1 <html> 2 <head> 3 <meta http-equiv="Content-Type" content="text/html; charset=ISO-8859-1"> 4 <title>11.Ptrcheck: an experimental heap, stack and global array overrun detector</title> 5 <link rel="stylesheet" href="vg_basic.css" type="text/css"> 6 <meta name="generator" content="DocBook XSL Stylesheets V1.75.2"> 7 <link rel="home" href="index.html" title="Valgrind Documentation"> 8 <link rel="up" href="manual.html" title="Valgrind User Manual"> 9 <link rel="prev" href="dh-manual.html" title="10.DHAT: a dynamic heap analysis tool"> 10 <link rel="next" href="bbv-manual.html" title="12.BBV: an experimental basic block vector generation tool"> 11 </head> 12 <body bgcolor="white" text="black" link="#0000FF" vlink="#840084" alink="#0000FF"> 13 <div><table class="nav" width="100%" cellspacing="3" cellpadding="3" border="0" summary="Navigation header"><tr> 14 <td width="22px" align="center" valign="middle"><a accesskey="p" href="dh-manual.html"><img src="images/prev.png" width="18" height="21" border="0" alt="Prev"></a></td> 15 <td width="25px" align="center" valign="middle"><a accesskey="u" href="manual.html"><img src="images/up.png" width="21" height="18" border="0" alt="Up"></a></td> 16 <td width="31px" align="center" valign="middle"><a accesskey="h" href="index.html"><img src="images/home.png" width="27" height="20" border="0" alt="Up"></a></td> 17 <th align="center" valign="middle">Valgrind User Manual</th> 18 <td width="22px" align="center" valign="middle"><a accesskey="n" href="bbv-manual.html"><img src="images/next.png" width="18" height="21" border="0" alt="Next"></a></td> 19 </tr></table></div> 20 <div class="chapter" title="11.Ptrcheck: an experimental heap, stack and global array overrun detector"> 21 <div class="titlepage"><div><div><h2 class="title"> 22 <a name="pc-manual"></a>11.Ptrcheck: an experimental heap, stack and global array overrun detector</h2></div></div></div> 23 <div class="toc"> 24 <p><b>Table of Contents</b></p> 25 <dl> 26 <dt><span class="sect1"><a href="pc-manual.html#pc-manual.overview">11.1. Overview</a></span></dt> 27 <dt><span class="sect1"><a href="pc-manual.html#pc-manual.options">11.2. Ptrcheck Command-line Options</a></span></dt> 28 <dt><span class="sect1"><a href="pc-manual.html#pc-manual.how-works.heap-checks">11.3. How Ptrcheck Works: Heap Checks</a></span></dt> 29 <dt><span class="sect1"><a href="pc-manual.html#pc-manual.how-works.sg-checks">11.4. How Ptrcheck Works: Stack and Global Checks</a></span></dt> 30 <dt><span class="sect1"><a href="pc-manual.html#pc-manual.cmp-w-memcheck">11.5. Comparison with Memcheck</a></span></dt> 31 <dt><span class="sect1"><a href="pc-manual.html#pc-manual.limitations">11.6. Limitations</a></span></dt> 32 <dt><span class="sect1"><a href="pc-manual.html#pc-manual.todo-user-visible">11.7. Still To Do: User-visible Functionality</a></span></dt> 33 <dt><span class="sect1"><a href="pc-manual.html#pc-manual.todo-implementation">11.8. Still To Do: Implementation Tidying</a></span></dt> 34 </dl> 35 </div> 36 <p>To use this tool, you must specify 37 <code class="option">--tool=exp-ptrcheck</code> on the Valgrind 38 command line.</p> 39 <div class="sect1" title="11.1.Overview"> 40 <div class="titlepage"><div><div><h2 class="title" style="clear: both"> 41 <a name="pc-manual.overview"></a>11.1.Overview</h2></div></div></div> 42 <p>Ptrcheck is a tool for finding overruns of heap, stack 43 and global arrays. Its functionality overlaps somewhat with 44 Memcheck's, but it is able to catch invalid accesses in a number of 45 cases that Memcheck would miss. A detailed comparison against 46 Memcheck is presented below.</p> 47 <p>Ptrcheck is composed of two almost completely independent tools 48 that have been glued together. One part, 49 in <code class="computeroutput">h_main.[ch]</code>, checks accesses 50 through heap-derived pointers. The other part, in 51 <code class="computeroutput">sg_main.[ch]</code>, checks accesses to 52 stack and global arrays. The remaining 53 files <code class="computeroutput">pc_{common,main}.[ch]</code>, provide 54 common error-management and coordination functions, so as to make it 55 appear as a single tool.