Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
96.71% |
499 / 516 |
|
87.10% |
54 / 62 |
CRAP | |
0.00% |
0 / 1 |
| TaintednessVisitor | |
96.71% |
499 / 516 |
|
87.10% |
54 / 62 |
150 | |
0.00% |
0 / 1 |
| analyzeNodeAndGetTaintedness | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| setCachedData | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| setCurTaintUnknown | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
1 | |||
| setCurTaintSafe | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
1 | |||
| visit | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
2 | |||
| visitClosure | |
85.71% |
6 / 7 |
|
0.00% |
0 / 1 |
2.01 | |||
| visitFuncDecl | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| visitMethod | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| visitArrowFunc | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| analyzeFunctionLike | |
75.00% |
3 / 4 |
|
0.00% |
0 / 1 |
4.25 | |||
| visitClassName | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| visitThrow | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| visitUnset | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
3 | |||
| handleUnsetDim | |
72.22% |
13 / 18 |
|
0.00% |
0 / 1 |
9.37 | |||
| visitClone | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| visitAssignOp | |
93.33% |
28 / 30 |
|
0.00% |
0 / 1 |
3.00 | |||
| visitStatic | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| visitAssignRef | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitAssign | |
100.00% |
17 / 17 |
|
100.00% |
1 / 1 |
2 | |||
| doVisitAssign | |
100.00% |
14 / 14 |
|
100.00% |
1 / 1 |
1 | |||
| visitBinaryOp | |
100.00% |
21 / 21 |
|
100.00% |
1 / 1 |
2 | |||
| getBinOpTaint | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
3 | |||
| visitDim | |
100.00% |
17 / 17 |
|
100.00% |
1 / 1 |
3 | |||
| visitPrint | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitExit | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitShellExec | |
100.00% |
11 / 11 |
|
100.00% |
1 / 1 |
1 | |||
| visitIncludeOrEval | |
100.00% |
8 / 8 |
|
100.00% |
1 / 1 |
2 | |||
| visitEcho | |
100.00% |
7 / 7 |
|
100.00% |
1 / 1 |
1 | |||
| visitSimpleSinkAndPropagate | |
100.00% |
14 / 14 |
|
100.00% |
1 / 1 |
4 | |||
| visitStaticCall | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitNew | |
100.00% |
40 / 40 |
|
100.00% |
1 / 1 |
8 | |||
| visitMethodCall | |
100.00% |
9 / 9 |
|
100.00% |
1 / 1 |
2 | |||
| analyzeCallNode | |
88.89% |
8 / 9 |
|
0.00% |
0 / 1 |
4.02 | |||
| visitNullsafeMethodCall | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitCall | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
2 | |||
| visitVar | |
91.67% |
22 / 24 |
|
0.00% |
0 / 1 |
4.01 | |||
| getHardcodedTaintednessForVar | |
100.00% |
37 / 37 |
|
100.00% |
1 / 1 |
14 | |||
| visitGlobal | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
6 | |||
| visitReturn | |
66.67% |
6 / 9 |
|
0.00% |
0 / 1 |
3.33 | |||
| setFuncTaintFromReturn | |
100.00% |
30 / 30 |
|
100.00% |
1 / 1 |
7 | |||
| visitArray | |
100.00% |
31 / 31 |
|
100.00% |
1 / 1 |
9 | |||
| visitClassConst | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| visitConst | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| visitStaticProp | |
80.00% |
8 / 10 |
|
0.00% |
0 / 1 |
2.03 | |||
| visitProp | |
100.00% |
34 / 34 |
|
100.00% |
1 / 1 |
7 | |||
| visitNullsafeProp | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitConditional | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
2 | |||
| visitName | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitNameList | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitUnaryOp | |
100.00% |
8 / 8 |
|
100.00% |
1 / 1 |
2 | |||
| visitPostInc | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitPreInc | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitPostDec | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitPreDec | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| analyzeIncOrDec | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| visitCast | |
100.00% |
14 / 14 |
|
100.00% |
1 / 1 |
2 | |||
| visitEncapsList | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
2 | |||
| visitIsset | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitEmpty | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitMagicConst | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitInstanceOf | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| visitMatch | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
3 | |||
| 1 | <?php declare( strict_types=1 ); |
| 2 | /** |
| 3 | * Copyright (C) 2017 Brian Wolff <bawolff@gmail.com> |
| 4 | * |
| 5 | * This program is free software; you can redistribute it and/or modify |
| 6 | * it under the terms of the GNU General Public License as published by |
| 7 | * the Free Software Foundation; either version 2 of the License, or |
| 8 | * (at your option) any later version. |
| 9 | * |
| 10 | * This program is distributed in the hope that it will be useful, |
| 11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
| 13 | * GNU General Public License for more details. |
| 14 | * |
| 15 | * You should have received a copy of the GNU General Public License along |
| 16 | * with this program; if not, write to the Free Software Foundation, Inc., |
| 17 | * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. |
| 18 | */ |
| 19 | |
| 20 | namespace SecurityCheckPlugin; |
| 21 | |
| 22 | use ast\Node; |
| 23 | use Phan\Analysis\BlockExitStatusChecker; |
| 24 | use Phan\AST\ContextNode; |
| 25 | use Phan\Debug; |
| 26 | use Phan\Exception\CodeBaseException; |
| 27 | use Phan\Exception\IssueException; |
| 28 | use Phan\Exception\NodeException; |
| 29 | use Phan\Language\Element\FunctionInterface; |
| 30 | use Phan\Language\Element\GlobalVariable; |
| 31 | use Phan\Language\FQSEN\FullyQualifiedClassName; |
| 32 | use Phan\Language\FQSEN\FullyQualifiedFunctionName; |
| 33 | use Phan\Language\Type\FunctionLikeDeclarationType; |
| 34 | use Phan\PluginV3\PluginAwarePostAnalysisVisitor; |
| 35 | |
| 36 | /** |
| 37 | * This class visits all the nodes in the ast. It has two jobs: |
| 38 | * |
| 39 | * 1) Return the taint value of the current node we are visiting. |
| 40 | * 2) In the event of an assignment (and similar things) propagate |
| 41 | * the taint value from the left hand side to the right hand side. |
| 42 | * |
| 43 | * For the moment, the taint values are stored in a "taintedness" |
| 44 | * property of various phan TypedElement objects. This is probably |
| 45 | * not the best solution for where to store the data, but its what |
| 46 | * this does for now. |
| 47 | * |
| 48 | * This also maintains some other properties, such as where the error |
| 49 | * originates, and dependencies in certain cases. |
| 50 | * |
| 51 | * @phan-file-suppress PhanUnusedPublicMethodParameter Many methods don't use $node |
