docs/ExceptionHandling.html

   1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01//EN"
   2                       "http://www.w3.org/TR/html4/strict.dtd">
   3 <html>
   4 <head>
   5   <title>Exception Handling in LLVM</title>
   6   <link rel="stylesheet" href="llvm.css" type="text/css">
   7 </head>
   8 <body>
   9
  10 <div class="doc_title">Exception Handling in LLVM</div>
  11
  12 <table class="layout" style="width:100%">
  13   <tr class="layout">
  14     <td class="left">
  15 <ul>
  16   <li><a href="#introduction">Introduction</a>
  17   <ol>
  18     <li><a href="#itanium">Itanium ABI Zero-cost Exception Handling</a></li>
  19     <li><a href="#overview">Overview</a></li>
  20   </ol></li>
  21   <li><a href="#codegen">LLVM Code Generation</a>
  22   <ol>
  23     <li><a href="#throw">Throw</a></li>
  24     <li><a href="#try_catch">Try/Catch</a></li>
  25     <li><a href="#cleanups">Cleanups</a></li>
  26     <li><a href="#throw_filters">Throw Filters</a></li>
  27     <li><a href="#restrictions">Restrictions</a></li>
  28   </ol></li>
  29   <li><a href="#format_common_intrinsics">Exception Handling Intrinsics</a>
  30   <ol>
  31         <li><a href="#llvm_eh_exception"><tt>llvm.eh.exception</tt></a></li>
  32         <li><a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a></li>
  33         <li><a href="#llvm_eh_typeid_for"><tt>llvm.eh.typeid.for</tt></a></li>
  34         <li><a href="#llvm_eh_sjlj_setjmp"><tt>llvm.eh.sjlj.setjmp</tt></a></li>
  35         <li><a href="#llvm_eh_sjlj_longjmp"><tt>llvm.eh.sjlj.longjmp</tt></a></li>
  36   </ol></li>
  37   <li><a href="#asm">Asm Table Formats</a>
  38   <ol>
  39     <li><a href="#unwind_tables">Exception Handling Frame</a></li>
  40     <li><a href="#exception_tables">Exception Tables</a></li>
  41   </ol></li>
  42   <li><a href="#todo">ToDo</a></li>
  43 </ul>
  44 </td>
  45 </tr></table>
  46
  47 <div class="doc_author">
  48   <p>Written by <a href="mailto:jlaskey@mac.com">Jim Laskey</a></p>
  49 </div>
  50
  51
  52 <!-- *********************************************************************** -->
  53 <div class="doc_section"><a name="introduction">Introduction</a></div>
  54 <!-- *********************************************************************** -->
  55
  56 <div class="doc_text">
  57
  58 <p>This document is the central repository for all information pertaining to
  59 exception handling in LLVM.  It describes the format that LLVM exception
  60 handling information takes, which is useful for those interested in creating
  61 front-ends or dealing directly with the information.  Further, this document
  62 provides specific examples of what exception handling information is used for
  63 C/C++.</p>
  64
  65 </div>
  66
  67 <!-- ======================================================================= -->
  68 <div class="doc_subsection">
  69   <a name="itanium">Itanium ABI Zero-cost Exception Handling</a>
  70 </div>
  71
  72 <div class="doc_text">
  73
  74 <p>Exception handling for most programming languages is designed to recover from
  75 conditions that rarely occur during general use of an application.  To that end,
  76 exception handling should not interfere with the main flow of an
  77 application's algorithm by performing checkpointing tasks such as saving
  78 the current pc or register state.</p>
  79
  80 <p>The Itanium ABI Exception Handling Specification defines a methodology for
  81 providing outlying data in the form of exception tables without inlining
  82 speculative exception handling code in the flow of an application's main
  83 algorithm.  Thus, the specification is said to add "zero-cost" to the normal
  84 execution of an application.</p>
  85
  86 <p>A more complete description of the Itanium ABI exception handling runtime
  87 support of can be found at <a
  88 href="http://www.codesourcery.com/cxx-abi/abi-eh.html">Itanium C++ ABI:
  89 Exception Handling.</a> A description of the exception frame format can be found
  90 at <a href="http://refspecs.freestandards.org/LSB_3.0.0/LSB-Core-generic/LSB-
  91 Core-generic/ehframechpt.html">Exception Frames</a>, with details of the Dwarf
  92 specification at <a href="http://www.eagercon.com/dwarf/dwarf3std.htm">Dwarf 3
  93 Standard.</a> A description for the C++ exception table formats can be found at
  94 <a href="http://www.codesourcery.com/cxx-abi/exceptions.pdf">Exception Handling
  95 Tables.</a></p>
  96
  97 </div>
  98
  99 <!-- ======================================================================= -->
 100 <div class="doc_subsection">
 101   <a name="overview">Overview</a>
 102 </div>
 103
 104 <div class="doc_text">
 105
 106 <p>When an exception is thrown in llvm code, the runtime does a best effort to
 107 find a handler suited to process the circumstance.</p>
 108
 109 <p>The runtime first attempts to find an <i>exception frame</i> corresponding to
 110 the function where the exception was thrown.  If the programming language (ex.