</p> 56 <p>The heap-check part is an extensively-hacked (largely rewritten) 57 version of the experimental "Annelid" tool developed and described by 58 Nicholas Nethercote and Jeremy Fitzhardinge. The stack- and global- 59 check part uses a heuristic approach derived from an observation about 60 the likely forms of stack and global array accesses, and, as far as is 61 known, is entirely novel.</p> 62 </div> 63 <div class="sect1" title="11.2.Ptrcheck Command-line Options"> 64 <div class="titlepage"><div><div><h2 class="title" style="clear: both"> 65 <a name="pc-manual.options"></a>11.2.Ptrcheck Command-line Options</h2></div></div></div> 66 <p>Ptrcheck-specific command-line options are:</p> 67 <div class="variablelist"> 68 <a name="pc.opts.list"></a><dl> 69 <dt> 70 <a name="opt.enable-sg-checks"></a><span class="term"> 71 <code class="option">--enable-sg-checks=no|yes 72 [default: yes] </code> 73 </span> 74 </dt> 75 <dd><p>By default, Ptrcheck checks for overruns of stack, global 76 and heap arrays. 77 With <code class="varname">--enable-sg-checks=no</code>, the stack and 78 global array checks are omitted, and only heap checking is 79 performed. This can be useful because the stack and global 80 checks are quite expensive, so omitting them speeds Ptrcheck up 81 a lot. 82 </p></dd> 83 <dt> 84 <a name="opt.partial-loads-ok"></a><span class="term"> 85 <code class="option">--partial-loads-ok=<yes|no> [default: no] </code> 86 </span> 87 </dt> 88 <dd> 89 <p>This option has the same meaning as it does for 90 Memcheck.</p> 91 <p>Controls how Ptrcheck handles word-sized, word-aligned 92 loads which partially overlap the end of heap blocks -- that is, 93 some of the bytes in the word are validly addressable, but 94 others are not. When <code class="varname">yes</code>, such loads do not 95 produce an address error. When <code class="varname">no</code> (the 96 default), loads from partially invalid addresses are treated the 97 same as loads from completely invalid addresses: an illegal heap 98 access error is issued. 99 </p> 100 <p>Note that code that behaves in this way is in violation of 101 the the ISO C/C++ standards, and should be considered broken. If 102 at all possible, such code should be fixed. This option should be 103 used only as a last resort.</p> 104 </dd> 105 </dl> 106 </div> 107 </div> 108 <div class="sect1" title="11.3.How Ptrcheck Works: Heap Checks"> 109 <div class="titlepage"><div><div><h2 class="title" style="clear: both"> 110 <a name="pc-manual.how-works.heap-checks"></a>11.3.How Ptrcheck Works: Heap Checks</h2></div></div></div> 111 <p>Ptrcheck can check for invalid uses of heap pointers, including 112 out of range accesses and accesses to freed memory. The mechanism is 113 however completely different from Memcheck's, and the checking is more 114 powerful.</p> 115 <p>For each pointer in the program, Ptrcheck keeps track of which 116 heap block (if any) it was derived from. Then, when an access is made 117 through that pointer, Ptrcheck compares the access address with the 118 bounds of the associated block, and reports an error if the address is 119 out of bounds, or if the block has been freed.</p> 120 <p>Of course it is rarely the case that one wants to access a block 121 only at the exact address returned by <code class="function">malloc</code> et al. 122 Ptrcheck understands that adding or subtracting offsets from a pointer to a 123 block results in a pointer to the same block.</p> 124 <p>At a fundamental level, this scheme works because a correct 125 program cannot make assumptions about the addresses returned by 126 <code class="function">malloc</code> et al. In particular it cannot make any 127 assumptions about the differences in addresses returned by subsequent calls 128 to <code class="function">malloc</code> et al. Hence there are very few ways to take 129 an address returned by <code class="function">malloc</code>, modify it, and still 130 have a valid address. In short, the only allowable operations are adding 131 and subtracting other non-pointer values. Almost all other operations 132 produce a value which cannot possibly be a valid pointer.