| 52 | */ |
| 53 | class TaintednessVisitor extends PluginAwarePostAnalysisVisitor { |
| 54 | use TaintednessBaseVisitor; |
| 55 | |
| 56 | /** |
| 57 | * Node kinds whose taintedness is not well-defined and for which we don't need a visit* method. |
| 58 | */ |
| 59 | public const INAPPLICABLE_NODES_WITHOUT_VISITOR = [ |
| 60 | \ast\AST_ARG_LIST => true, |
| 61 | \ast\AST_TYPE => true, |
| 62 | \ast\AST_NULLABLE_TYPE => true, |
| 63 | \ast\AST_PARAM_LIST => true, |
| 64 | // Params are handled in PreTaintednessVisitor |
| 65 | \ast\AST_PARAM => true, |
| 66 | \ast\AST_CLASS => true, |
| 67 | \ast\AST_USE_ELEM => true, |
| 68 | \ast\AST_STMT_LIST => true, |
| 69 | \ast\AST_CLASS_CONST_DECL => true, |
| 70 | \ast\AST_CLASS_CONST_GROUP => true, |
| 71 | \ast\AST_CONST_DECL => true, |
| 72 | \ast\AST_IF => true, |
| 73 | \ast\AST_IF_ELEM => true, |
| 74 | \ast\AST_PROP_DECL => true, |
| 75 | \ast\AST_CONST_ELEM => true, |
| 76 | \ast\AST_USE => true, |
| 77 | \ast\AST_USE_TRAIT => true, |
| 78 | \ast\AST_BREAK => true, |
| 79 | \ast\AST_CONTINUE => true, |
| 80 | \ast\AST_GOTO => true, |
| 81 | \ast\AST_CATCH => true, |
| 82 | \ast\AST_NAMESPACE => true, |
| 83 | \ast\AST_SWITCH => true, |
| 84 | \ast\AST_SWITCH_CASE => true, |
| 85 | \ast\AST_SWITCH_LIST => true, |
| 86 | \ast\AST_WHILE => true, |
| 87 | \ast\AST_DO_WHILE => true, |
| 88 | \ast\AST_FOR => true, |
| 89 | // Handled in TaintednessLoopVisitor |
| 90 | \ast\AST_FOREACH => true, |
| 91 | \ast\AST_EXPR_LIST => true, |
| 92 | \ast\AST_TRY => true, |
| 93 | // Array elems are handled directly in visitArray |
| 94 | \ast\AST_ARRAY_ELEM => true, |
| 95 | // Initializing the prop is done in preorder |
| 96 | \ast\AST_PROP_ELEM => true, |
| 97 | \ast\AST_PROP_GROUP => true, |
| 98 | // Variables are already handled in visitVar |
| 99 | \ast\AST_CLOSURE_VAR => true, |
| 100 | \ast\AST_CLOSURE_USES => true, |
| 101 | \ast\AST_LABEL => true, |
| 102 | \ast\AST_ATTRIBUTE => true, |
| 103 | \ast\AST_ATTRIBUTE_GROUP => true, |
| 104 | \ast\AST_ATTRIBUTE_LIST => true, |
| 105 | ]; |
| 106 | |
| 107 | /** |
| 108 | * Node kinds whose taintedness is not well-defined, but for which we still need a visit* method. |
| 109 | * Trying to get the taintedness of these nodes will still result in an error. |
| 110 | */ |
| 111 | public const INAPPLICABLE_NODES_WITH_VISITOR = [ |
| 112 | \ast\AST_GLOBAL => true, |
| 113 | \ast\AST_RETURN => true, |
| 114 | \ast\AST_STATIC => true, |
| 115 | \ast\AST_FUNC_DECL => true, |
| 116 | \ast\AST_METHOD => true, |
| 117 | ]; |
| 118 | |
| 119 | /** |
| 120 | * Map of node kinds whose taintedness is not well-defined, e.g. because that node |
| 121 | * cannot be used as an expression. Note that it's safe to use array plus here. |
| 122 | */ |
| 123 | private const INAPPLICABLE_NODES = self::INAPPLICABLE_NODES_WITHOUT_VISITOR + self::INAPPLICABLE_NODES_WITH_VISITOR; |
| 124 | |
| 125 | /** @var TaintednessWithError|null */ |
| 126 | private $curTaintWithError; |
| 127 | |
| 128 | /** |
| 129 | * @param Node $node |
| 130 | * @return TaintednessWithError |
| 131 | */ |
| 132 | public function analyzeNodeAndGetTaintedness( Node $node ): TaintednessWithError { |
| 133 | assert( |
| 134 | !isset( self::INAPPLICABLE_NODES[$node->kind] ), |
| 135 | 'Should not try to get taintedness of inapplicable nodes (got ' . Debug::nodeName( $node ) . ')' |
| 136 | ); |
| 137 | $this->__invoke( $node ); |
| 138 | $this->setCachedData( $node ); |
| 139 | return $this->curTaintWithError; |
| 140 | } |
| 141 | |
| 142 | /** |
| 143 | * Cache taintedness data in an AST node. Ideally we'd want this to happen at the end of __invoke, but phan |
| 144 | * calls visit* methods by name, so that doesn't work. |
| 145 | * Caching a node *may* improve the speed, but *will* increase the memory usage, so only do that for nodes |
| 146 | * whose taintedness: |
| 147 | * - Is not trivial to compute, and |
| 148 | * - Might be needed from another node (via getTaintednessNode) |
| 149 | * @param Node $node |
| 150 | */ |
| 151 | private function setCachedData( Node $node ): void { |
| 152 | // @phan-suppress-next-line PhanUndeclaredProperty |
| 153 | $node->taint = $this->curTaintWithError; |
| 154 | } |
| 155 | |
| 156 | /** |
| 157 | * Sets $this->curTaint to UNKNOWN. Shorthand to filter the usages of curTaint. |
| 158 | */ |
| 159 | private function setCurTaintUnknown(): void { |
| 160 | $this->curTaintWithError = new TaintednessWithError( |
| 161 | Taintedness::newUnknown(), |
| 162 | new CausedByLines(), |
| 163 | MethodLinks::newEmpty() |
| 164 | ); |
| 165 | } |
| 166 | |
| 167 | private function setCurTaintSafe(): void { |
| 168 | $this->curTaintWithError = new TaintednessWithError( |
| 169 | Taintedness::newSafe(), |
| 170 | new CausedByLines(), |
| 171 | MethodLinks::newEmpty() |
| 172 | ); |
| 173 | } |
| 174 | |
| 175 | /** |
| 176 | * Generic visitor when we haven't defined a more specific one. |
| 177 | * |
| 178 | * @param Node $node |
| 179 | */ |
| 180 | public function visit( Node $node ): void { |
| 181 | // This method will be called on all nodes for which |
| 182 | // there is no implementation of its kind visitor. |
| 183 | |
| 184 | if ( isset( self::INAPPLICABLE_NODES_WITHOUT_VISITOR[$node->kind] ) ) { |
| 185 | return; |
| 186 | } |
| 187 | |
| 188 | // To see what kinds of nodes are passing through here, |
| 189 | // you can run `Debug::printNode($node)`. |
| 190 | # Debug::printNode( $node ); |
| 191 | $this->debug( __METHOD__, "unhandled case " . Debug::nodeName( $node ) ); |
| 192 | $this->setCurTaintUnknown(); |
| 193 | } |
| 194 | |
| 195 | /** |
| 196 | * @param Node $node |
| 197 | */ |
| 198 | public function visitClosure( Node $node ): void { |
| 199 | // We cannot use getFunctionLikeInScope for closures |
| 200 | $closureFQSEN = FullyQualifiedFunctionName::fromClosureInContext( $this->context, $node ); |
| 201 | |
| 202 | if ( $this->code_base->hasFunctionWithFQSEN( $closureFQSEN ) ) { |
| 203 | $func = $this->code_base->getFunctionByFQSEN( $closureFQSEN ); |
| 204 | $this->analyzeFunctionLike( $func ); |
| 205 | } else { |
| 206 | $this->debug( __METHOD__, 'closure doesn\'t exist' ); |
| 207 | } |
| 208 | $this->setCurTaintSafe(); |
| 209 | $this->setCachedData( $node ); |
| 210 | } |
| 211 | |
| 212 | /** |
| 213 | * @param Node $node |
| 214 | */ |
| 215 | public function visitFuncDecl( Node $node ): void { |
| 216 | $func = $this->context->getFunctionLikeInScope( $this->code_base ); |
| 217 | $this->analyzeFunctionLike( $func ); |
| 218 | } |
| 219 | |
| 220 | /** |
| 221 | * Visit a method declaration |
| 222 | * |
| 223 | * @param Node $node |
| 224 | */ |
| 225 | public function visitMethod( Node $node ): void { |
| 226 | $method = $this->context->getFunctionLikeInScope( $this->code_base ); |
| 227 | $this->analyzeFunctionLike( $method ); |
