "/",
-                prefix, c->nice,
-                prefix, c->oom_adjust);
+                prefix, c->working_directory ? c->working_directory : "/",
+                prefix, c->root_directory ? c->root_directory : "/");
+
+        if (c->environment)
+                for (e = c->environment; *e; e++)
+                        fprintf(f, "%sEnvironment: %s\n", prefix, *e);
+
+        if (c->nice_set)
+                fprintf(f,
+                        "%sNice: %i\n",
+                        prefix, c->nice);
+
+        if (c->oom_adjust_set)
+                fprintf(f,
+                        "%sOOMAdjust: %i\n",
+                        prefix, c->oom_adjust);
+
+        for (i = 0; i < RLIM_NLIMITS; i++)
+                if (c->rlimit[i])
+                        fprintf(f, "%s%s: %llu\n", prefix, rlimit_to_string(i), (unsigned long long) c->rlimit[i]->rlim_max);
+
+        if (c->ioprio_set)
+                fprintf(f,
+                        "%sIOSchedulingClass: %s\n"
+                        "%sIOPriority: %i\n",
+                        prefix, ioprio_class_to_string(IOPRIO_PRIO_CLASS(c->ioprio)),
+                        prefix, (int) IOPRIO_PRIO_DATA(c->ioprio));
+
+        if (c->cpu_sched_set)
+                fprintf(f,
+                        "%sCPUSchedulingPolicy: %s\n"
+                        "%sCPUSchedulingPriority: %i\n"
+                        "%sCPUSchedulingResetOnFork: %s\n",
+                        prefix, sched_policy_to_string(c->cpu_sched_policy),
+                        prefix, c->cpu_sched_priority,
+                        prefix, yes_no(c->cpu_sched_reset_on_fork));
+
+        if (c->cpu_affinity_set) {
+                fprintf(f, "%sCPUAffinity:", prefix);
+                for (i = 0; i < CPU_SETSIZE; i++)
+                        if (CPU_ISSET(i, &c->cpu_affinity))
+                                fprintf(f, " %i", i);
+                fputs("\n", f);
+        }
+
+        if (c->timer_slack_ns_set)
+                fprintf(f, "%sTimerSlackNS: %lu\n", prefix, c->timer_slack_ns);
+
+        fprintf(f,
+                "%sInput: %s\n"
+                "%sOutput: %s\n",
+                prefix, exec_input_to_string(c->input),
+                prefix, exec_output_to_string(c->output));
+
+        if (c->output == EXEC_OUTPUT_SYSLOG || c->output == EXEC_OUTPUT_KERNEL)
+                fprintf(f,
+                        "%sSyslogFacility: %s\n"
+                        "%sSyslogLevel: %s\n",
+                        prefix, log_facility_to_string(LOG_FAC(c->syslog_priority)),
+                        prefix, log_level_to_string(LOG_PRI(c->syslog_priority)));
+
+        if (c->capabilities) {
+                char *t;
+                if ((t = cap_to_text(c->capabilities, NULL))) {
+                        fprintf(f, "%sCapabilities: %s\n",
+                                prefix, t);
+                        cap_free(t);
+                }
+        }
+
+        if (c->secure_bits)
+                fprintf(f, "%sSecure Bits:%s%s%s%s%s%s\n",
+                        prefix,
+                        (c->secure_bits & SECURE_KEEP_CAPS) ? " keep-caps" : "",
+                        (c->secure_bits & SECURE_KEEP_CAPS_LOCKED) ? " keep-caps-locked" : "",
+                        (c->secure_bits & SECURE_NO_SETUID_FIXUP) ? " no-setuid-fixup" : "",
+                        (c->secure_bits & SECURE_NO_SETUID_FIXUP_LOCKED) ? " no-setuid-fixup-locked" : "",
+                        (c->secure_bits & SECURE_NOROOT) ? " noroot" : "",
+                        (c->secure_bits & SECURE_NOROOT_LOCKED) ? "noroot-locked" : "");
+
+        if (c->capability_bounding_set_drop) {
+                fprintf(f, "%sCapabilityBoundingSetDrop:", prefix);
+
+                for (i = 0; i <= CAP_LAST_CAP; i++)
+                        if (c->capability_bounding_set_drop & (1 << i)) {
+                                char *t;
+
+                                if ((t = cap_to_name(i))) {
+                                        fprintf(f, " %s", t);
+                                        free(t);
+                                }
+                        }
+
+                fputs("\n", f);
+        }
+
+        if (c->user)
+                fprintf(f, "%sUser: %s", prefix, c->user);
+        if (c->group)
+                fprintf(f, "%sGroup: %s", prefix, c->group);
+
+        if (c->supplementary_groups) {
+                char **g;
+
+                fprintf(f, "%sSupplementaryGroups:", prefix);
+
+                STRV_FOREACH(g, c->supplementary_groups)
+                        fprintf(f, " %s", *g);
+
+                fputs("\n", f);
+        }
 }
 
 void exec_status_fill(ExecStatus *s, pid_t pid, int code, int status) {
@@ -394,3 +743,19 @@ void exec_command_dump_list(ExecCommand *c, FILE *f, const char *prefix) {
         LIST_FOREACH(command, c, c)
                 exec_command_dump(c, f, prefix);
 }
+
+static const char* const exec_output_table[_EXEC_OUTPUT_MAX] = {
+        [EXEC_OUTPUT_CONSOLE] = "console",
+        [EXEC_OUTPUT_NULL] = "null",
+        [EXEC_OUTPUT_SYSLOG] = "syslog",
+        [EXEC_OUTPUT_KERNEL] = "kernel"
+};
+
+DEFINE_STRING_TABLE_LOOKUP(exec_output, ExecOutput);
+
+static const char* const exec_input_table[_EXEC_INPUT_MAX] = {
+        [EXEC_INPUT_NULL] = "null",
+        [EXEC_INPUT_CONSOLE] = "console"
+};
+
+DEFINE_STRING_TABLE_LOOKUP(exec_input, ExecInput);