 111 C++) supports exception handling, the exception frame contains a reference to an
 112 exception table describing how to process the exception.  If the language (ex.
 113 C) does not support exception handling or if the exception needs to be forwarded
 114 to a prior activation, the exception frame contains information about how to
 115 unwind the current activation and restore the state of the prior activation.
 116 This process is repeated until the exception is handled.  If the exception is
 117 not handled and no activations remain, then the application is terminated with
 118 an appropriate error message.</p>
 119
 120 <p>Since different programming languages have different behaviors when handling
 121 exceptions, the exception handling ABI provides a mechanism for supplying
 122 <i>personalities.</i> An exception handling personality is defined by way of a
 123 <i>personality function</i> (ex. for C++ <tt>__gxx_personality_v0</tt>) which
 124 receives the context of the exception, an <i>exception structure</i> containing
 125 the exception object type and value, and a reference to the exception table for
 126 the current function.  The personality function for the current compile unit is
 127 specified in a <i>common exception frame</i>.</p>
 128
 129 <p>The organization of an exception table is language dependent.  For C++, an
 130 exception table is organized as a series of code ranges defining what to do if
 131 an exception occurs in that range.  Typically, the information associated with a
 132 range defines which types of exception objects (using C++ <i>type info</i>) that
 133 are handled in that range, and an associated action that should take place.
 134 Actions typically pass control to a <i>landing pad</i>.</p>
 135
 136 <p>A landing pad corresponds to the code found in the catch portion of a
 137 try/catch sequence.  When execution resumes at a landing pad, it receives the
 138 exception structure and a selector corresponding to the <i>type</i> of exception
 139 thrown.  The selector is then used to determine which catch should actually
 140 process the exception.</p>
 141
 142 </div>
 143
 144 <!-- ======================================================================= -->
 145 <div class="doc_section">
 146   <a name="codegen">LLVM Code Generation</a>
 147 </div>
 148
 149 <div class="doc_text">
 150
 151 <p>At the time of this writing, only C++ exception handling support is available
 152 in LLVM.  So the remainder of this document will be somewhat C++-centric.</p>
 153
 154 <p>From the C++ developers perspective, exceptions are defined in terms of the
 155 <tt>throw</tt> and <tt>try/catch</tt> statements.  In this section we will
 156 describe the implementation of llvm exception handling in terms of C++
 157 examples.</p>
 158
 159 </div>
 160
 161 <!-- ======================================================================= -->
 162 <div class="doc_subsection">
 163   <a name="throw">Throw</a>
 164 </div>
 165
 166 <div class="doc_text">
 167
 168 <p>Languages that support exception handling typically provide a <tt>throw</tt>
 169 operation to initiate the exception process.  Internally, a throw operation
 170 breaks down into two steps.  First, a request is made to allocate exception
 171 space for an exception structure.  This structure needs to survive beyond the
 172 current activation.  This structure will contain the type and value of the
 173 object being thrown.  Second, a call is made to the runtime to raise the
 174 exception, passing the exception structure as an argument.</p>
 175
 176 <p>In C++, the allocation of the exception structure is done by the
 177 <tt>__cxa_allocate_exception</tt> runtime function.  The exception raising is
 178 handled by <tt>__cxa_throw</tt>.  The type of the exception is represented using
 179 a C++ RTTI type info structure.</p>
 180
 181 </div>
 182
 183 <!-- ======================================================================= -->
 184 <div class="doc_subsection">
 185   <a name="try_catch">Try/Catch</a>
 186 </div>
 187
 188 <div class="doc_text">
 189
 190 <p>A call within the scope of a try statement can potentially raise an exception.