| 228 | } |
| 229 | |
| 230 | /** |
| 231 | * @param Node $node |
| 232 | */ |
| 233 | public function visitArrowFunc( Node $node ): void { |
| 234 | $this->visitClosure( $node ); |
| 235 | } |
| 236 | |
| 237 | /** |
| 238 | * Handles methods, functions and closures. |
| 239 | * |
| 240 | * @param FunctionInterface $func The func to analyze |
| 241 | */ |
| 242 | private function analyzeFunctionLike( FunctionInterface $func ): void { |
| 243 | if ( self::getFuncTaint( $func ) === null ) { |
| 244 | // If we still have no data, presumably the function doesn't return anything, so mark as safe. |
| 245 | if ( $func->hasReturn() || $func->hasYield() ) { |
| 246 | $this->debug( __METHOD__, "TODO: $func returns something but has no taint after analysis" ); |
| 247 | } |
| 248 | |
| 249 | // NOTE: If the method stores its arg to a class prop, and that class prop gets output later, |
| 250 | // the exec status of this won't be detected until the output is analyzed, we might miss some issues |
| 251 | // in the inbetween period. |
| 252 | self::doSetFuncTaint( $func, new FunctionTaintedness( Taintedness::newSafe() ) ); |
| 253 | } |
| 254 | } |
| 255 | |
| 256 | /** |
| 257 | * FooBar::class, presumably safe since class names cannot have special chars. |
| 258 | * |
| 259 | * @param Node $node |
| 260 | */ |
| 261 | public function visitClassName( Node $node ): void { |
| 262 | $this->setCurTaintSafe(); |
| 263 | $this->setCachedData( $node ); |
| 264 | } |
| 265 | |
| 266 | /** |
| 267 | * @note AST_THROW is an expression since PHP 8 |
| 268 | * @param Node $node |
| 269 | */ |
| 270 | public function visitThrow( Node $node ): void { |
| 271 | $this->setCurTaintSafe(); |
| 272 | $this->setCachedData( $node ); |
| 273 | } |
| 274 | |
| 275 | /** |
| 276 | * @param Node $node |
| 277 | */ |
| 278 | public function visitUnset( Node $node ): void { |
| 279 | $varNode = $node->children['var']; |
| 280 | if ( $varNode instanceof Node && $varNode->kind === \ast\AST_DIM ) { |
| 281 | $this->handleUnsetDim( $varNode ); |
| 282 | } |
| 283 | $this->setCurTaintSafe(); |
| 284 | } |
| 285 | |
| 286 | /** |
| 287 | * Analyzes expressions like unset( $arr['foo'] ) to infer shape mutations. |
| 288 | * @param Node $node |
| 289 | * @return void |
| 290 | */ |
| 291 | private function handleUnsetDim( Node $node ): void { |
| 292 | $expr = $node->children['expr']; |
| 293 | if ( !$expr instanceof Node ) { |
| 294 | // Syntax error. |
| 295 | return; |
| 296 | } |
| 297 | if ( $expr->kind !== \ast\AST_VAR ) { |
| 298 | // For now, we only handle a single offset. |
| 299 | // TODO actually recurse. |
| 300 | return; |
| 301 | } |
| 302 | |
| 303 | $keyNode = $node->children['dim']; |
| 304 | $key = $keyNode !== null ? $this->resolveOffset( $keyNode ) : null; |
| 305 | if ( $key instanceof Node ) { |
| 306 | // We can't tell what gets removed. |
| 307 | return; |
| 308 | } |
| 309 | |
| 310 | try { |
| 311 | $var = $this->getCtxN( $expr )->getVariable(); |
| 312 | } catch ( NodeException | IssueException $_ ) { |
| 313 | return; |
| 314 | } |
| 315 | |
| 316 | if ( $var instanceof GlobalVariable ) { |
| 317 | // Don't handle for now. |
| 318 | return; |
| 319 | } |
| 320 | |
| 321 | $curTaint = self::getTaintednessRaw( $var ); |
| 322 | if ( !$curTaint ) { |
| 323 | // Is this even possible? Don't do anything, just in case. |
| 324 | return; |
| 325 | } |
| 326 | self::setTaintednessRaw( $var, $curTaint->withoutKey( $key ) ); |
| 327 | } |
| 328 | |
| 329 | /** |
| 330 | * @param Node $node |
| 331 | */ |
| 332 | public function visitClone( Node $node ): void { |
| 333 | $this->curTaintWithError = clone $this->getTaintedness( $node->children['expr'] ); |
| 334 | $this->setCachedData( $node ); |
| 335 | } |
| 336 | |
| 337 | /** |
| 338 | * Assignment operators are: .=, +=, -=, /=, %=, *=, **=, ??=, |=, &=, ^=, <<=, >>= |
| 339 | * @param Node $node |
| 340 | */ |
| 341 | public function visitAssignOp( Node $node ): void { |
| 342 | $lhs = $node->children['var']; |
| 343 | if ( !$lhs instanceof Node ) { |
| 344 | // Syntax error, don't crash |
| 345 | $this->setCurTaintSafe(); |
| 346 | return; |
| 347 | } |
| 348 | $rhs = $node->children['expr']; |
| 349 | $lhsTaintedness = $this->getTaintedness( $lhs ); |
| 350 | $rhsTaintedness = $this->getTaintedness( $rhs ); |
| 351 | |
| 352 | if ( property_exists( $node, 'assignTaintMask' ) ) { |
| 353 | // @phan-suppress-next-line PhanUndeclaredProperty |
| 354 | $mask = $node->assignTaintMask; |
| 355 | // TODO Should we consume the value, since it depends on the union types? |
| 356 | } else { |
| 357 | $this->debug( __METHOD__, 'FIXME no preorder visit?' ); |
| 358 | $mask = SecurityCheckPlugin::ALL_TAINT_FLAGS; |
| 359 | } |
| 360 | |
| 361 | // Expand rhs to include implicit lhs ophand. |
| 362 | $allRHSTaint = $this->getBinOpTaint( |
| 363 | $lhsTaintedness->getTaintedness(), |
| 364 | $rhsTaintedness->getTaintedness(), |
| 365 | $node->flags, |
| 366 | $mask |
| 367 | ); |
| 368 | $allError = $lhsTaintedness->getError()->asMergedWith( $rhsTaintedness->getError() ); |
| 369 | $allLinks = $lhsTaintedness->getMethodLinks()->asMergedWith( $rhsTaintedness->getMethodLinks() ); |
| 370 | |
| 371 | $curTaint = $this->doVisitAssign( |
| 372 | $lhs, |
| 373 | $rhs, |
| 374 | $allRHSTaint, |
| 375 | $allError, |
| 376 | $allLinks, |
| 377 | $rhsTaintedness->getTaintedness(), |
| 378 | $rhsTaintedness->getMethodLinks() |
| 379 | ); |
| 380 | // TODO Links and error? |
| 381 | $this->curTaintWithError = new TaintednessWithError( $curTaint, new CausedByLines(), MethodLinks::newEmpty() ); |
| 382 | $this->setCachedData( $node ); |
| 383 | } |
| 384 | |
| 385 | /** |
| 386 | * `static $var = 'foo'` Handle it as an assignment of a safe value, to initialize the taintedness |
| 387 | * on $var. Ideally, we'd want to retain any taintedness on this object, but it's currently impossible |
| 388 | * (upstream has the same limitation with union types). |
| 389 | * |
| 390 | * @param Node $node |
| 391 | */ |
| 392 | public function visitStatic( Node $node ): void { |
| 393 | $var = $this->getCtxN( $node->children['var'] )->getVariable(); |
| 394 | $this->ensureTaintednessIsSet( $var ); |
| 395 | } |
| 396 | |
| 397 | /** |
| 398 | * @param Node $node |
| 399 | */ |
| 400 | public function visitAssignRef( Node $node ): void { |
| 401 | $this->visitAssign( $node ); |
| 402 | } |
| 403 | |
| 404 | /** |
| 405 | * @param Node $node |
| 406 | */ |
| 407 | public function visitAssign( Node $node ): void { |
| 408 | $lhs = $node->children['var']; |
| 409 | if ( !$lhs instanceof Node ) { |
| 410 | // Syntax error, don't crash |
| 411 | $this->setCurTaintSafe(); |
| 412 | return; |
| 413 | } |
| 414 | $rhs = $node->children['expr']; |
| 415 | |
| 416 | $rhsTaintedness = $this->getTaintedness( $rhs ); |
| 417 | |
