forked from Raynes/clojail
-
Notifications
You must be signed in to change notification settings - Fork 1
/
core.clj
301 lines (269 loc) · 11.6 KB
/
core.clj
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
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
;; Clojail is an easy way to sandbox your code. Whether you want to allow evaluation on a website,
;; in an IRC bot, or anything else you can think of, Clojail wants to be the easiest and most
;; comprehensive way to do that. But it isn't easy.
(ns clojail.core
(:use clojure.stacktrace
[clojure.walk :only [walk postwalk-replace]]
clojail.jvm)
(:import (java.util.concurrent TimeoutException TimeUnit FutureTask)
(clojure.lang LispReader$ReaderException)))
;; postwalk is like a magical recursive doall, to force lazy-seqs
;; within the timeout context; but since it doesn't maintain perfect
;; structure for *every* data type, we want to actually return the
;; original value after we force it, not the result of postwalk
;; replacement
(defn eagerly-consume
"Recursively force all lazy-seqs in val."
[val]
(try
(postwalk-replace {} val)
(catch Throwable _))
val)
;; It sucks to have to deal with TimeUnits. They're so damned long.
(def uglify-time-unit "Create a map of pretty keywords to ugly TimeUnits"
(into {} (for [[enum aliases] {TimeUnit/NANOSECONDS [:ns :nanoseconds]
TimeUnit/MICROSECONDS [:us :microseconds]
TimeUnit/MILLISECONDS [:ms :milliseconds]
TimeUnit/SECONDS [:s :sec :seconds]}
alias aliases]
{alias enum})))
;; This function uses some deprecated Java methods to stop threads, but the
;; reason they're deprecated doesn't really apply here. Just because people
;; don't use them properly doesn't mean they aren't useful.
;;
;; This function is useful in general, and that's why it is public.
(defn thunk-timeout
"Takes a function and an amount of time to wait for the function to finish
executing. The sandbox can do this for you. unit is any of :ns, :us, :ms,
or :s which correspond to TimeUnit/NANOSECONDS, MICROSECONDS, MILLISECONDS,
and SECONDS respectively."
([thunk ms]
(thunk-timeout thunk ms :ms)) ; Default to milliseconds, because that's pretty common.
([thunk time unit]
(thunk-timeout thunk time unit identity))
([thunk time unit transform]
(thunk-timeout thunk time unit identity nil))
([thunk time unit transform tg]
(let [task (FutureTask. (comp transform thunk))
thr (if tg (Thread. tg task) (Thread. task))]
(try
(.start thr)
(.get task time (or (uglify-time-unit unit) unit))
(catch TimeoutException e
(.cancel task true)
(.stop thr)
(throw (TimeoutException. "Execution timed out.")))
(catch Exception e
(.cancel task true)
(.stop thr)
(throw e))
(finally (when tg (.stop tg)))))))
(defn safe-resolve
"Resolves things safely."
[s nspace]
(try (if-let [resolved (ns-resolve nspace s)]
resolved
s)
(catch RuntimeException _ s)))
(defn flatten-all
"The core flatten doesn't flatten maps."
[form] (remove coll? (tree-seq coll? seq form)))
(defn- collify
"If form isn't a collection, wrap it in a vector."
[form] (if (coll? form) form [form]))
(defn- macroexpand-most
"Macroexpand most, but not all. Leave non-collections and quoted things alone."
[form]
(if (or
(not (coll? form))
(and (seq? form)
(= 'quote (first form))))
form
(walk macroexpand-most identity (macroexpand form))))
(defn- separate
"Take a collection and break it and its contents apart until we have
a set of things to check for badness against."
[s nspace]
(set
(flatten-all
(map #(if (symbol? %)
(let [resolved-s (safe-resolve % nspace)
s-meta (meta resolved-s)]
(if s-meta
[resolved-s ((juxt (comp symbol str :ns) :ns :name) s-meta)]
(let [[bottom] (map symbol (.split (str %) "/"))
resolved-s (safe-resolve bottom nspace)]
(if (class? resolved-s)
[resolved-s %]
%))))
%)
(flatten-all (collify (macroexpand-most s)))))))
;; Because the dot (.) interop form is a special form, we can't just rebind it or anything.
;; Instead, we need to replace it entirely with a safe macro of our own. To do this, we need
;; to replace all . symbols with 'dot', the name of our own safe dot macro.
(defn- dotify
"Replace all . symbols with 'dot."
[form]
(if-not (coll? form)
form
(let [recurse #(walk dotify identity %)]
(if-not (seq? form)
(recurse form)
(let [f (first form)]
(case f
quote form
. (cons 'dot (recurse (rest form)))
(recurse form)))))))
;; Compose our earlier functions.
(def ^{:private true} ensafen "Fix code to make interop safe."
(comp dotify macroexpand-most))
(defn- mutilate
"Macroexpand and separate pieces to create a set of symbols an")
;; The clojail equivalent of motion detectors.
(defn check-form
"Check a form to see if it trips a tester."
[form tester nspace]
(some tester (separate form nspace)))
;; We have to run the sandbox against packages as well as classes,
;; but macros can't embed Package objects in code by default. This
;; is a simple print-dup method so that we can embed them in our dot
;; macro.
(defmethod print-dup java.lang.Package
([p out]
(.write out (str "#=(java.lang.Package/getPackage \""
(.getName p)
"\")"))))
(defn security-exception [problem]
(throw
(SecurityException.