</p> 133 </div> 134 <div class="sect1" title="11.4.How Ptrcheck Works: Stack and Global Checks"> 135 <div class="titlepage"><div><div><h2 class="title" style="clear: both"> 136 <a name="pc-manual.how-works.sg-checks"></a>11.4.How Ptrcheck Works: Stack and Global Checks</h2></div></div></div> 137 <p>When a source file is compiled 138 with <code class="option">-g</code>, the compiler attaches DWARF3 139 debugging information which describes the location of all stack and 140 global arrays in the file.</p> 141 <p>Checking of accesses to such arrays would then be relatively 142 simple, if the compiler could also tell us which array (if any) each 143 memory referencing instruction was supposed to access. Unfortunately 144 the DWARF3 debugging format does not provide a way to represent such 145 information, so we have to resort to a heuristic technique to 146 approximate the same information. The key observation is that 147 <span class="emphasis"><em> 148 if a memory referencing instruction accesses inside a stack or 149 global array once, then it is highly likely to always access that 150 same array</em></span>.</p> 151 <p>To see how this might be useful, consider the following buggy 152 fragment:</p> 153 <pre class="programlisting"> 154 { int i, a[10]; // both are auto vars 155 for (i = 0; i <= 10; i++) 156 a[i] = 42; 157 } 158 </pre> 159 <p>At run time we will know the precise address 160 of <code class="computeroutput">a[]</code> on the stack, and so we can 161 observe that the first store resulting from <code class="computeroutput">a[i] = 162 42</code> writes <code class="computeroutput">a[]</code>, and 163 we will (correctly) assume that that instruction is intended always to 164 access <code class="computeroutput">a[]</code>. Then, on the 11th 165 iteration, it accesses somewhere else, possibly a different local, 166 possibly an un-accounted for area of the stack (eg, spill slot), so 167 Ptrcheck reports an error.</p> 168 <p>There is an important caveat.</p> 169 <p>Imagine a function such as <code class="function">memcpy</code>, which is used 170 to read and write many different areas of memory over the lifetime of the 171 program. If we insist that the read and write instructions in its memory 172 copying loop only ever access one particular stack or global variable, we 173 will be flooded with errors resulting from calls to 174 <code class="function">memcpy</code>.</p> 175 <p>To avoid this problem, Ptrcheck instantiates fresh likely-target 176 records for each entry to a function, and discards them on exit. This 177 allows detection of cases where (e.g.) <code class="function">memcpy</code> overflows 178 its source or destination buffers for any specific call, but does not carry 179 any restriction from one call to the next. Indeed, multiple threads may be 180 multiple simultaneous calls to (e.g.) <code class="function">memcpy</code> without 181 mutual interference.</p> 182 </div> 183 <div class="sect1" title="11.5.Comparison with Memcheck"> 184 <div class="titlepage"><div><div><h2 class="title" style="clear: both"> 185 <a name="pc-manual.cmp-w-memcheck"></a>11.5.Comparison with Memcheck</h2></div></div></div> 186 <p>Memcheck does not do any access checks for stack or global arrays, so 187 the presence of those in Ptrcheck is a straight win. (But see 188 "Limitations" below).</p> 189 <p>Memcheck and Ptrcheck use different approaches for checking heap 190 accesses. Memcheck maintains bitmaps telling it which areas of memory 191 are accessible and which are not. If a memory access falls in an 192 unaccessible area, it reports an error. By marking the 16 bytes 193 before and after an allocated block unaccessible, Memcheck is able to 194 detect small over- and underruns of the block. Similarly, by marking 195 freed memory as unaccessible, Memcheck can detect all accesses to 196 freed memory.</p> 197 <p>Memcheck's approach is simple. But it's also weak. It can't 198 catch block overruns beyond 16 bytes. And, more generally, because it 199 focusses only on the question "is the target address accessible", it 200 fails to detect invalid accesses which just happen to fall within some 201 other valid area. This is not improbable, especially in crowded areas 202 of the process' address space.