| 418 | $curTaint = $this->doVisitAssign( |
| 419 | $lhs, |
| 420 | $rhs, |
| 421 | clone $rhsTaintedness->getTaintedness(), |
| 422 | $rhsTaintedness->getError(), |
| 423 | $rhsTaintedness->getMethodLinks(), |
| 424 | $rhsTaintedness->getTaintedness(), |
| 425 | $rhsTaintedness->getMethodLinks() |
| 426 | ); |
| 427 | // TODO Links and error? |
| 428 | $this->curTaintWithError = new TaintednessWithError( $curTaint, new CausedByLines(), new MethodLinks() ); |
| 429 | $this->setCachedData( $node ); |
| 430 | } |
| 431 | |
| 432 | /** |
| 433 | * @param Node $lhs |
| 434 | * @param Node|mixed $rhs |
| 435 | * @param Taintedness $rhsTaint |
| 436 | * @param CausedByLines $rhsError |
| 437 | * @param MethodLinks $rhsLinks |
| 438 | * @param Taintedness $errorTaint |
| 439 | * @param MethodLinks $errorLinks |
| 440 | * @return Taintedness |
| 441 | */ |
| 442 | private function doVisitAssign( |
| 443 | Node $lhs, |
| 444 | $rhs, |
| 445 | Taintedness $rhsTaint, |
| 446 | CausedByLines $rhsError, |
| 447 | MethodLinks $rhsLinks, |
| 448 | Taintedness $errorTaint, |
| 449 | MethodLinks $errorLinks |
| 450 | ): Taintedness { |
| 451 | $vis = new TaintednessAssignVisitor( |
| 452 | $this->code_base, |
| 453 | $this->context, |
| 454 | clone $rhsTaint, |
| 455 | $rhsError, |
| 456 | $rhsLinks, |
| 457 | $errorTaint, |
| 458 | $errorLinks, |
| 459 | function () use ( $rhs ): bool { |
| 460 | return $this->nodeIsArray( $rhs ); |
| 461 | } |
| 462 | ); |
| 463 | $vis( $lhs ); |
| 464 | return $rhsTaint; |
| 465 | } |
| 466 | |
| 467 | /** |
| 468 | * @param Node $node |
| 469 | */ |
| 470 | public function visitBinaryOp( Node $node ): void { |
| 471 | $lhs = $node->children['left']; |
| 472 | $rhs = $node->children['right']; |
| 473 | $mask = $this->getBinOpTaintMask( $node, $lhs, $rhs ); |
| 474 | if ( $mask === SecurityCheckPlugin::NO_TAINT ) { |
| 475 | // If the operation is safe, don't waste time analyzing children.This might also create bugs |
| 476 | // like the test undeclaredvar2 |
| 477 | $this->setCurTaintSafe(); |
| 478 | $this->setCachedData( $node ); |
| 479 | return; |
| 480 | } |
| 481 | $leftTaint = $this->getTaintedness( $lhs ); |
| 482 | $rightTaint = $this->getTaintedness( $rhs ); |
| 483 | $curTaint = $this->getBinOpTaint( |
| 484 | $leftTaint->getTaintedness(), |
| 485 | $rightTaint->getTaintedness(), |
| 486 | $node->flags, |
| 487 | $mask |
| 488 | ); |
| 489 | $this->curTaintWithError = new TaintednessWithError( |
| 490 | $curTaint, |
| 491 | $leftTaint->getError()->asMergedWith( $rightTaint->getError() ), |
| 492 | $leftTaint->getMethodLinks()->asMergedWith( $rightTaint->getMethodLinks() ) |
| 493 | ); |
| 494 | $this->setCachedData( $node ); |
| 495 | } |
| 496 | |
| 497 | /** |
| 498 | * Get the taintedness of a binop, depending on the op type, applying the given flags |
| 499 | * @param Taintedness $leftTaint |
| 500 | * @param Taintedness $rightTaint |
| 501 | * @param int $op Represented by a flags in \ast\flags |
| 502 | * @param int $mask |
| 503 | * @return Taintedness |
| 504 | */ |
| 505 | private function getBinOpTaint( |
| 506 | Taintedness $leftTaint, |
| 507 | Taintedness $rightTaint, |
| 508 | int $op, |
| 509 | int $mask |
| 510 | ): Taintedness { |
| 511 | if ( $mask === SecurityCheckPlugin::NO_TAINT ) { |
| 512 | return Taintedness::newSafe(); |
| 513 | } |
| 514 | if ( $op === \ast\flags\BINARY_ADD ) { |
| 515 | // HACK: This means that a node can be array, so assume array plus |
| 516 | return $leftTaint->asArrayPlusWith( $rightTaint ); |
| 517 | } |
| 518 | return $leftTaint->asMergedWith( $rightTaint )->asCollapsed()->withOnly( $mask ); |
| 519 | } |
| 520 | |
| 521 | /** |
| 522 | * @param Node $node |
| 523 | */ |
| 524 | public function visitDim( Node $node ): void { |
| 525 | $varNode = $node->children['expr']; |
| 526 | if ( !$varNode instanceof Node ) { |
| 527 | // Accessing offset of a string literal |
| 528 | $this->setCurTaintSafe(); |
| 529 | $this->setCachedData( $node ); |
| 530 | return; |
| 531 | } |
| 532 | $nodeTaint = $this->getTaintednessNode( $varNode ); |
| 533 | if ( $node->children['dim'] === null ) { |
| 534 | // This should only happen in assignments: $x[] = 'foo'. Just return |
| 535 | // the taint of the whole object. |
| 536 | $this->curTaintWithError = clone $nodeTaint; |
| 537 | $this->setCachedData( $node ); |
| 538 | return; |
| 539 | } |
| 540 | $offset = $this->resolveOffset( $node->children['dim'] ); |
| 541 | $this->curTaintWithError = new TaintednessWithError( |
| 542 | clone $nodeTaint->getTaintedness()->getTaintednessForOffsetOrWhole( $offset ), |
| 543 | $nodeTaint->getError(), |
| 544 | $nodeTaint->getMethodLinks()->getForDim( $offset ) |
| 545 | ); |
| 546 | $this->setCachedData( $node ); |
| 547 | } |
| 548 | |
| 549 | /** |
| 550 | * @param Node $node |
| 551 | */ |
| 552 | public function visitPrint( Node $node ): void { |
| 553 | $this->visitEcho( $node ); |
| 554 | } |
| 555 | |
| 556 | /** |
| 557 | * This is for exit() and die(). If they're passed an argument, they behave the |
| 558 | * same as print. |
| 559 | * @param Node $node |
| 560 | */ |
| 561 | public function visitExit( Node $node ): void { |
| 562 | $this->visitEcho( $node ); |
| 563 | } |
| 564 | |
| 565 | /** |
| 566 | * Visits the backtick operator. Note that shell_exec() has a simple AST_CALL node. |
| 567 | * @param Node $node |
| 568 | */ |
| 569 | public function visitShellExec( Node $node ): void { |
| 570 | $this->visitSimpleSinkAndPropagate( |
| 571 | $node, |
| 572 | SecurityCheckPlugin::SHELL_EXEC_TAINT, |
| 573 | 'Backtick shell execution operator contains user controlled arg' |
| 574 | ); |
| 575 | // Its unclear if we should consider this tainted or not |
| 576 | $this->curTaintWithError = new TaintednessWithError( |
| 577 | Taintedness::newTainted(), |
| 578 | new CausedByLines(), |
| 579 | MethodLinks::newEmpty() |
| 580 | ); |
| 581 | $this->setCachedData( $node ); |
| 582 | } |
| 583 | |
| 584 | /** |
| 585 | * @param Node $node |
| 586 | */ |
| 587 | public function visitIncludeOrEval( Node $node ): void { |
| 588 | if ( $node->flags === \ast\flags\EXEC_EVAL ) { |
| 589 | $taintValue = SecurityCheckPlugin::CODE_EXEC_TAINT; |
| 590 | $msg = 'The code supplied to `eval` is user controlled'; |
| 591 | } else { |
| 592 | $taintValue = SecurityCheckPlugin::PATH_EXEC_TAINT; |
| 593 | $msg = 'The included path is user controlled'; |
| 594 | } |
| 595 | $this->visitSimpleSinkAndPropagate( $node, $taintValue, $msg ); |
| 596 | // Strictly speaking we have no idea if the result |
| 597 | // of an eval() or require() is safe. But given that we |
| 598 | // don't know, and at least in the require() case its |
| 599 | // fairly likely to be safe, no point in complaining. |
| 600 | $this->setCurTaintSafe(); |
| 601 | $this->setCachedData( $node ); |
| 602 | } |
| 603 | |
| 604 | /** |
| 605 | * Also handles exit() and print |
