]>
Commit | Line | Data |
---|---|---|
c061f33f KC |
1 | =========================================== |
2 | Seccomp BPF (SECure COMPuting with filters) | |
3 | =========================================== | |
8ac270d1 WD |
4 | |
5 | Introduction | |
c061f33f | 6 | ============ |
8ac270d1 WD |
7 | |
8 | A large number of system calls are exposed to every userland process | |
9 | with many of them going unused for the entire lifetime of the process. | |
10 | As system calls change and mature, bugs are found and eradicated. A | |
11 | certain subset of userland applications benefit by having a reduced set | |
12 | of available system calls. The resulting set reduces the total kernel | |
13 | surface exposed to the application. System call filtering is meant for | |
14 | use with those applications. | |
15 | ||
16 | Seccomp filtering provides a means for a process to specify a filter for | |
17 | incoming system calls. The filter is expressed as a Berkeley Packet | |
18 | Filter (BPF) program, as with socket filters, except that the data | |
19 | operated on is related to the system call being made: system call | |
20 | number and the system call arguments. This allows for expressive | |
21 | filtering of system calls using a filter program language with a long | |
22 | history of being exposed to userland and a straightforward data set. | |
23 | ||
24 | Additionally, BPF makes it impossible for users of seccomp to fall prey | |
25 | to time-of-check-time-of-use (TOCTOU) attacks that are common in system | |
26 | call interposition frameworks. BPF programs may not dereference | |
27 | pointers which constrains all filters to solely evaluating the system | |
28 | call arguments directly. | |
29 | ||
30 | What it isn't | |
c061f33f | 31 | ============= |
8ac270d1 WD |
32 | |
33 | System call filtering isn't a sandbox. It provides a clearly defined | |
34 | mechanism for minimizing the exposed kernel surface. It is meant to be | |
35 | a tool for sandbox developers to use. Beyond that, policy for logical | |
36 | behavior and information flow should be managed with a combination of | |
37 | other system hardening techniques and, potentially, an LSM of your | |
38 | choosing. Expressive, dynamic filters provide further options down this | |
39 | path (avoiding pathological sizes or selecting which of the multiplexed | |
40 | system calls in socketcall() is allowed, for instance) which could be | |
41 | construed, incorrectly, as a more complete sandboxing solution. | |
42 | ||
43 | Usage | |
c061f33f | 44 | ===== |
8ac270d1 WD |
45 | |
46 | An additional seccomp mode is added and is enabled using the same | |
47 | prctl(2) call as the strict seccomp. If the architecture has | |
c061f33f | 48 | ``CONFIG_HAVE_ARCH_SECCOMP_FILTER``, then filters may be added as below: |
8ac270d1 | 49 | |
c061f33f | 50 | ``PR_SET_SECCOMP``: |
8ac270d1 WD |
51 | Now takes an additional argument which specifies a new filter |
52 | using a BPF program. | |
53 | The BPF program will be executed over struct seccomp_data | |
54 | reflecting the system call number, arguments, and other | |
55 | metadata. The BPF program must then return one of the | |
56 | acceptable values to inform the kernel which action should be | |
57 | taken. | |
58 | ||
c061f33f KC |
59 | Usage:: |
60 | ||
8ac270d1 WD |
61 | prctl(PR_SET_SECCOMP, SECCOMP_MODE_FILTER, prog); |
62 | ||
63 | The 'prog' argument is a pointer to a struct sock_fprog which | |
64 | will contain the filter program. If the program is invalid, the | |
c061f33f | 65 | call will return -1 and set errno to ``EINVAL``. |
8ac270d1 | 66 | |
c061f33f | 67 | If ``fork``/``clone`` and ``execve`` are allowed by @prog, any child |
8ac270d1 WD |
68 | processes will be constrained to the same filters and system |
69 | call ABI as the parent. | |
70 | ||
c061f33f KC |
71 | Prior to use, the task must call ``prctl(PR_SET_NO_NEW_PRIVS, 1)`` or |
72 | run with ``CAP_SYS_ADMIN`` privileges in its namespace. If these are not | |
73 | true, ``-EACCES`` will be returned. This requirement ensures that filter | |
8ac270d1 WD |
74 | programs cannot be applied to child processes with greater privileges |
75 | than the task that installed them. | |
76 | ||
c061f33f | 77 | Additionally, if ``prctl(2)`` is allowed by the attached filter, |
8ac270d1 WD |
78 | additional filters may be layered on which will increase evaluation |
79 | time, but allow for further decreasing the attack surface during | |
80 | execution of a process. | |
81 | ||
82 | The above call returns 0 on success and non-zero on error. | |
83 | ||
