blob: 4e6e28e67daae03b80e1d457da33968a26198754 [file] [log] [blame]
njn43c799e2003-04-08 00:08:52 +00001
2/*--------------------------------------------------------------------*/
njn1d0825f2006-03-27 11:37:07 +00003/*--- The leak checker. mc_leakcheck.c ---*/
njn43c799e2003-04-08 00:08:52 +00004/*--------------------------------------------------------------------*/
5
6/*
nethercote137bc552003-11-14 17:47:54 +00007 This file is part of MemCheck, a heavyweight Valgrind tool for
njn1d0825f2006-03-27 11:37:07 +00008 detecting memory errors.
njn43c799e2003-04-08 00:08:52 +00009
sewardj0f157dd2013-10-18 14:27:36 +000010 Copyright (C) 2000-2013 Julian Seward
njn43c799e2003-04-08 00:08:52 +000011 jseward@acm.org
12
13 This program is free software; you can redistribute it and/or
14 modify it under the terms of the GNU General Public License as
15 published by the Free Software Foundation; either version 2 of the
16 License, or (at your option) any later version.
17
18 This program is distributed in the hope that it will be useful, but
19 WITHOUT ANY WARRANTY; without even the implied warranty of
20 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
21 General Public License for more details.
22
23 You should have received a copy of the GNU General Public License
24 along with this program; if not, write to the Free Software
25 Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
26 02111-1307, USA.
27
28 The GNU General Public License is contained in the file COPYING.
29*/
30
njnc7561b92005-06-19 01:24:32 +000031#include "pub_tool_basics.h"
sewardj4cfea4f2006-10-14 19:26:10 +000032#include "pub_tool_vki.h"
njnac1e0332009-05-08 00:39:31 +000033#include "pub_tool_aspacehl.h"
njn4802b382005-06-11 04:58:29 +000034#include "pub_tool_aspacemgr.h"
njn1d0825f2006-03-27 11:37:07 +000035#include "pub_tool_execontext.h"
36#include "pub_tool_hashtable.h"
njn97405b22005-06-02 03:39:33 +000037#include "pub_tool_libcbase.h"
njn132bfcc2005-06-04 19:16:06 +000038#include "pub_tool_libcassert.h"
njn36a20fa2005-06-03 03:08:39 +000039#include "pub_tool_libcprint.h"
njnde62cbf2005-06-10 22:08:14 +000040#include "pub_tool_libcsignal.h"
njn6ace3ea2005-06-17 03:06:27 +000041#include "pub_tool_machine.h"
njnc7561b92005-06-19 01:24:32 +000042#include "pub_tool_mallocfree.h"
43#include "pub_tool_options.h"
njn29a5c012009-05-06 06:15:55 +000044#include "pub_tool_oset.h"
philippe6643e962012-01-17 21:16:30 +000045#include "pub_tool_poolalloc.h"
46#include "pub_tool_signals.h" // Needed for mc_include.h
sewardj6c591e12011-04-11 16:17:51 +000047#include "pub_tool_libcsetjmp.h" // setjmp facilities
njn1d0825f2006-03-27 11:37:07 +000048#include "pub_tool_tooliface.h" // Needed for mc_include.h
njn43c799e2003-04-08 00:08:52 +000049
njn1d0825f2006-03-27 11:37:07 +000050#include "mc_include.h"
njnc7561b92005-06-19 01:24:32 +000051
njn8225cc02009-03-09 22:52:24 +000052/*------------------------------------------------------------*/
53/*--- An overview of leak checking. ---*/
54/*------------------------------------------------------------*/
njnc7561b92005-06-19 01:24:32 +000055
njn8225cc02009-03-09 22:52:24 +000056// Leak-checking is a directed-graph traversal problem. The graph has
57// two kinds of nodes:
58// - root-set nodes:
59// - GP registers of all threads;
60// - valid, aligned, pointer-sized data words in valid client memory,
61// including stacks, but excluding words within client heap-allocated
62// blocks (they are excluded so that later on we can differentiate
63// between heap blocks that are indirectly leaked vs. directly leaked).
64// - heap-allocated blocks. A block is a mempool chunk or a malloc chunk
65// that doesn't contain a mempool chunk. Nb: the terms "blocks" and
66// "chunks" are used interchangeably below.
