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 #ifndef _GNU_SOURCE
  25 #define _GNU_SOURCE 1
  26 #endif
  27 #include <alloca.h>
  28 #include <errno.h>
  29 #include <fcntl.h>
  30 #include <sched.h>
  31 #include <signal.h>
  32 #include <sys/param.h>
  33 #include <sys/stat.h>
  34 #include <sys/syscall.h>
  35 #include <sys/types.h>
  36 #include <unistd.h>
  37
  38 #include "config.h"
  39 #include "log.h"
  40 #include "namespace.h"
  41 #include "utils.h"
  42
  43 lxc_log_define(namespace, lxc);
  44
  45 struct clone_arg {
  46         int (*fn)(void *);
  47         void *arg;
  48 };
  49
  50 static int do_clone(void *arg)
  51 {
  52         struct clone_arg *clone_arg = arg;
  53         return clone_arg->fn(clone_arg->arg);
  54 }
  55
  56 pid_t lxc_clone(int (*fn)(void *), void *arg, int flags)
  57 {
  58         struct clone_arg clone_arg = {
  59                 .fn = fn,
  60                 .arg = arg,
  61         };
  62
  63         size_t stack_size = lxc_getpagesize();
  64         void *stack = alloca(stack_size);
  65         pid_t ret;
  66
  67 #ifdef __ia64__
  68         ret = __clone2(do_clone, stack, stack_size, flags | SIGCHLD, &clone_arg);
  69 #else
  70         ret = clone(do_clone, stack + stack_size, flags | SIGCHLD, &clone_arg);
  71 #endif
  72         if (ret < 0)
  73                 SYSERROR("Failed to clone (%#x)", flags);
  74
  75         return ret;
  76 }
  77
  78 /* Leave the user namespace at the first position in the array of structs so
  79  * that we always attach to it first when iterating over the struct and using
  80  * setns() to switch namespaces. This especially affects lxc_attach(): Suppose
  81  * you cloned a new user namespace and mount namespace as an unprivileged user
  82  * on the host and want to setns() to the mount namespace. This requires you to
  83  * attach to the user namespace first otherwise the kernel will fail this check:
  84  *
  85  *        if (!ns_capable(mnt_ns->user_ns, CAP_SYS_ADMIN) ||
  86  *            !ns_capable(current_user_ns(), CAP_SYS_CHROOT) ||
  87  *            !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
  88  *            return -EPERM;
  89  *
  90  *    in
  91  *
  92  *        linux/fs/namespace.c:mntns_install().
  93  */
  94 const struct ns_info ns_info[LXC_NS_MAX] = {
  95         [LXC_NS_USER]    = { "user",   CLONE_NEWUSER,   "CLONE_NEWUSER",   "LXC_USER_NS"    },
  96         [LXC_NS_MNT]    =  { "mnt",    CLONE_NEWNS,     "CLONE_NEWNS",     "LXC_MNT_NS"     },
  97         [LXC_NS_PID]    =  { "pid",    CLONE_NEWPID,    "CLONE_NEWPID",    "LXC_PID_NS"     },
  98         [LXC_NS_UTS]    =  { "uts",    CLONE_NEWUTS,    "CLONE_NEWUTS",    "LXC_UTS_NS"     },
  99         [LXC_NS_IPC]    =  { "ipc",    CLONE_NEWIPC,    "CLONE_NEWIPC",    "LXC_IPC_NS"     },
 100         [LXC_NS_NET]    =  { "net",    CLONE_NEWNET,    "CLONE_NEWNET",    "LXC_NET_NS"     },
 101         [LXC_NS_CGROUP] =  { "cgroup", CLONE_NEWCGROUP, "CLONE_NEWCGROUP", "LXC_CGROUP_NS"  }
 102 };
 103
 104 int lxc_namespace_2_cloneflag(const char *namespace)
 105 {
 106         int i;
 107
 108         for (i = 0; i < LXC_NS_MAX; i++)
 109                 if (!strcasecmp(ns_info[i].proc_name, namespace))
 110                         return ns_info[i].clone_flag;
 111
 112         ERROR("Invalid namespace name \"%s\"", namespace);
 113         return -EINVAL;
 114 }
 115
 116 int lxc_namespace_2_ns_idx(const char *namespace)
 117 {
 118         int i;
 119
 120         for (i = 0; i < LXC_NS_MAX; i++)
 121                 if (!strcmp(ns_info[i].proc_name, namespace))
 122                         return i;
 123
 124         ERROR("Invalid namespace name \"%s\"", namespace);
 125         return -EINVAL;
 126 }
 127
 128 extern int lxc_namespace_2_std_identifiers(char *namespaces)
 129 {
 130         char **it;
 131         char *del;
 132
 133         /* The identifiers for namespaces used with lxc-attach and lxc-unshare
 134          * as given on the manpage do not align with the standard identifiers.
 135          * This affects network, mount, and uts namespaces. The standard identifiers
 136          * are: "mnt", "uts", and "net" whereas lxc-attach and lxc-unshare uses
 137          * "MOUNT", "UTSNAME", and "NETWORK". So let's use some cheap memmove()s
 138          * to replace them by their standard identifiers.
 139          * Let's illustrate this with an example:
 140          * Assume the string:
 141          *
 142          *      "IPC|MOUNT|PID"
 143          *
 144          * then we memmove()
 145          *
 146          *      dest: del + 1 == OUNT|PID
 147          *      src:  del + 3 == NT|PID
 148          */
 149         if (!namespaces)
 150                 return -1;
 151
 152         while ((del = strstr(namespaces, "MOUNT")))
 153                 memmove(del + 1, del + 3, strlen(del) - 2);
 154
 155         for (it = (char *[]){"NETWORK", "UTSNAME", NULL}; it && *it; it++)
 156                 while ((del = strstr(namespaces, *it)))
 157                         memmove(del + 3, del + 7, strlen(del) - 6);
 158
 159         return 0;
 160 }
 161
 162 int lxc_fill_namespace_flags(char *flaglist, int *flags)
 163 {
 164         char *token;
 165         int aflag;
 166
 167         if (!flaglist) {
 168                 ERROR("At least one namespace is needed.");
 169                 return -1;
 170         }
 171
 172         lxc_iterate_parts(token, flaglist, "|") {
 173                 aflag = lxc_namespace_2_cloneflag(token);
 174                 if (aflag < 0)
 175                         return -1;
 176
 177                 *flags |= aflag;
 178         }
 179
 180         return 0;
 181 }