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 #include <alloca.h>
  25 #include <errno.h>
  26 #include <fcntl.h>
  27 #include <sched.h>
  28 #include <signal.h>
  29 #include <unistd.h>
  30 #include <sys/param.h>
  31 #include <sys/stat.h>
  32 #include <sys/syscall.h>
  33 #include <sys/types.h>
  34
  35 #include "log.h"
  36 #include "namespace.h"
  37 #include "utils.h"
  38
  39 lxc_log_define(lxc_namespace, lxc);
  40
  41 struct clone_arg {
  42         int (*fn)(void *);
  43         void *arg;
  44 };
  45
  46 static int do_clone(void *arg)
  47 {
  48         struct clone_arg *clone_arg = arg;
  49         return clone_arg->fn(clone_arg->arg);
  50 }
  51
  52 pid_t lxc_clone(int (*fn)(void *), void *arg, int flags)
  53 {
  54         struct clone_arg clone_arg = {
  55                 .fn = fn,
  56                 .arg = arg,
  57         };
  58
  59         size_t stack_size = lxc_getpagesize();
  60         void *stack = alloca(stack_size);
  61         pid_t ret;
  62
  63 #ifdef __ia64__
  64         ret = __clone2(do_clone, stack, stack_size, flags | SIGCHLD, &clone_arg);
  65 #else
  66         ret = clone(do_clone, stack  + stack_size, flags | SIGCHLD, &clone_arg);
  67 #endif
  68         if (ret < 0)
  69                 ERROR("Failed to clone (%#x): %s.", flags, strerror(errno));
  70
  71         return ret;
  72 }
  73
  74 /**
  75  * This is based on raw_clone in systemd but adapted to our needs. This uses
  76  * copy on write semantics and doesn't pass a stack. CLONE_VM is tricky and
  77  * doesn't really matter to us so disallow it.
  78  *
  79  * The nice thing about this is that we get fork() behavior. That is
  80  * lxc_raw_clone() returns 0 in the child and the child pid in the parent.
  81  */
  82 pid_t lxc_raw_clone(unsigned long flags)
  83 {
  84
  85         /* These flags don't interest at all so we don't jump through any hoopes
  86          * of retrieving them and passing them to the kernel.
  87          */
  88         errno = EINVAL;
  89         if ((flags & (CLONE_VM | CLONE_PARENT_SETTID | CLONE_CHILD_SETTID |
  90                       CLONE_CHILD_CLEARTID | CLONE_SETTLS)))
  91                 return -EINVAL;
  92
  93 #if defined(__s390x__) || defined(__s390__) || defined(__CRIS__)
  94         /* On s390/s390x and cris the order of the first and second arguments
  95          * of the system call is reversed.
  96          */
  97         return (int)syscall(__NR_clone, NULL, flags | SIGCHLD);
  98 #elif defined(__sparc__) && defined(__arch64__)
  99         {
 100                 /**
 101                  * sparc64 always returns the other process id in %o0, and
 102                  * a boolean flag whether this is the child or the parent in
 103                  * %o1. Inline assembly is needed to get the flag returned
 104                  * in %o1.
 105                  */
 106                 int in_child;
 107                 int child_pid;
 108                 asm volatile("mov %2, %%g1\n\t"
 109                              "mov %3, %%o0\n\t"
 110                              "mov 0 , %%o1\n\t"
 111                              "t 0x6d\n\t"
 112                              "mov %%o1, %0\n\t"
 113                              "mov %%o0, %1"
 114                              : "=r"(in_child), "=r"(child_pid)
 115                              : "i"(__NR_clone), "r"(flags | SIGCHLD)
 116                              : "%o1", "%o0", "%g1");
 117                 if (in_child)
 118                         return 0;
 119                 else
 120                         return child_pid;
 121         }
 122 #elif defined(__ia64__)
 123         /* On ia64 the stack and stack size are passed as separate arguments. */
 124         return (int)syscall(__NR_clone, flags | SIGCHLD, NULL, 0);
 125 #else
 126         return (int)syscall(__NR_clone, flags | SIGCHLD, NULL);
 127 #endif
 128 }
 129
 130 /* Leave the user namespace at the first position in the array of structs so
 131  * that we always attach to it first when iterating over the struct and using
 132  * setns() to switch namespaces. This especially affects lxc_attach(): Suppose
 133  * you cloned a new user namespace and mount namespace as an unprivileged user
 134  * on the host and want to setns() to the mount namespace. This requires you to
 135  * attach to the user namespace first otherwise the kernel will fail this check:
 136  *
 137  *        if (!ns_capable(mnt_ns->user_ns, CAP_SYS_ADMIN) ||
 138  *            !ns_capable(current_user_ns(), CAP_SYS_CHROOT) ||
 139  *            !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
 140  *            return -EPERM;
 141  *
 142  *    in
 143  *
 144  *        linux/fs/namespace.c:mntns_install().
 145  */
 146 const struct ns_info ns_info[LXC_NS_MAX] = {
 147         [LXC_NS_USER]    = { "user",   CLONE_NEWUSER,   "CLONE_NEWUSER",   "LXC_USER_NS"    },
 148         [LXC_NS_MNT]    =  { "mnt",    CLONE_NEWNS,     "CLONE_NEWNS",     "LXC_MNT_NS"     },
 149         [LXC_NS_PID]    =  { "pid",    CLONE_NEWPID,    "CLONE_NEWPID",    "LXC_PID_NS"     },
 150         [LXC_NS_UTS]    =  { "uts",    CLONE_NEWUTS,    "CLONE_NEWUTS",    "LXC_UTS_NS"     },
 151         [LXC_NS_IPC]    =  { "ipc",    CLONE_NEWIPC,    "CLONE_NEWIPC",    "LXC_IPC_NS"     },
 152         [LXC_NS_NET]    =  { "net",    CLONE_NEWNET,    "CLONE_NEWNET",    "LXC_NET_NS"     },
 153         [LXC_NS_CGROUP] =  { "cgroup", CLONE_NEWCGROUP, "CLONE_NEWCGROUP", "LXC_CGROUP_NS"  }
 154 };
 155
 156 int lxc_namespace_2_cloneflag(const char *namespace)
 157 {
 158         int i;
 159         for (i = 0; i < LXC_NS_MAX; i++)
 160                 if (!strcasecmp(ns_info[i].proc_name, namespace))
 161                         return ns_info[i].clone_flag;
 162
 163         ERROR("Invalid namespace name \"%s\"", namespace);
 164         return -EINVAL;
 165 }
 166
 167 int lxc_namespace_2_ns_idx(const char *namespace)
 168 {
 169         int i;
 170         for (i = 0; i < LXC_NS_MAX; i++)
 171                 if (!strcmp(ns_info[i].proc_name, namespace))
 172                         return i;
 173
 174         ERROR("Invalid namespace name \"%s\"", namespace);
 175         return -EINVAL;
 176 }
 177
 178 int lxc_fill_namespace_flags(char *flaglist, int *flags)
 179 {
 180         char *token, *saveptr = NULL;
 181         int aflag;
 182
 183         if (!flaglist) {
 184                 ERROR("At least one namespace is needed.");
 185                 return -1;
 186         }
 187
 188         token = strtok_r(flaglist, "|", &saveptr);
 189         while (token) {
 190
 191                 aflag = lxc_namespace_2_cloneflag(token);
 192                 if (aflag < 0)
 193                         return -1;
 194
 195                 *flags |= aflag;
 196
 197                 token = strtok_r(NULL, "|", &saveptr);
 198         }
 199         return 0;
 200 }