</p> 203 <p>Ptrcheck's approach is to keep track of pointers derived from 204 heap blocks. It tracks pointers which are derived directly from calls 205 to <code class="function">malloc</code> et al, but also ones derived indirectly, by 206 adding or subtracting offsets from the directly-derived pointers. When a 207 pointer is finally used to access memory, Ptrcheck compares the access 208 address with that of the block it was originally derived from, and 209 reports an error if the access address is not within the block 210 bounds.</p> 211 <p>Consequently Ptrcheck can detect any out of bounds access 212 through a heap-derived pointer, no matter how far from the original 213 block it is.</p> 214 <p>A second advantage is that Ptrcheck is better at detecting 215 accesses to blocks freed very far in the past. Memcheck can detect 216 these too, but only for blocks freed relatively recently. To detect 217 accesses to a freed block, Memcheck must make it inaccessible, hence 218 requiring a space overhead proportional to the size of the block. If 219 the blocks are large, Memcheck will have to make them available for 220 re-allocation relatively quickly, thereby losing the ability to detect 221 invalid accesses to them.</p> 222 <p>By contrast, Ptrcheck has a constant per-block space requirement 223 of four machine words, for detection of accesses to freed blocks. A 224 freed block can be reallocated immediately, yet Ptrcheck can still 225 detect all invalid accesses through any pointers derived from the old 226 allocation, providing only that the four-word descriptor for the old 227 allocation is stored. For example, on a 64-bit machine, to detect 228 accesses in any of the most recently freed 10 million blocks, Ptrcheck 229 will require only 320MB of extra storage. Achieving the same level of 230 detection with Memcheck is close to impossible and would likely 231 involve several gigabytes of extra storage.</p> 232 <p>Having said all that, remember that Memcheck performs uninitialised 233 value checking, invalid and mismatched free checking, overlap checking, and 234 leak checking, none of which Ptrcheck do. Memcheck has also benefitted from 235 years of refinement, tuning, and experience with production-level usage, and 236 so is much faster than Ptrcheck as it currently stands. 237 </p> 238 <p>Consequently we recommend you first make your programs run Memcheck 239 clean. Once that's done, try Ptrcheck to see if you can shake out any 240 further heap, global or stack errors.</p> 241 </div> 242 <div class="sect1" title="11.6.Limitations"> 243 <div class="titlepage"><div><div><h2 class="title" style="clear: both"> 244 <a name="pc-manual.limitations"></a>11.6.Limitations</h2></div></div></div> 245 <p>This is an experimental tool, which relies rather too heavily on some 246 not-as-robust-as-I-would-like assumptions on the behaviour of correct 247 programs. There are a number of limitations which you should be aware 248 of.</p> 249 <div class="itemizedlist"><ul class="itemizedlist" type="disc"> 250 <li class="listitem"><p>Heap checks: Ptrcheck can occasionally lose track of, or 251 become confused about, which heap block a given pointer has been 252 derived from. This can cause it to falsely report errors, or to 253 miss some errors. This is not believed to be a serious 254 problem.</p></li> 255 <li class="listitem"><p>Heap checks: Ptrcheck only tracks pointers that are stored 256 properly aligned in memory. If a pointer is stored at a misaligned 257 address, and then later read again, Ptrcheck will lose track of 258 what it points at. Similar problem if a pointer is split into 259 pieces and later reconsitituted.</p></li> 260 <li class="listitem"><p>Heap checks: Ptrcheck needs to "understand" which system 261 calls return pointers and which don't. Many, but not all system 262 calls are handled. If an unhandled one is encountered, Ptrcheck will 263 abort. Fortunately, adding support for a new syscall is very 264 easy.