 191 In those circumstances, the LLVM C++ front-end replaces the call with an
 192 <tt>invoke</tt> instruction.  Unlike a call, the invoke has two potential
 193 continuation points; where to continue when the call succeeds as per normal, and
 194 where to continue if the call raises an exception, either by a throw or the
 195 unwinding of a throw.</p>
 196
 197 <p>The term used to define a the place where an invoke continues after an
 198 exception is called a <i>landing pad</i>.  LLVM landing pads are conceptually
 199 alternative function entry points where a exception structure reference and a type
 200 info index are passed in as arguments.  The landing pad saves the exception
 201 structure reference and then proceeds to select the catch block that corresponds
 202 to the type info of the exception object.</p>
 203
 204 <p>Two llvm intrinsic functions are used convey information about the landing
 205 pad to the back end.</p>
 206
 207 <p><a href="#llvm_eh_exception"><tt>llvm.eh.exception</tt></a> takes no
 208 arguments and returns a pointer to the exception structure.  This only returns a
 209 sensible value if called after an invoke has branched to a landing pad.  Due to
 210 codegen limitations, it must currently be called in the landing pad itself.</p>
 211
 212 <p><a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> takes a minimum of
 213 three arguments.  The first argument is the reference to the exception
 214 structure. The second argument is a reference to the personality function to be
 215 used for this try catch sequence. Each of the remaining arguments is either a
 216 reference to the type info for a catch statement,
 217 a <a href="#throw_filters">filter</a> expression,
 218 or the number zero representing a <a href="#cleanups">cleanup</a>.
 219 The exception is tested against the arguments sequentially from first to last.
 220 The result of the <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> is a
 221 positive number if the exception matched a type info, a negative number if it matched
 222 a filter, and zero if it matched a cleanup.  If nothing is matched, the behaviour of
 223 the program is <a href="#restrictions">undefined</a>.
 224 This only returns a sensible value if called after an invoke has branched to a
 225 landing pad.  Due to codegen limitations, it must currently be called in the
 226 landing pad itself.
 227 If a type info matched then the selector value is the index of the type info in
 228 the exception table, which can be obtained using the
 229 <a href="#llvm_eh_typeid_for"><tt>llvm.eh.typeid.for</tt></a> intrinsic.</p>
 230
 231 <p>Once the landing pad has the type info selector, the code branches to the
 232 code for the first catch.  The catch then checks the value of the type info
 233 selector against the index of type info for that catch.  Since the type info
 234 index is not known until all the type info have been gathered in the backend,
 235 the catch code will call the <a
 236 href="#llvm_eh_typeid_for"><tt>llvm.eh.typeid.for</tt></a> intrinsic to
 237 determine the index for a given type info.  If the catch fails to match the
 238 selector then control is passed on to the next catch. Note: Since the landing
 239 pad will not be used if there is no match in the list of type info on the call
 240 to <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a>, then neither the
 241 last catch nor <i>catch all</i> need to perform the the check against the
 242 selector.</p>
 243
 244 <p>Finally, the entry and exit of catch code is bracketed with calls to
 245 <tt>__cxa_begin_catch</tt> and <tt>__cxa_end_catch</tt>.
 246 <tt>__cxa_begin_catch</tt> takes a exception structure reference as an argument
 247 and returns the value of the exception object. <tt>__cxa_end_catch</tt>
 248 takes a exception structure reference as an argument. This function clears the
 249 exception from the exception space.  Note: a rethrow from within the catch may
 250 replace this call with a <tt>__cxa_rethrow</tt>.</p>
 251
 252 </div>
 253
 254 <!-- ======================================================================= -->
 255 <div class="doc_subsection">
 256   <a name="cleanups">Cleanups</a>
 257 </div>
 258
 259 <div class="doc_text">
 260
 261 <p>To handle destructors and cleanups in try code, control may not run directly
 262 from a landing pad to the first catch.  Control may actually flow from the
 263 landing pad to clean up code and then to the first catch.  Since the required
 264 clean up for each invoke in a try may be different (ex., intervening
 265 constructor), there may be several landing pads for a given try.  If cleanups
 266 need to be run, the number zero should be passed as the last
 267 <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> argument.
 268 However for C++ a <tt>null i8*</tt> <a href="#restrictions">must</a> be passed
 269 instead.
 270 </p>
 271
 272 </div>
 273
 274 <!-- ======================================================================= -->
 275 <div class="doc_subsection">
 276   <a name="throw_filters">Throw Filters</a>
 277 </div>
 278
 279 <div class="doc_text">
 280
 281 <p>C++ allows the specification of which exception types can be thrown from
 282 a function.  To represent this a top level landing pad may exist to filter out
 283 invalid types.  To express this in LLVM code the landing pad will call <a
 284 href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a>.  The arguments are a
 285 reference to the exception structure, a reference to the personality function,
 286 the length of the filter expression (the number of type infos plus one),
 287 followed by the type infos themselves.