| 606 | * |
| 607 | * We assume a web based system, where outputting HTML via echo |
| 608 | * is bad. This will have false positives in a CLI environment. |
| 609 | * |
| 610 | * @param Node $node |
| 611 | */ |
| 612 | public function visitEcho( Node $node ): void { |
| 613 | $this->visitSimpleSinkAndPropagate( |
| 614 | $node, |
| 615 | SecurityCheckPlugin::HTML_EXEC_TAINT, |
| 616 | 'Echoing expression that was not html escaped' |
| 617 | ); |
| 618 | $this->setCurTaintSafe(); |
| 619 | $this->setCachedData( $node ); |
| 620 | } |
| 621 | |
| 622 | /** |
| 623 | * @param Node $node |
| 624 | * @param int $sinkTaintInt |
| 625 | * @param string $issueMsg |
| 626 | */ |
| 627 | private function visitSimpleSinkAndPropagate( Node $node, int $sinkTaintInt, string $issueMsg ): void { |
| 628 | if ( !isset( $node->children['expr'] ) ) { |
| 629 | return; |
| 630 | } |
| 631 | $expr = $node->children['expr']; |
| 632 | $exprTaint = $this->getTaintedness( $expr ); |
| 633 | |
| 634 | $sinkTaint = new Taintedness( $sinkTaintInt ); |
| 635 | $rhsTaint = $exprTaint->getTaintedness(); |
| 636 | $this->maybeEmitIssue( |
| 637 | $sinkTaint, |
| 638 | $rhsTaint, |
| 639 | "$issueMsg{DETAILS}", |
| 640 | [ $exprTaint->getError() ] |
| 641 | ); |
| 642 | |
| 643 | if ( $expr instanceof Node && !$rhsTaint->has( Taintedness::flagsAsExecToYesTaint( $sinkTaintInt ) ) ) { |
| 644 | $this->backpropagateArgTaint( $expr, $sinkTaint ); |
| 645 | } |
| 646 | } |
| 647 | |
| 648 | /** |
| 649 | * @param Node $node |
| 650 | */ |
| 651 | public function visitStaticCall( Node $node ): void { |
| 652 | $this->visitMethodCall( $node ); |
| 653 | } |
| 654 | |
| 655 | /** |
| 656 | * @param Node $node |
| 657 | */ |
| 658 | public function visitNew( Node $node ): void { |
| 659 | $ctxNode = $this->getCtxN( $node ); |
| 660 | if ( !$node->children['class'] instanceof Node ) { |
| 661 | // Syntax error, don't crash |
| 662 | $this->setCurTaintSafe(); |
| 663 | return; |
| 664 | } |
| 665 | |
| 666 | // We check the __construct() method first, but the |
| 667 | // final resulting taint is from the __toString() |
| 668 | // method. This is a little hacky. |
| 669 | try { |
| 670 | // First do __construct() |
| 671 | $constructor = $ctxNode->getMethod( |
| 672 | '__construct', |
| 673 | false, |
| 674 | false, |
| 675 | true |
| 676 | ); |
| 677 | } catch ( NodeException | CodeBaseException | IssueException $_ ) { |
| 678 | $constructor = null; |
| 679 | } |
| 680 | |
| 681 | if ( $constructor ) { |
| 682 | $this->handleMethodCall( |
| 683 | $constructor, |
| 684 | $constructor->getFQSEN(), |
| 685 | $node->children['args']->children, |
| 686 | false |
| 687 | ); |
| 688 | } |
| 689 | |
| 690 | // Now return __toString() |
| 691 | try { |
| 692 | $clazzes = $ctxNode->getClassList( |
| 693 | false, |
| 694 | ContextNode::CLASS_LIST_ACCEPT_OBJECT_OR_CLASS_NAME, |
| 695 | null, |
| 696 | false |
| 697 | ); |
| 698 | } catch ( CodeBaseException | IssueException $e ) { |
| 699 | $this->debug( __METHOD__, 'Cannot get class: ' . $this->getDebugInfo( $e ) ); |
| 700 | $this->setCurTaintUnknown(); |
| 701 | $this->setCachedData( $node ); |
| 702 | return; |
| 703 | } |
| 704 | |
| 705 | foreach ( $clazzes as $clazz ) { |
| 706 | try { |
| 707 | $toString = $clazz->getMethodByName( $this->code_base, '__toString' ); |
| 708 | } catch ( CodeBaseException $_ ) { |
| 709 | // No __toString() in this class |
| 710 | continue; |
| 711 | } |
| 712 | |
| 713 | $toStringTaint = $this->handleMethodCall( $toString, $toString->getFQSEN(), [] ); |
| 714 | if ( !$this->curTaintWithError ) { |
| 715 | $this->curTaintWithError = $toStringTaint; |
| 716 | } else { |
| 717 | $this->curTaintWithError->mergeWith( $toStringTaint ); |
| 718 | } |
| 719 | } |
| 720 | |
| 721 | // If we find no __toString(), then presumably the object can't be outputted, so should be safe. |
| 722 | $this->curTaintWithError ??= TaintednessWithError::newEmpty(); |
| 723 | |
| 724 | $this->setCachedData( $node ); |
| 725 | } |
| 726 | |
| 727 | /** |
| 728 | * @param Node $node |
| 729 | */ |
| 730 | public function visitMethodCall( Node $node ): void { |
| 731 | $methodName = $node->children['method']; |
| 732 | $isStatic = $node->kind === \ast\AST_STATIC_CALL; |
| 733 | try { |
| 734 | $method = $this->getCtxN( $node )->getMethod( $methodName, $isStatic, true ); |
| 735 | } catch ( NodeException | CodeBaseException | IssueException $e ) { |
| 736 | $this->debug( __METHOD__, "Cannot find method in node. " . $this->getDebugInfo( $e ) ); |
| 737 | $this->setCurTaintUnknown(); |
| 738 | $this->setCachedData( $node ); |
| 739 | return; |
| 740 | } |
| 741 | |
| 742 | $this->analyzeCallNode( $node, [ $method ] ); |
| 743 | } |
| 744 | |
| 745 | /** |
| 746 | * @param Node $node |
| 747 | * @param iterable<FunctionInterface> $funcs |
| 748 | */ |
| 749 | protected function analyzeCallNode( Node $node, iterable $funcs ): void { |
| 750 | $args = $node->children['args']->children; |
| 751 | foreach ( $funcs as $func ) { |
| 752 | // No point in analyzing abstract function declarations |
| 753 | if ( !$func instanceof FunctionLikeDeclarationType ) { |
| 754 | $callTaint = $this->handleMethodCall( $func, $func->getFQSEN(), $args ); |
| 755 | if ( !$this->curTaintWithError ) { |
| 756 | $this->curTaintWithError = $callTaint; |
| 757 | } else { |
| 758 | $this->curTaintWithError->mergeWith( $callTaint ); |
| 759 | } |
| 760 | } |
| 761 | } |
| 762 | $this->curTaintWithError ??= TaintednessWithError::newEmpty(); |
| 763 | $this->setCachedData( $node ); |
| 764 | } |
| 765 | |
| 766 | /** |
| 767 | * @param Node $node |
| 768 | */ |
| 769 | public function visitNullsafeMethodCall( Node $node ): void { |
| 770 | $this->visitMethodCall( $node ); |
| 771 | } |
| 772 | |
| 773 | /** |
| 774 | * A function call |
| 775 | * |
| 776 | * @param Node $node |
| 777 | */ |
| 778 | public function visitCall( Node $node ): void { |
| 779 | $funcs = $this->getCtxN( $node->children['expr'] )->getFunctionFromNode(); |
| 780 | |
| 781 | if ( !$funcs ) { |
| 782 | $this->setCurTaintUnknown(); |
| 783 | $this->setCachedData( $node ); |
| 784 | return; |
| 785 | } |
| 786 | |
| 787 | $this->analyzeCallNode( $node, $funcs ); |
| 788 | } |
| 789 | |
| 790 | /** |
| 791 | * A variable (e.g. $foo) |
| 792 | * |
| 793 | * This always considers superglobals as tainted |
| 794 | * |
| 795 | * @param Node $node |
| 796 | */ |
| 797 | public function visitVar( Node $node ): void { |
| 798 | $varName = $this->getCtxN( $node )->getVariableName(); |
| 799 | if ( $varName === '' ) { |
| 800 | // Something that phan can't understand, e.g. `$$foo` with unknown `$foo`. |
| 801 | $this->setCurTaintUnknown(); |
| 802 | return; |
| 803 | } |
| 804 | |
| 805 | $hardcodedTaint = $this->getHardcodedTaintednessForVar( $varName ); |
| 806 | if ( $hardcodedTaint ) { |