</p></li> 265 <li class="listitem"><p>Stack checks: It follows from the description above (<a class="xref" href="pc-manual.html#pc-manual.how-works.sg-checks" title="11.4.How Ptrcheck Works: Stack and Global Checks">How Ptrcheck Works: Stack and Global Checks</a>) that the first access by a 266 memory referencing instruction to a stack or global array creates an 267 association between that instruction and the array, which is checked on 268 subsequent accesses by that instruction, until the containing function 269 exits. Hence, the first access by an instruction to an array (in any 270 given function instantiation) is not checked for overrun, since Ptrcheck 271 uses that as the "example" of how subsequent accesses should 272 behave.</p></li> 273 <li class="listitem"> 274 <p>Stack checks: Similarly, and more serious, it is clearly 275 possible to write legitimate pieces of code which break the basic 276 assumption upon which the stack/global checking rests. For 277 example:</p> 278 <pre class="programlisting"> 279 { int a[10], b[10], *p, i; 280 for (i = 0; i < 10; i++) { 281 p = /* arbitrary condition */ ? &a[i] : &b[i]; 282 *p = 42; 283 } 284 } 285 </pre> 286 <p>In this case the store sometimes 287 accesses <code class="computeroutput">a[]</code> and 288 sometimes <code class="computeroutput">b[]</code>, but in no cases is 289 the addressed array overrun. Nevertheless the change in target 290 will cause an error to be reported.</p> 291 <p>It is hard to see how to get around this problem. The only 292 mitigating factor is that such constructions appear very rare, at 293 least judging from the results using the tool so far. Such a 294 construction appears only once in the Valgrind sources (running 295 Valgrind on Valgrind) and perhaps two or three times for a start 296 and exit of Firefox. The best that can be done is to suppress the 297 errors.</p> 298 </li> 299 <li class="listitem"><p>Performance: the stack/global checks require reading all of 300 the DWARF3 type and variable information on the executable and its 301 shared objects. This is computationally expensive and makes 302 startup quite slow. You can expect debuginfo reading time to be in 303 the region of a minute for an OpenOffice sized application, on a 304 2.4 GHz Core 2 machine. Reading this information also requires a 305 lot of memory. To make it viable, Ptrcheck goes to considerable 306 trouble to compress the in-memory representation of the DWARF3 307 data, which is why the process of reading it appears slow.</p></li> 308 <li class="listitem"><p>Performance: Ptrcheck runs slower than Memcheck. This is 309 partly due to a lack of tuning, but partly due to algorithmic 310 difficulties. The heap-check side is potentially quite fast. The 311 stack and global checks can sometimes require a number of range 312 checks per memory access, and these are difficult to short-circuit 313 (despite considerable efforts having been made). 314 </p></li> 315 <li class="listitem"> 316 <p>Coverage: the heap checking is relatively robust, requiring 317 only that Ptrcheck can see calls to <code class="function">malloc</code> et al. 318 In that sense it has debug-info requirements comparable with Memcheck, 319 and is able to heap-check programs even with no debugging information 320 attached.</p> 321 <p>Stack/global checking is much more fragile. If a shared 322 object does not have debug information attached, then Ptrcheck will 323 not be able to determine the bounds of any stack or global arrays 324 defined within that shared object, and so will not be able to check 325 accesses to them. This is true even when those arrays are accessed 326 from some other shared object which was compiled with debug 327 info.</p> 328 <p>At the moment Ptrcheck accepts objects lacking debuginfo 329 without comment. This is dangerous as it causes Ptrcheck to 330 silently skip stack and global checking for such objects. It would 331 be better to print a warning in such circumstances.</p> 332 </li> 333 <li class="listitem"><p>Coverage: Ptrcheck checks that the areas read or written by 334 system calls do not overrun heap blocks. But it doesn't currently 335 check them for overruns stack and global arrays. This would be 336 easy to add.