84 | Return values | |
c061f33f KC |
85 | ============= |
86 | ||
8ac270d1 WD |
87 | A seccomp filter may return any of the following values. If multiple |
88 | filters exist, the return value for the evaluation of a given system | |
89 | call will always use the highest precedent value. (For example, | |
c061f33f | 90 | ``SECCOMP_RET_KILL`` will always take precedence.) |
8ac270d1 WD |
91 | |
92 | In precedence order, they are: | |
93 | ||
c061f33f | 94 | ``SECCOMP_RET_KILL``: |
8ac270d1 | 95 | Results in the task exiting immediately without executing the |
c061f33f KC |
96 | system call. The exit status of the task (``status & 0x7f``) will |
97 | be ``SIGSYS``, not ``SIGKILL``. | |
8ac270d1 | 98 | |
c061f33f KC |
99 | ``SECCOMP_RET_TRAP``: |
100 | Results in the kernel sending a ``SIGSYS`` signal to the triggering | |
101 | task without executing the system call. ``siginfo->si_call_addr`` | |
87b526d3 | 102 | will show the address of the system call instruction, and |
c061f33f | 103 | ``siginfo->si_syscall`` and ``siginfo->si_arch`` will indicate which |
87b526d3 AL |
104 | syscall was attempted. The program counter will be as though |
105 | the syscall happened (i.e. it will not point to the syscall | |
106 | instruction). The return value register will contain an arch- | |
107 | dependent value -- if resuming execution, set it to something | |
108 | sensible. (The architecture dependency is because replacing | |
c061f33f | 109 | it with ``-ENOSYS`` could overwrite some useful information.) |
8ac270d1 | 110 | |
c061f33f KC |
111 | The ``SECCOMP_RET_DATA`` portion of the return value will be passed |
112 | as ``si_errno``. | |
8ac270d1 | 113 | |
c061f33f | 114 | ``SIGSYS`` triggered by seccomp will have a si_code of ``SYS_SECCOMP``. |
8ac270d1 | 115 | |
c061f33f | 116 | ``SECCOMP_RET_ERRNO``: |
8ac270d1 WD |
117 | Results in the lower 16-bits of the return value being passed |
118 | to userland as the errno without executing the system call. | |
119 | ||
c061f33f | 120 | ``SECCOMP_RET_TRACE``: |
8ac270d1 | 121 | When returned, this value will cause the kernel to attempt to |
c061f33f KC |
122 | notify a ``ptrace()``-based tracer prior to executing the system |
123 | call. If there is no tracer present, ``-ENOSYS`` is returned to | |
8ac270d1 WD |
124 | userland and the system call is not executed. |
125 | ||
c061f33f KC |
126 | A tracer will be notified if it requests ``PTRACE_O_TRACESECCOM``P |
127 | using ``ptrace(PTRACE_SETOPTIONS)``. The tracer will be notified | |
128 | of a ``PTRACE_EVENT_SECCOMP`` and the ``SECCOMP_RET_DATA`` portion of | |
8ac270d1 | 129 | the BPF program return value will be available to the tracer |
c061f33f | 130 | via ``PTRACE_GETEVENTMSG``. |
8ac270d1 | 131 | |
87b526d3 AL |
132 | The tracer can skip the system call by changing the syscall number |
133 | to -1. Alternatively, the tracer can change the system call | |
134 | requested by changing the system call to a valid syscall number. If | |
135 | the tracer asks to skip the system call, then the system call will | |
136 | appear to return the value that the tracer puts in the return value | |
137 | register. | |
138 | ||
139 | The seccomp check will not be run again after the tracer is | |
140 | notified. (This means that seccomp-based sandboxes MUST NOT | |
141 | allow use of ptrace, even of other sandboxed processes, without | |
142 | extreme care; ptracers can use this mechanism to escape.) | |
143 | ||
c061f33f | 144 | ``SECCOMP_RET_ALLOW``: |
8ac270d1 WD |
145 | Results in the system call being executed. |
146 | ||
147 | If multiple filters exist, the return value for the evaluation of a | |
148 | given system call will always use the highest precedent value. | |
149 | ||
c061f33f | 150 | Precedence is only determined using the ``SECCOMP_RET_ACTION`` mask. When |
8ac270d1 | 151 | multiple filters return values of the same precedence, only the |
c061f33f | 152 | ``SECCOMP_RET_DATA`` from the most recently installed filter will be |
8ac270d1 WD |
153 | returned. |
154 | ||
155 | Pitfalls | |
c061f33f | 156 | ======== |
8ac270d1 WD |
157 | |
158 | The biggest pitfall to avoid during use is filtering on system call | |
159 | number without checking the architecture value. Why? On any | |
160 | architecture that supports multiple system call invocation conventions, | |
161 | the system call numbers may vary based on the specific invocation. If | |
162 | the numbers in the different calling conventions overlap, then checks in | |
163 | the filters may be abused. Always check the arch value! | |
164 | ||
165 | Example | |
