Mercurial > emacs
comparison gc/doc/debugging.html @ 51488:5de98dce4bd1
*** empty log message ***
author | Dave Love <fx@gnu.org> |
---|---|
date | Thu, 05 Jun 2003 17:49:22 +0000 |
parents | |
children |
comparison
equal
deleted
inserted
replaced
51487:01d68b199093 | 51488:5de98dce4bd1 |
---|---|
1 <HTML> | |
2 <HEAD> | |
3 <TITLE>Debugging Garbage Collector Related Problems</title> | |
4 </head> | |
5 <BODY> | |
6 <H1>Debugging Garbage Collector Related Problems</h1> | |
7 This page contains some hints on | |
8 debugging issues specific to | |
9 the Boehm-Demers-Weiser conservative garbage collector. | |
10 It applies both to debugging issues in client code that manifest themselves | |
11 as collector misbehavior, and to debugging the collector itself. | |
12 <P> | |
13 If you suspect a bug in the collector itself, it is strongly recommended | |
14 that you try the latest collector release, even if it is labelled as "alpha", | |
15 before proceeding. | |
16 <H2>Bus Errors and Segmentation Violations</h2> | |
17 <P> | |
18 If the fault occurred in GC_find_limit, or with incremental collection enabled, | |
19 this is probably normal. The collector installs handlers to take care of | |
20 these. You will not see these unless you are using a debugger. | |
21 Your debugger <I>should</i> allow you to continue. | |
22 It's often preferable to tell the debugger to ignore SIGBUS and SIGSEGV | |
23 ("<TT>handle SIGSEGV SIGBUS nostop noprint</tt>" in gdb, | |
24 "<TT>ignore SIGSEGV SIGBUS</tt>" in most versions of dbx) | |
25 and set a breakpoint in <TT>abort</tt>. | |
26 The collector will call abort if the signal had another cause, | |
27 and there was not other handler previously installed. | |
28 <P> | |
29 We recommend debugging without incremental collection if possible. | |
30 (This applies directly to UNIX systems. | |
31 Debugging with incremental collection under win32 is worse. See README.win32.) | |
32 <P> | |
33 If the application generates an unhandled SIGSEGV or equivalent, it may | |
34 often be easiest to set the environment variable GC_LOOP_ON_ABORT. On many | |
35 platforms, this will cause the collector to loop in a handler when the | |
36 SIGSEGV is encountered (or when the collector aborts for some other reason), | |
37 and a debugger can then be attached to the looping | |
38 process. This sidesteps common operating system problems related | |
39 to incomplete core files for multithreaded applications, etc. | |
40 <H2>Other Signals</h2> | |
41 On most platforms, the multithreaded version of the collector needs one or | |
42 two other signals for internal use by the collector in stopping threads. | |
43 It is normally wise to tell the debugger to ignore these. On Linux, | |
44 the collector currently uses SIGPWR and SIGXCPU by default. | |
45 <H2>Warning Messages About Needing to Allocate Blacklisted Blocks</h2> | |
46 The garbage collector generates warning messages of the form | |
47 <PRE> | |
48 Needed to allocate blacklisted block at 0x... | |
49 </pre> | |
50 when it needs to allocate a block at a location that it knows to be | |
51 referenced by a false pointer. These false pointers can be either permanent | |
52 (<I>e.g.</i> a static integer variable that never changes) or temporary. | |
53 In the latter case, the warning is largely spurious, and the block will | |
54 eventually be reclaimed normally. | |
55 In the former case, the program will still run correctly, but the block | |
56 will never be reclaimed. Unless the block is intended to be | |
57 permanent, the warning indicates a memory leak. | |
58 <OL> | |
59 <LI>Ignore these warnings while you are using GC_DEBUG. Some of the routines | |
60 mentioned below don't have debugging equivalents. (Alternatively, write | |
61 the missing routines and send them to me.) | |
62 <LI>Replace allocator calls that request large blocks with calls to | |
63 <TT>GC_malloc_ignore_off_page</tt> or | |
64 <TT>GC_malloc_atomic_ignore_off_page</tt>. You may want to set a | |
65 breakpoint in <TT>GC_default_warn_proc</tt> to help you identify such calls. | |
66 Make sure that a pointer to somewhere near the beginning of the resulting block | |
67 is maintained in a (preferably volatile) variable as long as | |
68 the block is needed. | |
69 <LI> | |
70 If the large blocks are allocated with realloc, we suggest instead allocating | |
71 them with something like the following. Note that the realloc size increment | |
