src/lxc/namespace.c

   1 /*
   2  * lxc: linux Container library
   3  *
   4  * (C) Copyright IBM Corp. 2007, 2009
   5  *
   6  * Authors:
   7  * Daniel Lezcano <daniel.lezcano at free.fr>
   8  *
   9  * This library is free software; you can redistribute it and/or
  10  * modify it under the terms of the GNU Lesser General Public
  11  * License as published by the Free Software Foundation; either
  12  * version 2.1 of the License, or (at your option) any later version.
  13  *
  14  * This library is distributed in the hope that it will be useful,
  15  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  16  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  17  * Lesser General Public License for more details.
  18  *
  19  * You should have received a copy of the GNU Lesser General Public
  20  * License along with this library; if not, write to the Free Software
  21  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  22  */
  23
  24 #define _GNU_SOURCE
  25 #include <alloca.h>
  26 #include <errno.h>
  27 #include <fcntl.h>
  28 #include <sched.h>
  29 #include <signal.h>
  30 #include <unistd.h>
  31 #include <sys/param.h>
  32 #include <sys/stat.h>
  33 #include <sys/syscall.h>
  34 #include <sys/types.h>
  35
  36 #include "log.h"
  37 #include "namespace.h"
  38 #include "utils.h"
  39
  40 lxc_log_define(namespace, lxc);
  41
  42 struct clone_arg {
  43         int (*fn)(void *);
  44         void *arg;
  45 };
  46
  47 static int do_clone(void *arg)
  48 {
  49         struct clone_arg *clone_arg = arg;
  50         return clone_arg->fn(clone_arg->arg);
  51 }
  52
  53 pid_t lxc_clone(int (*fn)(void *), void *arg, int flags)
  54 {
  55         struct clone_arg clone_arg = {
  56                 .fn = fn,
  57                 .arg = arg,
  58         };
  59
  60         size_t stack_size = lxc_getpagesize();
  61         void *stack = alloca(stack_size);
  62         pid_t ret;
  63
  64 #ifdef __ia64__
  65         ret = __clone2(do_clone, stack, stack_size, flags | SIGCHLD, &clone_arg);
  66 #else
  67         ret = clone(do_clone, stack  + stack_size, flags | SIGCHLD, &clone_arg);
  68 #endif
  69         if (ret < 0)
  70                 SYSERROR("Failed to clone (%#x)", flags);
  71
  72         return ret;
  73 }
  74
  75 /**
  76  * This is based on raw_clone in systemd but adapted to our needs. This uses
  77  * copy on write semantics and doesn't pass a stack. CLONE_VM is tricky and
  78  * doesn't really matter to us so disallow it.
  79  *
  80  * The nice thing about this is that we get fork() behavior. That is
  81  * lxc_raw_clone() returns 0 in the child and the child pid in the parent.
  82  */
  83 pid_t lxc_raw_clone(unsigned long flags)
  84 {
  85
  86         /* These flags don't interest at all so we don't jump through any hoopes
  87          * of retrieving them and passing them to the kernel.
  88          */
  89         errno = EINVAL;
  90         if ((flags & (CLONE_VM | CLONE_PARENT_SETTID | CLONE_CHILD_SETTID |
  91                       CLONE_CHILD_CLEARTID | CLONE_SETTLS)))
  92                 return -EINVAL;
  93
  94 #if defined(__s390x__) || defined(__s390__) || defined(__CRIS__)
  95         /* On s390/s390x and cris the order of the first and second arguments
  96          * of the system call is reversed.
  97          */
  98         return (int)syscall(__NR_clone, NULL, flags | SIGCHLD);
  99 #elif defined(__sparc__) && defined(__arch64__)
 100         {
 101                 /**
 102                  * sparc64 always returns the other process id in %o0, and
 103                  * a boolean flag whether this is the child or the parent in
 104                  * %o1. Inline assembly is needed to get the flag returned
 105                  * in %o1.