(format "You tripped the alarm! %s is bad!" problem))))
(defn- make-dot
"Returns a safe . macro."
[tester-str]
`(defmacro ~'dot [object# method# & args#]
`(let [~'tester-obj# (binding [*read-eval* true] (read-string ~~tester-str))
~'obj# ~object#
~'obj-class# (class ~'obj#)]
(if-let [~'bad# (some ~'tester-obj# [~'obj-class# ~'obj# (.getPackage ~'obj-class#)])]
(security-exception ~'bad#)
(. ~object# ~method# ~@args#)))))
(defn- user-defs
"Find get a set of all the symbols of vars defined in a namespace."
[nspace] (set (keys (ns-interns nspace))))
(defn- bulk-unmap
"Unmap a bunch of vars."
[nspace vars]
(doseq [n vars]
(binding [*ns* nspace]
(eval `(ns-unmap *ns* '~n)))))
(defn- wipe-defs
"Unmap vals in the sandbox only if the count of them is higher than max-defs."
[init-defs old-defs max-defs nspace]
(let [defs (remove init-defs (user-defs nspace))
new-defs (remove old-defs defs)]
(when (> (count defs) max-defs)
(bulk-unmap nspace (remove init-defs old-defs)))
(when (> (count new-defs) max-defs)
(bulk-unmap nspace new-defs))))
(defn- read-tester [tester]
(with-out-str (binding [*print-dup* true] (pr tester))))
(defn- evaluator [code tester-str context nspace bindings]
(fn []
(binding [*ns* nspace
*read-eval* false]
(let [bindings (or bindings {})
code `(do ~(make-dot tester-str)
~(ensafen code))]
(with-bindings bindings (jvm-sandbox #(eval code) context))))))
(defn set-security-manager
"Sets the system security manager to whatever you pass. Passing nil is
the equivalent of turning it off entirely (which is usually how the JVM
starts up)."
[s] (System/setSecurityManager s))
(defn sandbox*
"This function creates a sandbox function that takes a tester. A tester is a set of objects
that you don't want to be allowed in code. It is a blacklist.
Optional arguments are as follows:
:timeout, default is 10000 MS or 10 seconds. If the expression evaluated in the sandbox
takes longer than the timeout, an error will be thrown and the thread running the code
will be stopped.
:namespace, the namespace of the sandbox. The default is (gensym \"sandbox\").
:context, the context for the JVM sandbox to run in. Only relevant if :jvm? is true. It
has a sane default, so you shouldn't need to worry about this.
:jvm?, if set to true, the JVM sandbox will be employed. It defaults to true.
:transform, a function to call on the result returned from the sandboxed code, before
returning it, while still within the timeout context.
:init, some (quoted) code to run in the sandbox's namespace, but outside of the sandbox.
:refer-clojure, true or false. If true (the default), automatically refer-clojure in the ns.
You might want to set this to false at some point if you're working with the namespace in the
:init key.
This function will return a new function that you should bind to something. You can call
this function with code and it will be evaluated in the sandbox. The function also takes
an optional second parameter which is a hashmap of vars to values that will be passed to
with-bindings. Since Clojure 1.3, only vars explicitly declared as dynamic are able to be
rebound. As a result, only those vars will work here. If this doesn't work for you,
read about the :init key."
[& {:keys [timeout namespace context jvm transform
init ns-init max-defs refer-clojure]
:or {timeout 10000
namespace (gensym "sandbox")
context (-> (permissions) domain context)
jvm true
transform eagerly-consume
refer-clojure true
max-defs 5}}]
(let [nspace (create-ns namespace)]
(binding [*ns* nspace]
(when refer-clojure (clojure.core/refer-clojure))
(eval init))
(let [init-defs (conj (user-defs nspace) 'dot)]
(fn [code tester & [bindings]]
(let [tester-str (read-tester tester)
old-defs (user-defs nspace)]
(when jvm (set-security-manager (SecurityManager.)))
(try
(let [result (if-let [problem (check-form code tester nspace)]
(security-exception problem)
(thunk-timeout
(evaluator code tester-str context nspace bindings)
timeout
:ms
transform
(ThreadGroup. "sandbox")))]
result)
(finally (wipe-defs init-defs old-defs max-defs nspace))))))))
(defn sandbox
"Convenience wrapper function around sandbox* to create a sandbox function out of a tester.
Takes the same arguments as sandbox* with the addition of the tester argument. Returns a
sandbox function like sandbox* returns, the difference being that the tester is hardcoded
and doesn't need to be passed to the created function."
[tester & args]
(let [sb (apply sandbox* args)]
#(apply sb % tester %&)))
(defn safe-read
"Read a string from an untrusted source. Mainly just disables read-eval,
but also repackages thrown exceptions to make it easier to
discriminate among them. read-eval errors will be thrown as
IllegalStateException; other exceptions will be thrown unchanged."
([]
(binding [*read-eval* false]
(let [repackage (fn [e]
(let [msg (str (.getName (class e))
": "
(.getMessage (root-cause e)))]
(if (.contains msg "EvalReader")
(IllegalStateException. msg)
e)))]
(try
(read)
(catch LispReader$ReaderException e
(throw (repackage e)))
(catch Throwable e
(let [cause (.getCause e)]
(cond
(not cause) (throw e)
(not (instance? LispReader$ReaderException cause)) (throw e)
:else (throw (repackage cause)))))))))
([str]
(with-in-str str
(safe-read))))