</p></li> 337 <li class="listitem"><p>Platforms: the stack/global checks won't work properly on any 338 PowerPC platforms, only on x86 and amd64 targets. That's because 339 the stack and global checking requires tracking function calls and 340 exits reliably, and there's no obvious way to do it with the PPC 341 ABIs. (In comparison, with the x86 and amd64 ABIs this is relatively 342 straightforward.)</p></li> 343 <li class="listitem"><p>Robustness: related to the previous point. Function 344 call/exit tracking for x86/amd64 is believed to work properly even 345 in the presence of longjmps within the same stack (although this 346 has not been tested). However, code which switches stacks is 347 likely to cause breakage/chaos.</p></li> 348 </ul></div> 349 </div> 350 <div class="sect1" title="11.7.Still To Do: User-visible Functionality"> 351 <div class="titlepage"><div><div><h2 class="title" style="clear: both"> 352 <a name="pc-manual.todo-user-visible"></a>11.7.Still To Do: User-visible Functionality</h2></div></div></div> 353 <div class="itemizedlist"><ul class="itemizedlist" type="disc"> 354 <li class="listitem"><p>Extend system call checking to work on stack and global arrays.</p></li> 355 <li class="listitem"><p>Print a warning if a shared object does not have debug info 356 attached, or if, for whatever reason, debug info could not be 357 found, or read.</p></li> 358 </ul></div> 359 </div> 360 <div class="sect1" title="11.8.Still To Do: Implementation Tidying"> 361 <div class="titlepage"><div><div><h2 class="title" style="clear: both"> 362 <a name="pc-manual.todo-implementation"></a>11.8.Still To Do: Implementation Tidying</h2></div></div></div> 363 <p>Items marked CRITICAL are considered important for correctness: 364 non-fixage of them is liable to lead to crashes or assertion failures 365 in real use.</p> 366 <div class="itemizedlist"><ul class="itemizedlist" type="disc"> 367 <li class="listitem"><p>h_main.c: make N_FREED_SEGS command-line configurable.</p></li> 368 <li class="listitem"><p> sg_main.c: Improve the performance of the stack / global 369 checks by doing some up-front filtering to ignore references in 370 areas which "obviously" can't be stack or globals. This will 371 require using information that m_aspacemgr knows about the address 372 space layout.</p></li> 373 <li class="listitem"><p>h_main.c: get rid of the last_seg_added hack; add suitable 374 plumbing to the core/tool interface to do this cleanly.</p></li> 375 <li class="listitem"><p>h_main.c: move vast amounts of arch-dependent ugliness 376 (get_IntRegInfo et al) to its own source file, a la 377 mc_machine.c.</p></li> 378 <li class="listitem"><p>h_main.c: make the lossage-check stuff work again, as a way 379 of doing quality assurance on the implementation.</p></li> 380 <li class="listitem"><p>h_main.c: schemeEw_Atom: don't generate a call to 381 nonptr_or_unknown, this is really stupid, since it could be done at 382 translation time instead.</p></li> 383 <li class="listitem"><p>CRITICAL: h_main.c: h_instrument (main instrumentation fn): 384 generate shadows for word-sized temps defined in the block's 385 preamble. (Why does this work at all, as it stands?)</p></li> 386 <li class="listitem"><p>sg_main.c: fix compute_II_hash to make it a bit more sensible 387 for ppc32/64 targets (except that sg_ doesn't work on ppc32/64 388 targets, so this is a bit academic at the moment).</p></li> 389 </ul></div> 390 </div> 391 </div> 392 <div> 393 <br><table class="nav" width="100%" cellspacing="3" cellpadding="2" border="0" summary="Navigation footer"> 394 <tr> 395 <td rowspan="2" width="40%" align="left"> 396 <a accesskey="p" href="dh-manual.html"><<10.DHAT: a dynamic heap analysis tool</a></td> 397 <td width="20%" align="center"><a accesskey="u" href="manual.html">Up</a></td> 398 <td rowspan="2" width="40%" align="right"><a accesskey="n" href="bbv-manual.html">12.BBV: an experimental basic block vector generation tool>></a> 399 </td> 400 </tr> 401 <tr><td width="20%" align="center"><a accesskey="h" href="index.html">Home</a></td></tr> 402 </table> 403 </div> 404 </body> 405 </html> 406