67//
68// There are two kinds of edges:
69// - start-pointers, i.e. pointers to the start of a block;
70// - interior-pointers, i.e. pointers to the interior of a block.
71//
72// We use "pointers" rather than "edges" below.
73//
74// Root set nodes only point to blocks. Blocks only point to blocks;
75// a block can point to itself.
76//
77// The aim is to traverse the graph and determine the status of each block.
78//
79// There are 9 distinct cases. See memcheck/docs/mc-manual.xml for details.
80// Presenting all nine categories to the user is probably too much.
81// Currently we do this:
82// - definitely lost: case 3
83// - indirectly lost: case 4, 9
84// - possibly lost: cases 5..8
85// - still reachable: cases 1, 2
86//
87// It's far from clear that this is the best possible categorisation; it's
88// accreted over time without any central guiding principle.
89
90/*------------------------------------------------------------*/
91/*--- XXX: Thoughts for improvement. ---*/
92/*------------------------------------------------------------*/
93
94// From the user's point of view:
95// - If they aren't using interior-pointers, they just have to fix the
96// directly lost blocks, and the indirectly lost ones will be fixed as
97// part of that. Any possibly lost blocks will just be due to random
98// pointer garbage and can be ignored.
99//
100// - If they are using interior-pointers, the fact that they currently are not
101// being told which ones might be directly lost vs. indirectly lost makes
102// it hard to know where to begin.
103//
104// All this makes me wonder if new option is warranted:
105// --follow-interior-pointers. By default it would be off, the leak checker
106// wouldn't follow interior-pointers and there would only be 3 categories:
107// R, DL, IL.
108//
109// If turned on, then it would show 7 categories (R, DL, IL, DR/DL, IR/IL,
110// IR/IL/DL, IL/DL). That output is harder to understand but it's your own
111// damn fault for using interior-pointers...
112//
113// ----
114//
115// Also, why are two blank lines printed between each loss record?
njnc2f8b1b2009-08-10 06:47:00 +0000116// [bug 197930]
njn8225cc02009-03-09 22:52:24 +0000117//
118// ----
119//
120// Also, --show-reachable is a bad name because it also turns on the showing
121// of indirectly leaked blocks(!) It would be better named --show-all or
122// --show-all-heap-blocks, because that's the end result.
philippe2193a7c2012-12-08 17:54:16 +0000123// We now have the option --show-leak-kinds=... which allows to specify =all.
njn8225cc02009-03-09 22:52:24 +0000124//
125// ----
126//
127// Also, the VALGRIND_LEAK_CHECK and VALGRIND_QUICK_LEAK_CHECK aren't great
128// names. VALGRIND_FULL_LEAK_CHECK and VALGRIND_SUMMARY_LEAK_CHECK would be
129// better.
130//
131// ----
132//
133// Also, VALGRIND_COUNT_LEAKS and VALGRIND_COUNT_LEAK_BLOCKS aren't great as
134// they combine direct leaks and indirect leaks into one. New, more precise
135// ones (they'll need new names) would be good. If more categories are
136// used, as per the --follow-interior-pointers option, they should be
137// updated accordingly. And they should use a struct to return the values.
138//
139// ----
140//
141// Also, for this case:
142//
143// (4) p4 BBB ---> AAA
144//
145// BBB is definitely directly lost. AAA is definitely indirectly lost.
146// Here's the relevant loss records printed for a full check (each block is
147// 16 bytes):
148//
149// ==20397== 16 bytes in 1 blocks are indirectly lost in loss record 9 of 15
150// ==20397== at 0x4C2694E: malloc (vg_replace_malloc.c:177)
151// ==20397== by 0x400521: mk (leak-cases.c:49)
152// ==20397== by 0x400578: main (leak-cases.c:72)
153//
154// ==20397== 32 (16 direct, 16 indirect) bytes in 1 blocks are definitely
155// lost in loss record 14 of 15
156// ==20397== at 0x4C2694E: malloc (vg_replace_malloc.c:177)
157// ==20397== by 0x400521: mk (leak-cases.c:49)
158// ==20397== by 0x400580: main (leak-cases.c:72)
159//
160// The first one is fine -- it describes AAA.
161//
162// The second one is for BBB. It's correct in that 16 bytes in 1 block are
163// directly lost. It's also correct that 16 are indirectly lost as a result,
164// but it means that AAA is being counted twice in the loss records. (It's
165// not, thankfully, counted twice in the summary counts). Argh.