72 should be fairly large (e.g. a factor of 3/2) for this to exhibit reasonable | |
73 performance. But we all know we should do that anyway. | |
74 <PRE> | |
75 void * big_realloc(void *p, size_t new_size) | |
76 { | |
77 size_t old_size = GC_size(p); | |
78 void * result; | |
79 | |
80 if (new_size <= 10000) return(GC_realloc(p, new_size)); | |
81 if (new_size <= old_size) return(p); | |
82 result = GC_malloc_ignore_off_page(new_size); | |
83 if (result == 0) return(0); | |
84 memcpy(result,p,old_size); | |
85 GC_free(p); | |
86 return(result); | |
87 } | |
88 </pre> | |
89 | |
90 <LI> In the unlikely case that even relatively small object | |
91 (<20KB) allocations are triggering these warnings, then your address | |
92 space contains lots of "bogus pointers", i.e. values that appear to | |
93 be pointers but aren't. Usually this can be solved by using GC_malloc_atomic | |
94 or the routines in gc_typed.h to allocate large pointer-free regions of bitmaps, etc. Sometimes the problem can be solved with trivial changes of encoding | |
95 in certain values. It is possible, to identify the source of the bogus | |
96 pointers by building the collector with <TT>-DPRINT_BLACK_LIST</tt>, | |
97 which will cause it to print the "bogus pointers", along with their location. | |
98 | |
99 <LI> If you get only a fixed number of these warnings, you are probably only | |
100 introducing a bounded leak by ignoring them. If the data structures being | |
101 allocated are intended to be permanent, then it is also safe to ignore them. | |
102 The warnings can be turned off by calling GC_set_warn_proc with a procedure | |
103 that ignores these warnings (e.g. by doing absolutely nothing). | |
104 </ol> | |
105 | |
106 <H2>The Collector References a Bad Address in <TT>GC_malloc</tt></h2> | |
107 | |
108 This typically happens while the collector is trying to remove an entry from | |
109 its free list, and the free list pointer is bad because the free list link | |
110 in the last allocated object was bad. | |
111 <P> | |
112 With > 99% probability, you wrote past the end of an allocated object. | |
113 Try setting <TT>GC_DEBUG</tt> before including <TT>gc.h</tt> and | |
114 allocating with <TT>GC_MALLOC</tt>. This will try to detect such | |
115 overwrite errors. | |
116 | |
117 <H2>Unexpectedly Large Heap</h2> | |
118 | |
119 Unexpected heap growth can be due to one of the following: | |
120 <OL> | |
121 <LI> Data structures that are being unintentionally retained. This | |
122 is commonly caused by data structures that are no longer being used, | |
123 but were not cleared, or by caches growing without bounds. | |
124 <LI> Pointer misidentification. The garbage collector is interpreting | |
125 integers or other data as pointers and retaining the "referenced" | |
126 objects. | |
127 <LI> Heap fragmentation. This should never result in unbounded growth, | |
128 but it may account for larger heaps. This is most commonly caused | |
129 by allocation of large objects. On some platforms it can be reduced | |
130 by building with -DUSE_MUNMAP, which will cause the collector to unmap | |
131 memory corresponding to pages that have not been recently used. | |
132 <LI> Per object overhead. This is usually a relatively minor effect, but | |
133 it may be worth considering. If the collector recognizes interior | |
134 pointers, object sizes are increased, so that one-past-the-end pointers | |
135 are correctly recognized. The collector can be configured not to do this | |
136 (<TT>-DDONT_ADD_BYTE_AT_END</tt>). | |
137 <P> | |
138 The collector rounds up object sizes so the result fits well into the | |
139 chunk size (<TT>HBLKSIZE</tt>, normally 4K on 32 bit machines, 8K | |
140 on 64 bit machines) used by the collector. Thus it may be worth avoiding | |
141 objects of size 2K + 1 (or 2K if a byte is being added at the end.) | |
142 </ol> | |
143 The last two cases can often be identified by looking at the output | |
144 of a call to <TT>GC_dump()</tt>. Among other things, it will print the | |
145 list of free heap blocks, and a very brief description of all chunks in | |
146 the heap, the object sizes they correspond to, and how many live objects | |
147 were found in the chunk at the last collection. | |
148 <P> | |
149 Growing data structures can usually be identified by | |
150 <OL> | |