 288 <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> will return a negative
 289 value if the exception does not match any of the type infos.  If no match is
 290 found then a call to <tt>__cxa_call_unexpected</tt> should be made, otherwise
 291 <tt>_Unwind_Resume</tt>.  Each of these functions requires a reference to the
 292 exception structure.  Note that the most general form of an
 293 <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> call can contain
 294 any number of type infos, filter expressions and cleanups (though having more
 295 than one cleanup is pointless).  The LLVM C++ front-end can generate such
 296 <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> calls due to inlining
 297 creating nested exception handling scopes.</p>
 298
 299 </div>
 300
 301 <!-- ======================================================================= -->
 302 <div class="doc_subsection">
 303   <a name="restrictions">Restrictions</a>
 304 </div>
 305
 306 <div class="doc_text">
 307
 308 <p>The semantics of the invoke instruction require that any exception that
 309 unwinds through an invoke call should result in a branch to the invoke's unwind
 310 label.  However such a branch will only happen if the
 311 <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> matches.
 312 Thus in order to ensure correct operation, the front-end must only generate
 313 <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> calls that are
 314 guaranteed to always match whatever exception unwinds through the invoke.
 315 For most languages it is enough to pass zero, indicating the presence of
 316 a <a href="#cleanups">cleanup</a>, as the last
 317 <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> argument.
 318 However for C++ this is not sufficient, because the C++ personality function
 319 will terminate the program if it detects that unwinding the exception only
 320 results in matches with cleanups.  For C++ a <tt>null i8*</tt> should
 321 be passed as the last
 322 <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> argument instead.
 323 This is interpreted as a catch-all by the C++ personality function, and will
 324 always match.
 325 </p>
 326
 327 </div>
 328
 329 <!-- ======================================================================= -->
 330 <div class="doc_section">
 331   <a name="format_common_intrinsics">Exception Handling Intrinsics</a>
 332 </div>
 333
 334 <div class="doc_text">
 335
 336 <p>LLVM uses several intrinsic functions (name prefixed with "llvm.eh") to
 337 provide exception handling information at various points in generated code.</p>
 338
 339 </div>
 340
 341 <!-- ======================================================================= -->
 342 <div class="doc_subsubsection">
 343   <a name="llvm_eh_exception">llvm.eh.exception</a>
 344 </div>
 345
 346 <div class="doc_text">
 347 <pre>
 348   i8* %<a href="#llvm_eh_exception">llvm.eh.exception</a>( )
 349 </pre>
 350
 351 <p>This intrinsic returns a pointer to the exception structure.</p>
 352
 353 </div>
 354
 355 <!-- ======================================================================= -->
 356 <div class="doc_subsubsection">
 357   <a name="llvm_eh_selector">llvm.eh.selector</a>
 358 </div>
 359
 360 <div class="doc_text">
 361 <pre>
 362   i32 %<a href="#llvm_eh_selector">llvm.eh.selector.i32</a>(i8*, i8*, i8*, ...)
 363   i64 %<a href="#llvm_eh_selector">llvm.eh.selector.i64</a>(i8*, i8*, i8*, ...)
 364 </pre>
 365
 366 <p>This intrinsic is used to compare the exception with the given type infos,
 367 filters and cleanups.</p>
 368
 369 <p><a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> takes a minimum of
 370 three arguments.  The first argument is the reference to the exception
 371 structure. The second argument is a reference to the personality function to be
 372 used for this try catch sequence. Each of the remaining arguments is either a
 373 reference to the type info for a catch statement,
 374 a <a href="#throw_filters">filter</a> expression,
 375 or the number zero representing a <a href="#cleanups">cleanup</a>.
 376 The exception is tested against the arguments sequentially from first to last.
 377 The result of the <a href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a> is a
 378 positive number if the exception matched a type info, a negative number if it matched
 379 a filter, and zero if it matched a cleanup.  If nothing is matched, the behaviour of
 380 the program is <a href="#restrictions">undefined</a>.