166//
167// This would be less confusing for the second one:
168//
169// ==20397== 16 bytes in 1 blocks are definitely lost in loss record 14
170// of 15 (and 16 bytes in 1 block are indirectly lost as a result; they
philippe2193a7c2012-12-08 17:54:16 +0000171// are mentioned elsewhere (if --show-reachable=yes or indirect is given
172// in --show-leak-kinds=... !))
njn8225cc02009-03-09 22:52:24 +0000173// ==20397== at 0x4C2694E: malloc (vg_replace_malloc.c:177)
174// ==20397== by 0x400521: mk (leak-cases.c:49)
175// ==20397== by 0x400580: main (leak-cases.c:72)
176//
177// But ideally we'd present the loss record for the directly lost block and
178// then the resultant indirectly lost blocks and make it clear the
179// dependence. Double argh.
180
181/*------------------------------------------------------------*/
182/*--- The actual algorithm. ---*/
183/*------------------------------------------------------------*/
184
185// - Find all the blocks (a.k.a. chunks) to check. Mempool chunks require
186// some special treatment because they can be within malloc'd blocks.
187// - Scan every word in the root set (GP registers and valid
188// non-heap memory words).
189// - First, we skip if it doesn't point to valid memory.
190// - Then, we see if it points to the start or interior of a block. If
191// so, we push the block onto the mark stack and mark it as having been
192// reached.
193// - Then, we process the mark stack, repeating the scanning for each block;
194// this can push more blocks onto the mark stack. We repeat until the
195// mark stack is empty. Each block is marked as definitely or possibly
196// reachable, depending on whether interior-pointers were required to
197// reach it.
198// - At this point we know for every block if it's reachable or not.
199// - We then push each unreached block onto the mark stack, using the block
200// number as the "clique" number.
201// - We process the mark stack again, this time grouping blocks into cliques
202// in order to facilitate the directly/indirectly lost categorisation.
203// - We group blocks by their ExeContexts and categorisation, and print them
204// if --leak-check=full. We also print summary numbers.
205//
206// A note on "cliques":
207// - A directly lost block is one with no pointers to it. An indirectly
208// lost block is one that is pointed to by a directly or indirectly lost
209// block.
210// - Each directly lost block has zero or more indirectly lost blocks
211// hanging off it. All these blocks together form a "clique". The
212// directly lost block is called the "clique leader". The clique number
213// is the number (in lc_chunks[]) of the clique leader.
214// - Actually, a directly lost block may be pointed to if it's part of a
215// cycle. In that case, there may be more than one choice for the clique
216// leader, and the choice is arbitrary. Eg. if you have A-->B and B-->A
217// either A or B could be the clique leader.
218// - Cliques cannot overlap, and will be truncated to avoid this. Eg. if we
219// have A-->C and B-->C, the two cliques will be {A,C} and {B}, or {A} and
220// {B,C} (again the choice is arbitrary). This is because we don't want
221// to count a block as indirectly lost more than once.
222//
223// A note on 'is_prior_definite':
224// - This is a boolean used in various places that indicates if the chain
225// up to the prior node (prior to the one being considered) is definite.
226// - In the clique == -1 case:
227// - if True it means that the prior node is a root-set node, or that the
228// prior node is a block which is reachable from the root-set via
229// start-pointers.
230// - if False it means that the prior node is a block that is only
231// reachable from the root-set via a path including at least one
232// interior-pointer.
233// - In the clique != -1 case, currently it's always True because we treat
234// start-pointers and interior-pointers the same for direct/indirect leak
235// checking. If we added a PossibleIndirectLeak state then this would
236// change.
237
238
239// Define to debug the memory-leak-detector.
sewardjb5f6f512005-03-10 23:59:00 +0000240#define VG_DEBUG_LEAKCHECK 0
njn8225cc02009-03-09 22:52:24 +0000241#define VG_DEBUG_CLIQUE 0
242
sewardjb5f6f512005-03-10 23:59:00 +0000243
njn43c799e2003-04-08 00:08:52 +0000244/*------------------------------------------------------------*/
njn8225cc02009-03-09 22:52:24 +0000245/*--- Getting the initial chunks, and searching them. ---*/
njn43c799e2003-04-08 00:08:52 +0000246/*------------------------------------------------------------*/
247
njn8225cc02009-03-09 22:52:24 +0000248// Compare the MC_Chunks by 'data' (i.e. the address of the block).