151 <LI> Building the collector with <TT>-DKEEP_BACK_PTRS</tt>, | |
152 <LI> Preferably using debugging allocation (defining <TT>GC_DEBUG</tt> | |
153 before including <TT>gc.h</tt> and allocating with <TT>GC_MALLOC</tt>), | |
154 so that objects will be identified by their allocation site, | |
155 <LI> Running the application long enough so | |
156 that most of the heap is composed of "leaked" memory, and | |
157 <LI> Then calling <TT>GC_generate_random_backtrace()</tt> from backptr.h | |
158 a few times to determine why some randomly sampled objects in the heap are | |
159 being retained. | |
160 </ol> | |
161 <P> | |
162 The same technique can often be used to identify problems with false | |
163 pointers, by noting whether the reference chains printed by | |
164 <TT>GC_generate_random_backtrace()</tt> involve any misidentified pointers. | |
165 An alternate technique is to build the collector with | |
166 <TT>-DPRINT_BLACK_LIST</tt> which will cause it to report values that | |
167 are almost, but not quite, look like heap pointers. It is very likely that | |
168 actual false pointers will come from similar sources. | |
169 <P> | |
170 In the unlikely case that false pointers are an issue, it can usually | |
171 be resolved using one or more of the following techniques: | |
172 <OL> | |
173 <LI> Use <TT>GC_malloc_atomic</tt> for objects containing no pointers. | |
174 This is especially important for large arrays containing compressed data, | |
175 pseudo-random numbers, and the like. It is also likely to improve GC | |
176 performance, perhaps drastically so if the application is paging. | |
177 <LI> If you allocate large objects containing only | |
178 one or two pointers at the beginning, either try the typed allocation | |
179 primitives is <TT>gc_typed.h</tt>, or separate out the pointerfree component. | |
180 <LI> Consider using <TT>GC_malloc_ignore_off_page()</tt> | |
181 to allocate large objects. (See <TT>gc.h</tt> and above for details. | |
182 Large means > 100K in most environments.) | |
183 </ol> | |
184 <H2>Prematurely Reclaimed Objects</h2> | |
185 The usual symptom of this is a segmentation fault, or an obviously overwritten | |
186 value in a heap object. This should, of course, be impossible. In practice, | |
187 it may happen for reasons like the following: | |
188 <OL> | |
189 <LI> The collector did not intercept the creation of threads correctly in | |
190 a multithreaded application, <I>e.g.</i> because the client called | |
191 <TT>pthread_create</tt> without including <TT>gc.h</tt>, which redefines it. | |
192 <LI> The last pointer to an object in the garbage collected heap was stored | |
193 somewhere were the collector couldn't see it, <I>e.g.</i> in an | |
194 object allocated with system <TT>malloc</tt>, in certain types of | |
195 <TT>mmap</tt>ed files, | |
196 or in some data structure visible only to the OS. (On some platforms, | |
197 thread-local storage is one of these.) | |
198 <LI> The last pointer to an object was somehow disguised, <I>e.g.</i> by | |
199 XORing it with another pointer. | |
200 <LI> Incorrect use of <TT>GC_malloc_atomic</tt> or typed allocation. | |
201 <LI> An incorrect <TT>GC_free</tt> call. | |
202 <LI> The client program overwrote an internal garbage collector data structure. | |
203 <LI> A garbage collector bug. | |
204 <LI> (Empirically less likely than any of the above.) A compiler optimization | |
205 that disguised the last pointer. | |
206 </ol> | |
207 The following relatively simple techniques should be tried first to narrow | |
208 down the problem: | |
209 <OL> | |
210 <LI> If you are using the incremental collector try turning it off for | |
211 debugging. | |
212 <LI> If you are using shared libraries, try linking statically. If that works, | |
213 ensure that DYNAMIC_LOADING is defined on your platform. | |
214 <LI> Try to reproduce the problem with fully debuggable unoptimized code. | |
215 This will eliminate the last possibility, as well as making debugging easier. | |
216 <LI> Try replacing any suspect typed allocation and <TT>GC_malloc_atomic</tt> | |
217 calls with calls to <TT>GC_malloc</tt>. | |
218 <LI> Try removing any GC_free calls (<I>e.g.</i> with a suitable | |
219 <TT>#define</tt>). | |
220 <LI> Rebuild the collector with <TT>-DGC_ASSERTIONS</tt>. | |
221 <LI> If the following works on your platform (i.e. if gctest still works | |
222 if you do this), try building the collector with | |
