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 <errno.h>
  28 #include <fcntl.h>
  29 #include <sched.h>
  30 #include <signal.h>
  31 #include <sys/param.h>
  32 #include <sys/stat.h>
  33 #include <sys/syscall.h>
  34 #include <sys/types.h>
  35 #include <unistd.h>
  36
  37 #include "config.h"
  38 #include "log.h"
  39 #include "memory_utils.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 #define __LXC_STACK_SIZE 4096
  57 pid_t lxc_clone(int (*fn)(void *), void *arg, int flags, int *pidfd)
  58 {
  59         pid_t ret;
  60         struct clone_arg clone_arg = {
  61             .fn = fn,
  62             .arg = arg,
  63         };
  64         void *stack;
  65
  66         stack = malloc(__LXC_STACK_SIZE);
  67         if (!stack) {
  68                 SYSERROR("Failed to allocate clone stack");
  69                 return -ENOMEM;
  70         }
  71
  72 #ifdef __ia64__
  73         ret = __clone2(fn, stack, __LXC_STACK_SIZE, flags | SIGCHLD, &clone_arg, pidfd);
  74 #else
  75         ret = clone(fn, stack + __LXC_STACK_SIZE, flags | SIGCHLD, &clone_arg, pidfd);
  76 #endif
  77         if (ret < 0)
  78                 SYSERROR("Failed to clone (%#x)", flags);
  79
  80         return ret;
  81 }
  82
  83 /* Leave the user namespace at the first position in the array of structs so
  84  * that we always attach to it first when iterating over the struct and using
  85  * setns() to switch namespaces. This especially affects lxc_attach(): Suppose
  86  * you cloned a new user namespace and mount namespace as an unprivileged user
  87  * on the host and want to setns() to the mount namespace. This requires you to
  88  * attach to the user namespace first otherwise the kernel will fail this check:
  89  *
  90  *        if (!ns_capable(mnt_ns->user_ns, CAP_SYS_ADMIN) ||
  91  *            !ns_capable(current_user_ns(), CAP_SYS_CHROOT) ||
  92  *            !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
  93  *            return -EPERM;
  94  *
  95  *    in
  96  *
  97  *        linux/fs/namespace.c:mntns_install().
  98  */
  99 const struct ns_info ns_info[LXC_NS_MAX] = {
 100         [LXC_NS_USER]    = { "user",   CLONE_NEWUSER,   "CLONE_NEWUSER",   "LXC_USER_NS"    },
 101         [LXC_NS_MNT]    =  { "mnt",    CLONE_NEWNS,     "CLONE_NEWNS",     "LXC_MNT_NS"     },
 102         [LXC_NS_PID]    =  { "pid",    CLONE_NEWPID,    "CLONE_NEWPID",    "LXC_PID_NS"     },
 103         [LXC_NS_UTS]    =  { "uts",    CLONE_NEWUTS,    "CLONE_NEWUTS",    "LXC_UTS_NS"     },
 104         [LXC_NS_IPC]    =  { "ipc",    CLONE_NEWIPC,    "CLONE_NEWIPC",    "LXC_IPC_NS"     },
 105         [LXC_NS_NET]    =  { "net",    CLONE_NEWNET,    "CLONE_NEWNET",    "LXC_NET_NS"     },
 106         [LXC_NS_CGROUP] =  { "cgroup", CLONE_NEWCGROUP, "CLONE_NEWCGROUP", "LXC_CGROUP_NS"  }
 107 };
 108
 109 int lxc_namespace_2_cloneflag(const char *namespace)
 110 {
 111         int i;
 112
 113         for (i = 0; i < LXC_NS_MAX; i++)
 114                 if (!strcasecmp(ns_info[i].proc_name, namespace))
 115                         return ns_info[i].clone_flag;
 116
 117         ERROR("Invalid namespace name \"%s\"", namespace);
 118         return -EINVAL;
 119 }
 120
 121 int lxc_namespace_2_ns_idx(const char *namespace)
 122 {
 123         int i;
 124
 125         for (i = 0; i < LXC_NS_MAX; i++)
 126                 if (!strcmp(ns_info[i].proc_name, namespace))
 127                         return i;
 128
 129         ERROR("Invalid namespace name \"%s\"", namespace);
 130         return -EINVAL;
 131 }
 132
 133 extern int lxc_namespace_2_std_identifiers(char *namespaces)
 134 {
 135         char **it;
 136         char *del;
 137
 138         /* The identifiers for namespaces used with lxc-attach and lxc-unshare
 139          * as given on the manpage do not align with the standard identifiers.
 140          * This affects network, mount, and uts namespaces. The standard identifiers
 141          * are: "mnt", "uts", and "net" whereas lxc-attach and lxc-unshare uses
 142          * "MOUNT", "UTSNAME", and "NETWORK". So let's use some cheap memmove()s
 143          * to replace them by their standard identifiers.
 144          * Let's illustrate this with an example:
 145          * Assume the string:
 146          *
 147          *      "IPC|MOUNT|PID"
 148          *
 149          * then we memmove()
 150          *
 151          *      dest: del + 1 == OUNT|PID
 152          *      src:  del + 3 == NT|PID
 153          */
 154         if (!namespaces)
 155                 return -1;
 156
 157         while ((del = strstr(namespaces, "MOUNT")))
 158                 memmove(del + 1, del + 3, strlen(del) - 2);
 159
 160         for (it = (char *[]){"NETWORK", "UTSNAME", NULL}; it && *it; it++)
 161                 while ((del = strstr(namespaces, *it)))
 162                         memmove(del + 3, del + 7, strlen(del) - 6);
 163
 164         return 0;
 165 }
 166
 167 int lxc_fill_namespace_flags(char *flaglist, int *flags)
 168 {
 169         char *token;
 170         int aflag;
 171
 172         if (!flaglist) {
 173                 ERROR("At least one namespace is needed.");
 174                 return -1;
 175         }
 176
 177         lxc_iterate_parts(token, flaglist, "|") {
 178                 aflag = lxc_namespace_2_cloneflag(token);
 179                 if (aflag < 0)
 180                         return -1;
 181
 182                 *flags |= aflag;
 183         }
 184
 185         return 0;
 186 }