| 807 | $this->curTaintWithError = new TaintednessWithError( |
| 808 | $hardcodedTaint, |
| 809 | new CausedByLines(), |
| 810 | MethodLinks::newEmpty() |
| 811 | ); |
| 812 | $this->setCachedData( $node ); |
| 813 | return; |
| 814 | } |
| 815 | if ( !$this->context->getScope()->hasVariableWithName( $varName ) ) { |
| 816 | // Probably the var just isn't in scope yet. |
| 817 | // $this->debug( __METHOD__, "No var with name \$$varName in scope (Setting Unknown taint)" ); |
| 818 | $this->setCurTaintUnknown(); |
| 819 | $this->setCachedData( $node ); |
| 820 | return; |
| 821 | } |
| 822 | $variableObj = $this->context->getScope()->getVariableByName( $varName ); |
| 823 | $this->curTaintWithError = new TaintednessWithError( |
| 824 | $this->getTaintednessPhanObj( $variableObj ), |
| 825 | self::getCausedByRawCloneOrEmpty( $variableObj ), |
| 826 | self::getMethodLinksCloneOrEmpty( $variableObj ) |
| 827 | ); |
| 828 | $this->setCachedData( $node ); |
| 829 | } |
| 830 | |
| 831 | /** |
| 832 | * If we hardcode taintedness for the given var name, return that taintedness; return null otherwise. |
| 833 | * This is currently used for superglobals, since they're always tainted, regardless of whether they're in |
| 834 | * the current scope: `function foo() use ($argv)` puts $argv in the local scope, but it retains its |
| 835 | * taintedness (see test closure2). |
| 836 | * |
| 837 | * @param string $varName |
| 838 | * @return Taintedness|null |
| 839 | */ |
| 840 | private function getHardcodedTaintednessForVar( string $varName ): ?Taintedness { |
| 841 | switch ( $varName ) { |
| 842 | case '_GET': |
| 843 | case '_POST': |
| 844 | case 'argc': |
| 845 | case 'argv': |
| 846 | case 'http_response_header': |
| 847 | case '_COOKIE': |
| 848 | case '_REQUEST': |
| 849 | // It's not entirely clear what $_ENV and $_SESSION should be considered |
| 850 | case '_ENV': |
| 851 | case '_SESSION': |
| 852 | // Hopefully we don't need to specify all keys for $_SERVER... |
| 853 | case '_SERVER': |
| 854 | return Taintedness::newTainted(); |
| 855 | case '_FILES': |
| 856 | $ret = Taintedness::newSafe(); |
| 857 | $ret->addKeysTaintedness( SecurityCheckPlugin::YES_TAINT ); |
| 858 | $elTaint = Taintedness::newFromArray( [ |
| 859 | 'name' => Taintedness::newTainted(), |
| 860 | 'type' => Taintedness::newTainted(), |
| 861 | 'tmp_name' => Taintedness::newSafe(), |
| 862 | 'error' => Taintedness::newSafe(), |
| 863 | 'size' => Taintedness::newSafe(), |
| 864 | ] ); |
| 865 | // Use 'null' as fake offset to set unknownDims |
| 866 | $ret->setOffsetTaintedness( null, $elTaint ); |
| 867 | return $ret; |
| 868 | case 'GLOBALS': |
| 869 | // Ideally this would recurse properly, but hopefully nobody is using $GLOBALS in complex ways |
| 870 | // that wouldn't be covered by this approximation. |
| 871 | |
| 872 | $filesTaintedness = $this->getHardcodedTaintednessForVar( '_FILES' ); |
| 873 | assert( $filesTaintedness !== null ); |
| 874 | return Taintedness::newFromArray( [ |
| 875 | '_GET' => Taintedness::newTainted(), |
| 876 | '_POST' => Taintedness::newTainted(), |
| 877 | '_SERVER' => Taintedness::newTainted(), |
| 878 | '_COOKIE' => Taintedness::newTainted(), |
| 879 | '_SESSION' => Taintedness::newTainted(), |
| 880 | '_REQUEST' => Taintedness::newTainted(), |
| 881 | '_ENV' => Taintedness::newTainted(), |
| 882 | '_FILES' => $filesTaintedness, |
| 883 | 'GLOBALS' => Taintedness::newTainted() |
| 884 | ] ); |
| 885 | default: |
| 886 | return null; |
| 887 | } |
| 888 | } |
| 889 | |
| 890 | /** |
| 891 | * A global declaration. Assume most globals are untainted. |
| 892 | * |
| 893 | * @param Node $node |
| 894 | */ |
| 895 | public function visitGlobal( Node $node ): void { |
| 896 | assert( isset( $node->children['var'] ) && $node->children['var']->kind === \ast\AST_VAR ); |
| 897 | |
| 898 | $varName = $node->children['var']->children['name']; |
| 899 | if ( !is_string( $varName ) || !$this->context->getScope()->hasVariableWithName( $varName ) ) { |
| 900 | // Something like global $$indirectReference; or the variable wasn't created somehow |
| 901 | return; |
| 902 | } |
| 903 | // Copy taintedness data from the actual global into the scoped clone |
| 904 | $gvar = $this->context->getScope()->getVariableByName( $varName ); |
| 905 | if ( !$gvar instanceof GlobalVariable ) { |
| 906 | // Likely a superglobal, nothing to do. |
| 907 | return; |
| 908 | } |
| 909 | $actualGlobal = $gvar->getElement(); |
| 910 | self::setTaintednessRaw( $gvar, self::getTaintednessRawClone( $actualGlobal ) ?: Taintedness::newSafe() ); |
| 911 | self::setCausedByRaw( $gvar, self::getCausedByRawCloneOrEmpty( $actualGlobal ) ); |
| 912 | self::setMethodLinks( $gvar, self::getMethodLinksCloneOrEmpty( $actualGlobal ) ); |
| 913 | } |
| 914 | |
| 915 | /** |
| 916 | * Set the taint of the function based on what's returned |
| 917 | * |
| 918 | * This attempts to match the return value up to the argument |
| 919 | * to figure out which argument might taint the function. This won't |
| 920 | * work in complex cases though. |
| 921 | * |
| 922 | * @param Node $node |
| 923 | */ |
| 924 | public function visitReturn( Node $node ): void { |
| 925 | if ( !$this->context->isInFunctionLikeScope() ) { |
| 926 | // E.g. a file that can be included. |
| 927 | $this->setCurTaintUnknown(); |
| 928 | $this->setCachedData( $node ); |
| 929 | return; |
| 930 | } |
| 931 | |
| 932 | $curFunc = $this->context->getFunctionLikeInScope( $this->code_base ); |
| 933 | |
| 934 | $this->setFuncTaintFromReturn( $node, $curFunc ); |
| 935 | |
| 936 | if ( $node->children['expr'] instanceof Node ) { |
| 937 | $collector = new ReturnObjectsCollectVisitor( $this->code_base, $this->context ); |
| 938 | self::addRetObjs( $curFunc, $collector->collectFromNode( $node ) ); |
| 939 | } |
| 940 | } |
| 941 | |
| 942 | /** |
| 943 | * @param Node $node |
| 944 | * @param FunctionInterface $func |
| 945 | */ |
| 946 | private function setFuncTaintFromReturn( Node $node, FunctionInterface $func ): void { |
| 947 | assert( $node->kind === \ast\AST_RETURN ); |
| 948 | $retExpr = $node->children['expr']; |
| 949 | $retTaintednessWithError = $this->getTaintedness( $retExpr ); |
| 950 | // Ensure we don't transmit any EXEC flag. |
| 951 | $retTaintedness = $retTaintednessWithError->getTaintedness()->withOnly( SecurityCheckPlugin::ALL_TAINT ); |
| 952 | if ( !$retExpr instanceof Node ) { |
| 953 | assert( $retTaintedness->isSafe() ); |
| 954 | $this->ensureFuncTaintIsSet( $func ); |
| 955 | return; |
| 956 | } |
| 957 | |
| 958 | $overallFuncTaint = $retTaintedness; |
| 959 | // Note, it's important that we only use the real type here (e.g. from typehints) and NOT |
| 960 | // the PHPDoc type, as it may be wrong. |
| 961 | $retTaintMask = $this->getTaintMaskForType( $func->getRealReturnType() ); |