223 <TT>-DREDIRECT_MALLOC=GC_malloc_uncollectable</tt>. This will cause | |
224 the collector to scan memory allocated with malloc. | |
225 </ol> | |
226 If all else fails, you will have to attack this with a debugger. | |
227 Suggested steps: | |
228 <OL> | |
229 <LI> Call <TT>GC_dump()</tt> from the debugger around the time of the failure. Verify | |
230 that the collectors idea of the root set (i.e. static data regions which | |
231 it should scan for pointers) looks plausible. If not, i.e. if it doesn't | |
232 include some static variables, report this as | |
233 a collector bug. Be sure to describe your platform precisely, since this sort | |
234 of problem is nearly always very platform dependent. | |
235 <LI> Especially if the failure is not deterministic, try to isolate it to | |
236 a relatively small test case. | |
237 <LI> Set a break point in <TT>GC_finish_collection</tt>. This is a good | |
238 point to examine what has been marked, i.e. found reachable, by the | |
239 collector. | |
240 <LI> If the failure is deterministic, run the process | |
241 up to the last collection before the failure. | |
242 Note that the variable <TT>GC_gc_no</tt> counts collections and can be used | |
243 to set a conditional breakpoint in the right one. It is incremented just | |
244 before the call to GC_finish_collection. | |
245 If object <TT>p</tt> was prematurely recycled, it may be helpful to | |
246 look at <TT>*GC_find_header(p)</tt> at the failure point. | |
247 The <TT>hb_last_reclaimed</tt> field will identify the collection number | |
248 during which its block was last swept. | |
249 <LI> Verify that the offending object still has its correct contents at | |
250 this point. | |
251 Then call <TT>GC_is_marked(p)</tt> from the debugger to verify that the | |
252 object has not been marked, and is about to be reclaimed. Note that | |
253 <TT>GC_is_marked(p)</tt> expects the real address of an object (the | |
254 address of the debug header if there is one), and thus it may | |
255 be more appropriate to call <TT>GC_is_marked(GC_base(p))</tt> | |
256 instead. | |
257 <LI> Determine a path from a root, i.e. static variable, stack, or | |
258 register variable, | |
259 to the reclaimed object. Call <TT>GC_is_marked(q)</tt> for each object | |
260 <TT>q</tt> along the path, trying to locate the first unmarked object, say | |
261 <TT>r</tt>. | |
262 <LI> If <TT>r</tt> is pointed to by a static root, | |
263 verify that the location | |
264 pointing to it is part of the root set printed by <TT>GC_dump()</tt>. If it | |
265 is on the stack in the main (or only) thread, verify that | |
266 <TT>GC_stackbottom</tt> is set correctly to the base of the stack. If it is | |
267 in another thread stack, check the collector's thread data structure | |
268 (<TT>GC_thread[]</tt> on several platforms) to make sure that stack bounds | |
269 are set correctly. | |
270 <LI> If <TT>r</tt> is pointed to by heap object <TT>s</tt>, check that the | |
271 collector's layout description for <TT>s</tt> is such that the pointer field | |
272 will be scanned. Call <TT>*GC_find_header(s)</tt> to look at the descriptor | |
273 for the heap chunk. The <TT>hb_descr</tt> field specifies the layout | |
274 of objects in that chunk. See gc_mark.h for the meaning of the descriptor. | |
275 (If it's low order 2 bits are zero, then it is just the length of the | |
276 object prefix to be scanned. This form is always used for objects allocated | |
277 with <TT>GC_malloc</tt> or <TT>GC_malloc_atomic</tt>.) | |
278 <LI> If the failure is not deterministic, you may still be able to apply some | |
279 of the above technique at the point of failure. But remember that objects | |
280 allocated since the last collection will not have been marked, even if the | |
281 collector is functioning properly. On some platforms, the collector | |
282 can be configured to save call chains in objects for debugging. | |
283 Enabling this feature will also cause it to save the call stack at the | |
284 point of the last GC in GC_arrays._last_stack. | |
285 <LI> When looking at GC internal data structures remember that a number | |
286 of <TT>GC_</tt><I>xxx</i> variables are really macro defined to | |
287 <TT>GC_arrays._</tt><I>xxx</i>, so that | |
288 the collector can avoid scanning them. | |
289 </ol> | |
290 </body> | |
291 </html> | |
292 | |
293 | |
294 | |
295 |