-
Notifications
You must be signed in to change notification settings - Fork 42
/
README.SETUID
85 lines (66 loc) · 3.94 KB
/
README.SETUID
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
hercifc and hercules as setuid root programs
Rationale :
Historically, hercifc is usually setuid root so that it can make changes
to the tun and tap interfaces on the system.
There are also some reasons to have the hercules executable itself setuid
root, for example, to allow the timer thread to run with real time priority.
This however leads to a number of security issues. hercifc is probably not
hardened enough and besides any possible bug that may get introduced in it,
there is also a possibility of bugs in the C library itself which may lead
to possible privilege escalations.
Having hercules itself run setuid 0 is even more of a problem since it is
now possible to load arbitrary code to support custom user made
implementation of some features and/or I/O devices).
In order to mitigate this, hercules now makes use of the Posix draft 1003.1e
capabilities in order to restrict the executables to just the necessary
privileges. For hercules, this means the CAP_SYS_NICE capability, and for
hercifc, this means CAP_NET_ADMIN capability.
This is still not an ideal situation since a setuid 0 hercules will allow
anyone to modify dispatching/scheduling priority for process it owns
beyond what is normally allowed for such a user. This is however less critical
than having hercules run as a full blown superuser.
Note that this only applies if hercules is setuid root. If this is not the
case, then hercules won't be able to set the timer thread priority (or any
other thread priority for that matter) to anything lower than 0 and the user
will not acquire any further privilege.
Having hercifc setuid root is pretty much mandatory if it is desired to
do some networking in the guest. hercifc already restricts on what interface
a controlling program may act. However, limiting its capabilities to just
CAP_NET_ADMIN restricts any escalation exploit to just THAT privilege.
On systems that do not have the capabilities feature implemented (or lacking
the necessary library) setting hercules setuid root is strongly discouraged
on a multi user machine (where users have normal access but not access to
superuser privilege).
On systems that DO have the capabilities feature implemented and enabled,
and providing hercules has been built with the support enabled, setting
hercules setuid root still allows normal users to potentially have the
capacity to set their own processes to negative nice values. So this also
should be taken into consideration.
Providing users have the authority to open a tun or tap device, having
hercifc setuid root allows those users to create network interfaces with
arbitrary network parameters. If this is a concern, those users should
either not be allowed access to the tun/tap interface or hercifc should
not be setuid root. Using capabilities with hercifc just guarantees that
should an exploit exist in hercifc to execute arbitrary code, then the
user will be limited to the CAP_NET_ADMIN privilege.
Building hercules with capabilities feature :
A new configure flag has been added : --[enable|disable]-capabilities
This flag allows building hercules without capabilities features even on a
system that would allow it. This may be necessary if the intended target may
be missing the necessary implementation (pre 2.2 linux kernel) and/or library
(libcap).
In order for the build to actually implement capabilities, the following files
are necessary :
libcap.a in the libraries directory
sys/capability.h in the includes directory
sys/prctl.h in the includes directory (this one is usually std on
linux systems)
In order to run a version of hercules built for capabilities support, the
following file is necessary :
libcap.so.x in the libraries directory (x is currently "1" at the time of
writing).
On linux systems, these files usually come with the "libcap1" package for
the runtime environment and the "libcap-dev" package for the build
environment. Depending on your linux distribution, these packages may have
other names.
--Ivan Warren, 2/17/2008