1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
|
#include "file-descriptor.hh"
#include "file-system.hh"
#include "logging.hh"
#include "namespaces.hh"
#include "processes.hh"
#include "strings.hh"
#include <sys/mount.h>
#if __linux__
# include <mutex>
# include <sys/resource.h>
#endif
namespace nix {
#if __linux__
static AutoCloseFD fdSavedMountNamespace;
static AutoCloseFD fdSavedRoot;
#endif
void saveMountNamespace()
{
#if __linux__
static std::once_flag done;
std::call_once(done, []() {
fdSavedMountNamespace = AutoCloseFD{open("/proc/self/ns/mnt", O_RDONLY)};
if (!fdSavedMountNamespace)
throw SysError("saving parent mount namespace");
fdSavedRoot = AutoCloseFD{open("/proc/self/root", O_RDONLY)};
});
#endif
}
void restoreMountNamespace()
{
#if __linux__
try {
auto savedCwd = absPath(".");
if (fdSavedMountNamespace && setns(fdSavedMountNamespace.get(), CLONE_NEWNS) == -1)
throw SysError("restoring parent mount namespace");
if (fdSavedRoot) {
if (fchdir(fdSavedRoot.get()))
throw SysError("chdir into saved root");
if (chroot("."))
throw SysError("chroot into saved root");
}
if (chdir(savedCwd.c_str()) == -1)
throw SysError("restoring cwd");
} catch (Error & e) {
debug(e.msg());
}
#endif
}
void unshareFilesystem()
{
#ifdef __linux__
if (unshare(CLONE_FS) != 0 && errno != EPERM)
throw SysError("unsharing filesystem state in download thread");
#endif
}
#if __linux__
static void diagnoseUserNamespaces()
{
if (!pathExists("/proc/self/ns/user")) {
warn("'/proc/self/ns/user' does not exist; your kernel was likely built without CONFIG_USER_NS=y");
}
Path maxUserNamespaces = "/proc/sys/user/max_user_namespaces";
if (!pathExists(maxUserNamespaces) ||
trim(readFile(maxUserNamespaces)) == "0")
{
warn("user namespaces appear to be disabled; check '/proc/sys/user/max_user_namespaces'");
}
Path procSysKernelUnprivilegedUsernsClone = "/proc/sys/kernel/unprivileged_userns_clone";
if (pathExists(procSysKernelUnprivilegedUsernsClone)
&& trim(readFile(procSysKernelUnprivilegedUsernsClone)) == "0")
{
warn("user namespaces appear to be disabled for unprivileged users; check '/proc/sys/kernel/unprivileged_userns_clone'");
}
}
bool userNamespacesSupported()
{
static auto res = [&]() -> bool
{
try {
Pid pid = startProcess([&]() { _exit(0); }, {.cloneFlags = CLONE_NEWUSER});
auto r = pid.wait();
assert(!r);
} catch (SysError & e) {
warn("user namespaces do not work on this system: %s", e.msg());
diagnoseUserNamespaces();
return false;
}
return true;
}();
return res;
}
bool mountAndPidNamespacesSupported()
{
static auto res = [&]() -> bool
{
try {
Pid pid = startProcess([&]() {
/* Make sure we don't remount the parent's /proc. */
if (mount(0, "/", 0, MS_PRIVATE | MS_REC, 0) == -1)
_exit(1);
/* Test whether we can remount /proc. The kernel disallows
this if /proc is not fully visible, i.e. if there are
filesystems mounted on top of files inside /proc. See
https://lore.kernel.org/lkml/87tvsrjai0.fsf@xmission.com/T/. */
if (mount("none", "/proc", "proc", 0, 0) == -1)
_exit(2);
_exit(0);
}, {
.cloneFlags = CLONE_NEWNS | CLONE_NEWPID | (userNamespacesSupported() ? CLONE_NEWUSER : 0)
});
if (pid.wait()) {
debug("PID namespaces do not work on this system: cannot remount /proc");
return false;
}
} catch (SysError & e) {
debug("mount namespaces do not work on this system: %s", e.msg());
return false;
}
return true;
}();
return res;
}
#endif
}
|