| 962 | if ( $retTaintMask !== null ) { |
| 963 | $overallFuncTaint->keepOnly( $retTaintMask->get() ); |
| 964 | } |
| 965 | |
| 966 | $paramTaint = new FunctionTaintedness( $overallFuncTaint ); |
| 967 | $funcError = new FunctionCausedByLines(); |
| 968 | |
| 969 | $links = $retTaintednessWithError->getMethodLinks(); |
| 970 | $retError = $retTaintednessWithError->getError(); |
| 971 | // Note, not forCaller, as that doesn't see variadic parameters |
| 972 | $calleeParamList = $func->getParameterList(); |
| 973 | foreach ( $calleeParamList as $i => $param ) { |
| 974 | $presTaint = $retTaintMask === null || !$retTaintMask->isSafe() |
| 975 | ? $links->asPreservedTaintednessForFuncParam( $func, $i ) |
| 976 | : PreservedTaintedness::newEmpty(); |
| 977 | $paramError = $retError->asFilteredForFuncAndParam( $func, $i ); |
| 978 | if ( $param->isVariadic() ) { |
| 979 | $paramTaint->setVariadicParamPreservedTaint( $i, $presTaint ); |
| 980 | $funcError->setVariadicParamPreservedLines( $i, $paramError ); |
| 981 | } else { |
| 982 | $paramTaint->setParamPreservedTaint( $i, $presTaint ); |
| 983 | $funcError->setParamPreservedLines( $i, $paramError ); |
| 984 | } |
| 985 | } |
| 986 | |
| 987 | $funcError->setGenericLines( $retError->getLinesForGenericReturn() ); |
| 988 | $this->addFuncTaint( $func, $paramTaint ); |
| 989 | $newFuncTaint = self::getFuncTaint( $func ); |
| 990 | assert( $newFuncTaint !== null ); |
| 991 | $this->maybeAddFuncError( $func, null, $paramTaint, $newFuncTaint, $links ); |
| 992 | // Note: adding the error after setting the taintedness means that the return line comes before |
| 993 | // the other lines |
| 994 | $this->mergeFuncError( $func, $funcError, $newFuncTaint ); |
| 995 | } |
| 996 | |
| 997 | /** |
| 998 | * @param Node $node |
| 999 | */ |
| 1000 | public function visitArray( Node $node ): void { |
| 1001 | $curTaint = Taintedness::newSafe(); |
| 1002 | $curError = new CausedByLines(); |
| 1003 | $links = MethodLinks::newEmpty(); |
| 1004 | // Current numeric key in the array |
| 1005 | $curNumKey = 0; |
| 1006 | foreach ( $node->children as $child ) { |
| 1007 | if ( $child === null ) { |
| 1008 | // Happens for list( , $x ) = foo() |
| 1009 | continue; |
| 1010 | } |
| 1011 | if ( $child->kind === \ast\AST_UNPACK ) { |
| 1012 | // PHP 7.4's in-place unpacking. |
| 1013 | // TODO Do something? |
| 1014 | continue; |
| 1015 | } |
| 1016 | assert( $child->kind === \ast\AST_ARRAY_ELEM ); |
| 1017 | $key = $child->children['key']; |
| 1018 | $keyTaintAll = $this->getTaintedness( $key ); |
| 1019 | $keyTaint = $keyTaintAll->getTaintedness(); |
| 1020 | $value = $child->children['value']; |
| 1021 | $valTaintAll = $this->getTaintedness( $value ); |
| 1022 | $valTaint = $valTaintAll->getTaintedness(); |
| 1023 | $sqlTaint = SecurityCheckPlugin::SQL_TAINT; |
| 1024 | |
| 1025 | if ( |
| 1026 | ( $keyTaint->has( $sqlTaint ) ) || ( |
| 1027 | ( $key === null || $this->nodeCanBeIntKey( $key ) ) |
| 1028 | && $valTaint->has( $sqlTaint ) |
| 1029 | && $this->nodeCanBeString( $value ) |
| 1030 | ) |
| 1031 | ) { |
| 1032 | $curTaint->add( SecurityCheckPlugin::SQL_NUMKEY_TAINT ); |
| 1033 | } |
| 1034 | // FIXME This will fail with in-place spread and when some numeric keys are specified |
| 1035 | // explicitly (at least). |
| 1036 | $offset = $key ?? $curNumKey++; |
| 1037 | $offset = $this->resolveOffset( $offset ); |
| 1038 | $curTaint->setOffsetTaintedness( $offset, $valTaint ); |
| 1039 | $curTaint->addKeysTaintedness( $keyTaint->get() ); |
| 1040 | $curError->mergeWith( $keyTaintAll->getError() ); |
| 1041 | $curError->mergeWith( $valTaintAll->getError() ); |
| 1042 | $links->addKeysLinks( $keyTaintAll->getMethodLinks()->getLinksCollapsing() ); |
| 1043 | $links->setAtDim( $offset, $valTaintAll->getMethodLinks() ); |
| 1044 | } |
| 1045 | $this->curTaintWithError = new TaintednessWithError( $curTaint, $curError, $links ); |
| 1046 | $this->setCachedData( $node ); |
| 1047 | } |
| 1048 | |
| 1049 | /** |
| 1050 | * @param Node $node |
| 1051 | */ |
| 1052 | public function visitClassConst( Node $node ): void { |
| 1053 | $this->setCurTaintSafe(); |
| 1054 | $this->setCachedData( $node ); |
| 1055 | } |
| 1056 | |
| 1057 | /** |
| 1058 | * @param Node $node |
| 1059 | */ |
| 1060 | public function visitConst( Node $node ): void { |
| 1061 | // We are going to assume nobody is doing stupid stuff like |
| 1062 | // define( "foo", $_GET['bar'] ); |
| 1063 | $this->setCurTaintSafe(); |
| 1064 | $this->setCachedData( $node ); |
| 1065 | } |
| 1066 | |
| 1067 | /** |
| 1068 | * The :: operator (for props) |
| 1069 | * @param Node $node |
| 1070 | */ |
| 1071 | public function visitStaticProp( Node $node ): void { |
| 1072 | $prop = $this->getPropFromNode( $node ); |
| 1073 | if ( !$prop ) { |
| 1074 | $this->setCurTaintUnknown(); |
| 1075 | return; |
| 1076 | } |
| 1077 | $this->curTaintWithError = new TaintednessWithError( |
| 1078 | $this->getTaintednessPhanObj( $prop ), |
| 1079 | self::getCausedByRawCloneOrEmpty( $prop ), |
| 1080 | self::getMethodLinksCloneOrEmpty( $prop ) |
| 1081 | ); |
| 1082 | $this->setCachedData( $node ); |
| 1083 | } |
| 1084 | |
| 1085 | /** |
| 1086 | * The -> operator (when not a method call) |
| 1087 | * @param Node $node |
| 1088 | */ |
| 1089 | public function visitProp( Node $node ): void { |
| 1090 | $nodeExpr = $node->children['expr']; |
| 1091 | if ( !$nodeExpr instanceof Node ) { |
| 1092 | // Syntax error. |
| 1093 | $this->setCurTaintSafe(); |
| 1094 | return; |
| 1095 | } |
| 1096 | |
| 1097 | // If the LHS expr can potentially be a stdClass, merge in its taintedness as well. |
| 1098 | // TODO Improve this (should similar to array offsets) |
| 1099 | $foundStdClass = false; |
| 1100 | $exprType = $this->getNodeType( $nodeExpr ); |
| 1101 | $stdClassType = FullyQualifiedClassName::getStdClassFQSEN()->asType(); |
| 1102 | if ( $exprType && $exprType->hasType( $stdClassType ) ) { |
| 1103 | $exprTaint = clone $this->getTaintedness( $nodeExpr ); |
| 1104 | $this->curTaintWithError = new TaintednessWithError( |
| 1105 | $exprTaint->getTaintedness(), |
| 1106 | $exprTaint->getError(), |
| 1107 | // TODO Links? |
| 1108 | MethodLinks::newEmpty() |
| 1109 | ); |
| 1110 | $foundStdClass = true; |
| 1111 | } |
| 1112 | |
| 1113 | $prop = $this->getPropFromNode( $node ); |
| 1114 | if ( !$prop ) { |
| 1115 | if ( !$foundStdClass ) { |
| 1116 | $this->setCurTaintUnknown(); |
| 1117 | } |
| 1118 | $this->setCachedData( $node ); |
| 1119 | return; |
| 1120 | } |
| 1121 | |
| 1122 | $objTaint = $this->getTaintednessPhanObj( $prop ); |
| 1123 | $objError = self::getCausedByRawCloneOrEmpty( $prop ); |
| 1124 | $objLinks = self::getMethodLinksCloneOrEmpty( $prop ); |
| 1125 | |
| 1126 | if ( $foundStdClass ) { |
| 1127 | $newTaint = $this->curTaintWithError->getTaintedness() |
