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 /*
  46  * Let's use the "standard stack limit" (i.e. glibc thread size default) for
  47  * stack sizes: 8MB.
  48  */
  49 #define __LXC_STACK_SIZE (8 * 1024 * 1024)
  50 pid_t lxc_clone(int (*fn)(void *), void *arg, int flags, int *pidfd)
  51 {
  52         pid_t ret;
  53         void *stack;
  54
  55         stack = malloc(__LXC_STACK_SIZE);
  56         if (!stack) {
  57                 SYSERROR("Failed to allocate clone stack");
  58                 return -ENOMEM;
  59         }
  60
  61 #ifdef __ia64__
  62         ret = __clone2(fn, stack, __LXC_STACK_SIZE, flags | SIGCHLD, arg, pidfd);
  63 #else
  64         ret = clone(fn, stack + __LXC_STACK_SIZE, flags | SIGCHLD, arg, pidfd);
  65 #endif
  66         if (ret < 0)
  67                 SYSERROR("Failed to clone (%#x)", flags);
  68
  69         return ret;
  70 }
  71
  72 /* Leave the user namespace at the first position in the array of structs so
  73  * that we always attach to it first when iterating over the struct and using
  74  * setns() to switch namespaces. This especially affects lxc_attach(): Suppose
  75  * you cloned a new user namespace and mount namespace as an unprivileged user
  76  * on the host and want to setns() to the mount namespace. This requires you to
  77  * attach to the user namespace first otherwise the kernel will fail this check:
  78  *
  79  *        if (!ns_capable(mnt_ns->user_ns, CAP_SYS_ADMIN) ||
  80  *            !ns_capable(current_user_ns(), CAP_SYS_CHROOT) ||
  81  *            !ns_capable(current_user_ns(), CAP_SYS_ADMIN))
  82  *            return -EPERM;
  83  *
  84  *    in
  85  *
  86  *        linux/fs/namespace.c:mntns_install().
  87  */
  88 const struct ns_info ns_info[LXC_NS_MAX] = {
  89         [LXC_NS_USER]    = { "user",   CLONE_NEWUSER,   "CLONE_NEWUSER",   "LXC_USER_NS"    },
  90         [LXC_NS_MNT]    =  { "mnt",    CLONE_NEWNS,     "CLONE_NEWNS",     "LXC_MNT_NS"     },
  91         [LXC_NS_PID]    =  { "pid",    CLONE_NEWPID,    "CLONE_NEWPID",    "LXC_PID_NS"     },
  92         [LXC_NS_UTS]    =  { "uts",    CLONE_NEWUTS,    "CLONE_NEWUTS",    "LXC_UTS_NS"     },
  93         [LXC_NS_IPC]    =  { "ipc",    CLONE_NEWIPC,    "CLONE_NEWIPC",    "LXC_IPC_NS"     },
  94         [LXC_NS_NET]    =  { "net",    CLONE_NEWNET,    "CLONE_NEWNET",    "LXC_NET_NS"     },
  95         [LXC_NS_CGROUP] =  { "cgroup", CLONE_NEWCGROUP, "CLONE_NEWCGROUP", "LXC_CGROUP_NS"  }
  96 };
  97
  98 int lxc_namespace_2_cloneflag(const char *namespace)
  99 {
 100         int i;
 101
 102         for (i = 0; i < LXC_NS_MAX; i++)
 103                 if (!strcasecmp(ns_info[i].proc_name, namespace))
 104                         return ns_info[i].clone_flag;
 105
 106         ERROR("Invalid namespace name \"%s\"", namespace);
 107         return -EINVAL;
 108 }
 109
 110 int lxc_namespace_2_ns_idx(const char *namespace)
 111 {
 112         int i;
 113
 114         for (i = 0; i < LXC_NS_MAX; i++)
 115                 if (!strcmp(ns_info[i].proc_name, namespace))
 116                         return i;
 117
 118         ERROR("Invalid namespace name \"%s\"", namespace);
 119         return -EINVAL;
 120 }
 121
 122 extern int lxc_namespace_2_std_identifiers(char *namespaces)
 123 {
 124         char **it;
 125         char *del;
 126
 127         /* The identifiers for namespaces used with lxc-attach and lxc-unshare
 128          * as given on the manpage do not align with the standard identifiers.
 129          * This affects network, mount, and uts namespaces. The standard identifiers
 130          * are: "mnt", "uts", and "net" whereas lxc-attach and lxc-unshare uses
 131          * "MOUNT", "UTSNAME", and "NETWORK". So let's use some cheap memmove()s
 132          * to replace them by their standard identifiers.
 133          * Let's illustrate this with an example:
 134          * Assume the string:
 135          *
 136          *      "IPC|MOUNT|PID"
 137          *
 138          * then we memmove()
 139          *
 140          *      dest: del + 1 == OUNT|PID
 141          *      src:  del + 3 == NT|PID
 142          */
 143         if (!namespaces)
 144                 return -1;
 145
 146         while ((del = strstr(namespaces, "MOUNT")))
 147                 memmove(del + 1, del + 3, strlen(del) - 2);
 148
 149         for (it = (char *[]){"NETWORK", "UTSNAME", NULL}; it && *it; it++)
 150                 while ((del = strstr(namespaces, *it)))
 151                         memmove(del + 3, del + 7, strlen(del) - 6);
 152
 153         return 0;
 154 }
 155
 156 int lxc_fill_namespace_flags(char *flaglist, int *flags)
 157 {
 158         char *token;
 159         int aflag;
 160
 161         if (!flaglist) {
 162                 ERROR("At least one namespace is needed.");
 163                 return -1;
 164         }
 165
 166         lxc_iterate_parts(token, flaglist, "|") {
 167                 aflag = lxc_namespace_2_cloneflag(token);
 168                 if (aflag < 0)
 169                         return -1;
 170
 171                 *flags |= aflag;
 172         }
 173
 174         return 0;
 175 }