c061f33f | 166 | ======= |
8ac270d1 | 167 | |
c061f33f | 168 | The ``samples/seccomp/`` directory contains both an x86-specific example |
8ac270d1 WD |
169 | and a more generic example of a higher level macro interface for BPF |
170 | program generation. | |
171 | ||
9ca58825 TH |
172 | Sysctls |
173 | ======= | |
174 | ||
175 | Seccomp's sysctl files can be found in the ``/proc/sys/kernel/seccomp/`` | |
176 | directory. Here's a description of each file in that directory: | |
177 | ||
178 | ``actions_avail``: | |
179 | A read-only ordered list of seccomp return values (refer to the | |
180 | ``SECCOMP_RET_*`` macros above) in string form. The ordering, from | |
181 | left-to-right, is the least permissive return value to the most | |
182 | permissive return value. | |
8ac270d1 | 183 | |
9ca58825 TH |
184 | The list represents the set of seccomp return values supported |
185 | by the kernel. A userspace program may use this list to | |
186 | determine if the actions found in the ``seccomp.h``, when the | |
187 | program was built, differs from the set of actions actually | |
188 | supported in the current running kernel. | |
8ac270d1 WD |
189 | |
190 | Adding architecture support | |
c061f33f | 191 | =========================== |
8ac270d1 | 192 | |
c061f33f | 193 | See ``arch/Kconfig`` for the authoritative requirements. In general, if an |
8ac270d1 | 194 | architecture supports both ptrace_event and seccomp, it will be able to |
c061f33f KC |
195 | support seccomp filter with minor fixup: ``SIGSYS`` support and seccomp return |
196 | value checking. Then it must just add ``CONFIG_HAVE_ARCH_SECCOMP_FILTER`` | |
8ac270d1 | 197 | to its arch-specific Kconfig. |
87b526d3 AL |
198 | |
199 | ||
200 | ||
201 | Caveats | |
c061f33f | 202 | ======= |
87b526d3 AL |
203 | |
204 | The vDSO can cause some system calls to run entirely in userspace, | |
205 | leading to surprises when you run programs on different machines that | |
206 | fall back to real syscalls. To minimize these surprises on x86, make | |
207 | sure you test with | |
c061f33f KC |
208 | ``/sys/devices/system/clocksource/clocksource0/current_clocksource`` set to |
209 | something like ``acpi_pm``. | |
87b526d3 AL |
210 | |
211 | On x86-64, vsyscall emulation is enabled by default. (vsyscalls are | |
c061f33f KC |
212 | legacy variants on vDSO calls.) Currently, emulated vsyscalls will |
213 | honor seccomp, with a few oddities: | |
87b526d3 | 214 | |
c061f33f | 215 | - A return value of ``SECCOMP_RET_TRAP`` will set a ``si_call_addr`` pointing to |
87b526d3 AL |
216 | the vsyscall entry for the given call and not the address after the |
217 | 'syscall' instruction. Any code which wants to restart the call | |
218 | should be aware that (a) a ret instruction has been emulated and (b) | |
219 | trying to resume the syscall will again trigger the standard vsyscall | |
220 | emulation security checks, making resuming the syscall mostly | |
221 | pointless. | |
222 | ||
c061f33f | 223 | - A return value of ``SECCOMP_RET_TRACE`` will signal the tracer as usual, |
87b526d3 AL |
224 | but the syscall may not be changed to another system call using the |
225 | orig_rax register. It may only be changed to -1 order to skip the | |
226 | currently emulated call. Any other change MAY terminate the process. | |
227 | The rip value seen by the tracer will be the syscall entry address; | |
228 | this is different from normal behavior. The tracer MUST NOT modify | |
229 | rip or rsp. (Do not rely on other changes terminating the process. | |
230 | They might work. For example, on some kernels, choosing a syscall | |
231 | that only exists in future kernels will be correctly emulated (by | |
c061f33f | 232 | returning ``-ENOSYS``). |
87b526d3 | 233 | |
c061f33f KC |
234 | To detect this quirky behavior, check for ``addr & ~0x0C00 == |
235 | 0xFFFFFFFFFF600000``. (For ``SECCOMP_RET_TRACE``, use rip. For | |
236 | ``SECCOMP_RET_TRAP``, use ``siginfo->si_call_addr``.) Do not check any other | |
87b526d3 AL |
237 | condition: future kernels may improve vsyscall emulation and current |
238 | kernels in vsyscall=native mode will behave differently, but the | |
c061f33f | 239 | instructions at ``0xF...F600{0,4,8,C}00`` will not be system calls in these |
87b526d3 AL |
240 | cases. |
241 | ||
242 | Note that modern systems are unlikely to use vsyscalls at all -- they | |
243 | are a legacy feature and they are considerably slower than standard | |
244 | syscalls. New code will use the vDSO, and vDSO-issued system calls | |
245 | are indistinguishable from normal system calls. |