| 1128 | ->asMergedWith( $objTaint->without( SecurityCheckPlugin::UNKNOWN_TAINT ) ); |
| 1129 | $this->curTaintWithError = new TaintednessWithError( |
| 1130 | $newTaint, |
| 1131 | $this->curTaintWithError->getError()->asMergedWith( $objError ), |
| 1132 | $this->curTaintWithError->getMethodLinks()->asMergedWith( $objLinks ) |
| 1133 | ); |
| 1134 | } else { |
| 1135 | $this->curTaintWithError = new TaintednessWithError( $objTaint, $objError, $objLinks ); |
| 1136 | } |
| 1137 | |
| 1138 | $this->setCachedData( $node ); |
| 1139 | } |
| 1140 | |
| 1141 | /** |
| 1142 | * @param Node $node |
| 1143 | */ |
| 1144 | public function visitNullsafeProp( Node $node ): void { |
| 1145 | $this->visitProp( $node ); |
| 1146 | } |
| 1147 | |
| 1148 | /** |
| 1149 | * Ternary operator. |
| 1150 | * @param Node $node |
| 1151 | */ |
| 1152 | public function visitConditional( Node $node ): void { |
| 1153 | if ( $node->children['true'] === null ) { |
| 1154 | // $foo ?: $bar; |
| 1155 | $trueTaint = $this->getTaintedness( $node->children['cond'] ); |
| 1156 | } else { |
| 1157 | $trueTaint = $this->getTaintedness( $node->children['true'] ); |
| 1158 | } |
| 1159 | $falseTaint = $this->getTaintedness( $node->children['false'] ); |
| 1160 | $this->curTaintWithError = $trueTaint->asMergedWith( $falseTaint ); |
| 1161 | $this->setCachedData( $node ); |
| 1162 | } |
| 1163 | |
| 1164 | /** |
| 1165 | * @param Node $node |
| 1166 | */ |
| 1167 | public function visitName( Node $node ): void { |
| 1168 | $this->setCurTaintSafe(); |
| 1169 | } |
| 1170 | |
| 1171 | /** |
| 1172 | * This is e.g. for class X implements Name,List |
| 1173 | * |
| 1174 | * @param Node $node |
| 1175 | */ |
| 1176 | public function visitNameList( Node $node ): void { |
| 1177 | $this->setCurTaintSafe(); |
| 1178 | } |
| 1179 | |
| 1180 | /** |
| 1181 | * @param Node $node |
| 1182 | */ |
| 1183 | public function visitUnaryOp( Node $node ): void { |
| 1184 | // ~ and @ are the only two unary ops |
| 1185 | // that can preserve taint (others cast bool or int) |
| 1186 | $unsafe = [ |
| 1187 | \ast\flags\UNARY_BITWISE_NOT, |
| 1188 | \ast\flags\UNARY_SILENCE |
| 1189 | ]; |
| 1190 | if ( in_array( $node->flags, $unsafe, true ) ) { |
| 1191 | $this->curTaintWithError = clone $this->getTaintedness( $node->children['expr'] ); |
| 1192 | } else { |
| 1193 | $this->setCurTaintSafe(); |
| 1194 | } |
| 1195 | $this->setCachedData( $node ); |
| 1196 | } |
| 1197 | |
| 1198 | /** |
| 1199 | * @param Node $node |
| 1200 | */ |
| 1201 | public function visitPostInc( Node $node ): void { |
| 1202 | $this->analyzeIncOrDec( $node ); |
| 1203 | } |
| 1204 | |
| 1205 | /** |
| 1206 | * @param Node $node |
| 1207 | */ |
| 1208 | public function visitPreInc( Node $node ): void { |
| 1209 | $this->analyzeIncOrDec( $node ); |
| 1210 | } |
| 1211 | |
| 1212 | /** |
| 1213 | * @param Node $node |
| 1214 | */ |
| 1215 | public function visitPostDec( Node $node ): void { |
| 1216 | $this->analyzeIncOrDec( $node ); |
| 1217 | } |
| 1218 | |
| 1219 | /** |
| 1220 | * @param Node $node |
| 1221 | */ |
| 1222 | public function visitPreDec( Node $node ): void { |
| 1223 | $this->analyzeIncOrDec( $node ); |
| 1224 | } |
| 1225 | |
| 1226 | /** |
| 1227 | * Handles all post/pre-increment/decrement operators. They have no effect on the |
| 1228 | * taintedness of a variable. |
| 1229 | * |
| 1230 | * @param Node $node |
| 1231 | */ |
| 1232 | private function analyzeIncOrDec( Node $node ): void { |
| 1233 | $this->curTaintWithError = clone $this->getTaintedness( $node->children['var'] ); |
| 1234 | $this->setCachedData( $node ); |
| 1235 | } |
| 1236 | |
| 1237 | /** |
| 1238 | * @param Node $node |
| 1239 | */ |
| 1240 | public function visitCast( Node $node ): void { |
| 1241 | // Casting between an array and object maintains |
| 1242 | // taint. Casting an object to a string calls __toString(). |
| 1243 | // Future TODO: handle the string case properly. |
| 1244 | $dangerousCasts = [ |
| 1245 | \ast\flags\TYPE_STRING, |
| 1246 | \ast\flags\TYPE_ARRAY, |
| 1247 | \ast\flags\TYPE_OBJECT |
| 1248 | ]; |
| 1249 | |
| 1250 | if ( !in_array( $node->flags, $dangerousCasts, true ) ) { |
| 1251 | $this->setCurTaintSafe(); |
| 1252 | } else { |
| 1253 | $exprTaint = $this->getTaintedness( $node->children['expr'] ); |
| 1254 | // Note, casting deletes shapes. |
| 1255 | $this->curTaintWithError = new TaintednessWithError( |
| 1256 | $exprTaint->getTaintedness()->asCollapsed(), |
| 1257 | $exprTaint->getError(), |
| 1258 | $exprTaint->getMethodLinks()->asCollapsed() |
| 1259 | ); |
| 1260 | } |
| 1261 | $this->setCachedData( $node ); |
| 1262 | } |
| 1263 | |
| 1264 | /** |
| 1265 | * The taint is the taint of all the child elements |
| 1266 | * |
| 1267 | * @param Node $node |
| 1268 | */ |
| 1269 | public function visitEncapsList( Node $node ): void { |
| 1270 | $this->curTaintWithError = TaintednessWithError::newEmpty(); |
| 1271 | foreach ( $node->children as $child ) { |
| 1272 | $this->curTaintWithError->mergeWith( $this->getTaintedness( $child ) ); |
| 1273 | } |
| 1274 | $this->setCachedData( $node ); |
| 1275 | } |
| 1276 | |
| 1277 | /** |
| 1278 | * Visit a node that is always safe |
| 1279 | * |
| 1280 | * @param Node $node |
| 1281 | */ |
| 1282 | public function visitIsset( Node $node ): void { |
| 1283 | $this->setCurTaintSafe(); |
| 1284 | } |
| 1285 | |
| 1286 | /** |
| 1287 | * Visits calls to empty(), which is always safe |
| 1288 | * |
| 1289 | * @param Node $node |
| 1290 | */ |
| 1291 | public function visitEmpty( Node $node ): void { |
| 1292 | $this->setCurTaintSafe(); |
| 1293 | } |
| 1294 | |
| 1295 | /** |
| 1296 | * Visit a node that is always safe |
| 1297 | * |
| 1298 | * @param Node $node |
| 1299 | */ |
| 1300 | public function visitMagicConst( Node $node ): void { |
| 1301 | $this->setCurTaintSafe(); |
| 1302 | } |
| 1303 | |
| 1304 | /** |
| 1305 | * Visit a node that is always safe |
| 1306 | * |
| 1307 | * @param Node $node |
| 1308 | */ |
| 1309 | public function visitInstanceOf( Node $node ): void { |
| 1310 | $this->setCurTaintSafe(); |
| 1311 | } |
| 1312 | |
| 1313 | /** |
| 1314 | * @param Node $node |
| 1315 | */ |
| 1316 | public function visitMatch( Node $node ): void { |
| 1317 | $this->curTaintWithError = TaintednessWithError::newEmpty(); |
| 1318 | // Based on UnionTypeVisitor |
| 1319 | foreach ( $node->children['stmts']->children as $armNode ) { |
| 1320 | // It sounds a bit weird to have to call this ourselves, but aight. |
| 1321 | if ( !BlockExitStatusChecker::willUnconditionallyThrowOrReturn( $armNode ) ) { |
| 1322 | // Note, we're straight using the expr to avoid implementing visitMatchArm |
| 1323 | $this->curTaintWithError->mergeWith( $this->getTaintedness( $armNode->children['expr'] ) ); |
| 1324 | } |
| 1325 | } |
| 1326 | |
| 1327 | $this->setCachedData( $node ); |
| 1328 | } |
| 1329 | } |