 106                  */
 107                 int in_child;
 108                 int child_pid;
 109                 asm volatile("mov %2, %%g1\n\t"
 110                              "mov %3, %%o0\n\t"
 111                              "mov 0 , %%o1\n\t"
 112                              "t 0x6d\n\t"
 113                              "mov %%o1, %0\n\t"
 114                              "mov %%o0, %1"
 115                              : "=r"(in_child), "=r"(child_pid)
 116                              : "i"(__NR_clone), "r"(flags | SIGCHLD)
 117                              : "%o1", "%o0", "%g1");
 118                 if (in_child)
 119                         return 0;
 120                 else
 121                         return child_pid;
 122         }
 123 #elif defined(__ia64__)
 124         /* On ia64 the stack and stack size are passed as separate arguments. */
 125         return (int)syscall(__NR_clone, flags | SIGCHLD, NULL, 0);
 126 #else
 127         return (int)syscall(__NR_clone, flags | SIGCHLD, NULL);
 128 #endif
 129 }
 130
 131 pid_t lxc_raw_clone_cb(int (*fn)(void *), void *args, unsigned long flags)
 132 {
 133         pid_t pid;
 134
 135         pid = lxc_raw_clone(flags);
 136         if (pid < 0)
 137                 return -1;
 138
 139         /* exit() is not thread-safe and might mess with the parent's signal
 140          * handlers and other stuff when exec() fails.
 141          */
 142         if (pid == 0)
 143                 _exit(fn(args));
 144
 145         return pid;
 146 }
 147
 148 /* Leave the user namespace at the first position in the array of structs so
 149  * that we always attach to it first when iterating over the struct and using
 150  * setns() to switch namespaces. This especially affects lxc_attach(): Suppose
 151  * you cloned a new user namespace and mount namespace as an unprivileged user
 152  * on the host and want to setns() to the mount namespace. This requires you to
 153  * attach to the user namespace first otherwise the kernel will fail this check:
 154  *
 155  *        if (!ns_capable(mnt_ns->user_ns, CAP_SYS_ADMIN) ||
 156  *            !ns_capable(current_user_ns(), CAP_SYS_CHROOT) ||
 157  *            !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
 158  *            return -EPERM;
 159  *
 160  *    in
 161  *
 162  *        linux/fs/namespace.c:mntns_install().
 163  */
 164 const struct ns_info ns_info[LXC_NS_MAX] = {
 165         [LXC_NS_USER]    = { "user",   CLONE_NEWUSER,   "CLONE_NEWUSER",   "LXC_USER_NS"    },
 166         [LXC_NS_MNT]    =  { "mnt",    CLONE_NEWNS,     "CLONE_NEWNS",     "LXC_MNT_NS"     },
 167         [LXC_NS_PID]    =  { "pid",    CLONE_NEWPID,    "CLONE_NEWPID",    "LXC_PID_NS"     },
 168         [LXC_NS_UTS]    =  { "uts",    CLONE_NEWUTS,    "CLONE_NEWUTS",    "LXC_UTS_NS"     },
 169         [LXC_NS_IPC]    =  { "ipc",    CLONE_NEWIPC,    "CLONE_NEWIPC",    "LXC_IPC_NS"     },
 170         [LXC_NS_NET]    =  { "net",    CLONE_NEWNET,    "CLONE_NEWNET",    "LXC_NET_NS"     },
 171         [LXC_NS_CGROUP] =  { "cgroup", CLONE_NEWCGROUP, "CLONE_NEWCGROUP", "LXC_CGROUP_NS"  }
 172 };
 173
 174 int lxc_namespace_2_cloneflag(const char *namespace)
 175 {
 176         int i;
 177         for (i = 0; i < LXC_NS_MAX; i++)
 178                 if (!strcasecmp(ns_info[i].proc_name, namespace))
 179                         return ns_info[i].clone_flag;
 180
 181         ERROR("Invalid namespace name \"%s\"", namespace);
 182         return -EINVAL;
 183 }
 184
 185 int lxc_namespace_2_ns_idx(const char *namespace)
 186 {
 187         int i;
 188         for (i = 0; i < LXC_NS_MAX; i++)
 189                 if (!strcmp(ns_info[i].proc_name, namespace))
 190                         return i;
 191
 192         ERROR("Invalid namespace name \"%s\"", namespace);
 193         return -EINVAL;
 194 }
 195
 196 int lxc_fill_namespace_flags(char *flaglist, int *flags)
 197 {
 198         char *token, *saveptr = NULL;
 199         int aflag;
 200
 201         if (!flaglist) {
 202                 ERROR("At least one namespace is needed.");
 203                 return -1;
 204         }
 205
 206         token = strtok_r(flaglist, "|", &saveptr);
 207         while (token) {
 208
 209                 aflag = lxc_namespace_2_cloneflag(token);
 210                 if (aflag < 0)
 211                         return -1;
 212
 213                 *flags |= aflag;
 214
 215                 token = strtok_r(NULL, "|", &saveptr);
 216         }
 217         return 0;
 218 }