 381 If a type info matched then the selector value is the index of the type info in
 382 the exception table, which can be obtained using the
 383 <a href="#llvm_eh_typeid_for"><tt>llvm.eh.typeid.for</tt></a> intrinsic.</p>
 384
 385 </div>
 386
 387 <!-- ======================================================================= -->
 388 <div class="doc_subsubsection">
 389   <a name="llvm_eh_typeid_for">llvm.eh.typeid.for</a>
 390 </div>
 391
 392 <div class="doc_text">
 393 <pre>
 394   i32 %<a href="#llvm_eh_typeid_for">llvm.eh.typeid.for.i32</a>(i8*)
 395   i64 %<a href="#llvm_eh_typeid_for">llvm.eh.typeid.for.i64</a>(i8*)
 396 </pre>
 397
 398 <p>This intrinsic returns the type info index in the exception table of the
 399 current function.  This value can be used to compare against the result of <a
 400 href="#llvm_eh_selector"><tt>llvm.eh.selector</tt></a>.  The single argument is
 401 a reference to a type info.</p>
 402
 403 </div>
 404
 405 <!-- ======================================================================= -->
 406 <div class="doc_subsubsection">
 407   <a name="llvm_eh_sjlj_setjmp">llvm.eh.sjlj.setjmp</a>
 408 </div>
 409
 410 <div class="doc_text">
 411 <pre>
 412   i32 %<a href="#llvm_eh_sjlj_setjmp">llvm.eh.sjlj.setjmp</a>(i8*)
 413 </pre>
 414
 415 <p>The SJLJ exception handling uses this intrinsic to force register saving
 416 for the current function and to store the address of the following instruction
 417 for use as a destination address by <a href="#llvm_eh_sjlj_setjmp">
 418 <tt>llvm.eh.sjlj.longjmp</tt></a>. The buffer format and the overall functioning
 419 of this intrinsic is compatible with the GCC <tt>__builtin_setjmp</tt>
 420 implementation, allowing code built with the two compilers to interoperate.</p>
 421
 422 <p>The single parameter is a pointer to a five word buffer in which the
 423 calling context is saved. The front end places the frame pointer in the
 424 first word, and the target implementation of this intrinsic should place the
 425 destination address for a <a href="#llvm_eh_sjlj_longjmp"><tt>
 426 llvm.eh.sjlj.longjmp</tt></a> in the second word. The following three words
 427 are available for use in a target-specific manner.</p>
 428
 429 <!-- ======================================================================= -->
 430 <div class="doc_section">
 431   <a name="asm">Asm Table Formats</a>
 432 </div>
 433
 434 <div class="doc_text">
 435
 436 <p>There are two tables that are used by the exception handling runtime to
 437 determine which actions should take place when an exception is thrown.</p>
 438
 439 </div>
 440
 441 <!-- ======================================================================= -->
 442 <div class="doc_subsection">
 443   <a name="unwind_tables">Exception Handling Frame</a>
 444 </div>
 445
 446 <div class="doc_text">
 447
 448 <p>An exception handling frame <tt>eh_frame</tt> is very similar to the unwind
 449 frame used by dwarf debug info.  The frame contains all the information
 450 necessary to tear down the current frame and restore the state of the prior
 451 frame.  There is an exception handling frame for each function in a compile
 452 unit, plus a common exception handling frame that defines information common to
 453 all functions in the unit.</p>
 454
 455 <p>Todo - Table details here.</p>
 456
 457 </div>
 458
 459 <!-- ======================================================================= -->
 460 <div class="doc_subsection">
 461   <a name="exception_tables">Exception Tables</a>
 462 </div>
 463
 464 <div class="doc_text">
 465
 466 <p>An exception table contains information about what actions to take when an
 467 exception is thrown in a particular part of a function's code.  There is
 468 one exception table per function except leaf routines and functions that have
 469 only calls to non-throwing functions will not need an exception table.</p>
 470
 471 <p>Todo - Table details here.</p>
 472
 473 </div>
 474
 475 <!-- ======================================================================= -->
 476 <div class="doc_section">
 477   <a name="todo">ToDo</a>
 478 </div>
 479
 480 <div class="doc_text">
 481
 482 <ol>
 483
 484 <li><p>Testing/Testing/Testing.</p></li>
 485
 486 </ol>
 487
 488 </div>
 489
 490 <!-- *********************************************************************** -->
 491
 492 <hr>
 493 <address>
 494   <a href="http://jigsaw.w3.org/css-validator/check/referer"><img
 495   src="http://jigsaw.w3.org/css-validator/images/vcss-blue" alt="Valid CSS"></a>
 496   <a href="http://validator.w3.org/check/referer"><img
 497   src="http://www.w3.org/Icons/valid-html401-blue" alt="Valid HTML 4.01"></a>
 498
 499   <a href="mailto:sabre@nondot.org">Chris Lattner</a><br>
 500   <a href="http://llvm.org">LLVM Compiler Infrastructure</a><br>
 501   Last modified: $Date$
 502 </address>
 503
 504 </body>
 505 </html>