florian6bd9dc12012-11-23 16:17:43 +0000249static Int compare_MC_Chunks(const void* n1, const void* n2)
njn43c799e2003-04-08 00:08:52 +0000250{
florian3e798632012-11-24 19:41:54 +0000251 const MC_Chunk* mc1 = *(const MC_Chunk *const *)n1;
252 const MC_Chunk* mc2 = *(const MC_Chunk *const *)n2;
njn8225cc02009-03-09 22:52:24 +0000253 if (mc1->data < mc2->data) return -1;
254 if (mc1->data > mc2->data) return 1;
255 return 0;
njn43c799e2003-04-08 00:08:52 +0000256}
257
njn8225cc02009-03-09 22:52:24 +0000258#if VG_DEBUG_LEAKCHECK
259// Used to sanity-check the fast binary-search mechanism.
260static
261Int find_chunk_for_OLD ( Addr ptr,
262 MC_Chunk** chunks,
263 Int n_chunks )
264
265{
266 Int i;
267 Addr a_lo, a_hi;
268 PROF_EVENT(70, "find_chunk_for_OLD");
269 for (i = 0; i < n_chunks; i++) {
270 PROF_EVENT(71, "find_chunk_for_OLD(loop)");
271 a_lo = chunks[i]->data;
272 a_hi = ((Addr)chunks[i]->data) + chunks[i]->szB;
273 if (a_lo <= ptr && ptr < a_hi)
274 return i;
275 }
276 return -1;
277}
278#endif
279
280// Find the i such that ptr points at or inside the block described by
281// chunks[i]. Return -1 if none found. This assumes that chunks[]
282// has been sorted on the 'data' field.
283static
284Int find_chunk_for ( Addr ptr,
285 MC_Chunk** chunks,
286 Int n_chunks )
287{
288 Addr a_mid_lo, a_mid_hi;
289 Int lo, mid, hi, retVal;
290 // VG_(printf)("find chunk for %p = ", ptr);
291 retVal = -1;
292 lo = 0;
293 hi = n_chunks-1;
294 while (True) {
295 // Invariant: current unsearched space is from lo to hi, inclusive.
296 if (lo > hi) break; // not found
297
298 mid = (lo + hi) / 2;
299 a_mid_lo = chunks[mid]->data;
300 a_mid_hi = chunks[mid]->data + chunks[mid]->szB;
301 // Extent of block 'mid' is [a_mid_lo .. a_mid_hi).
302 // Special-case zero-sized blocks - treat them as if they had
303 // size 1. Not doing so causes them to not cover any address
304 // range at all and so will never be identified as the target of
305 // any pointer, which causes them to be incorrectly reported as
306 // definitely leaked.
307 if (chunks[mid]->szB == 0)
308 a_mid_hi++;
309
310 if (ptr < a_mid_lo) {
311 hi = mid-1;
312 continue;
313 }
314 if (ptr >= a_mid_hi) {
315 lo = mid+1;
316 continue;
317 }
318 tl_assert(ptr >= a_mid_lo && ptr < a_mid_hi);
319 retVal = mid;
320 break;
321 }
322
323# if VG_DEBUG_LEAKCHECK
324 tl_assert(retVal == find_chunk_for_OLD ( ptr, chunks, n_chunks ));
325# endif
326 // VG_(printf)("%d\n", retVal);
327 return retVal;
328}
329
330
331static MC_Chunk**
florian54fe2022012-10-27 23:07:42 +0000332find_active_chunks(Int* pn_chunks)
njn8225cc02009-03-09 22:52:24 +0000333{
334 // Our goal is to construct a set of chunks that includes every
335 // mempool chunk, and every malloc region that *doesn't* contain a
336 // mempool chunk.
337 MC_Mempool *mp;
338 MC_Chunk **mallocs, **chunks, *mc;
339 UInt n_mallocs, n_chunks, m, s;
340 Bool *malloc_chunk_holds_a_pool_chunk;
341
342 // First we collect all the malloc chunks into an array and sort it.
343 // We do this because we want to query the chunks by interior
344 // pointers, requiring binary search.
345 mallocs = (MC_Chunk**) VG_(HT_to_array)( MC_(malloc_list), &n_mallocs );
346 if (n_mallocs == 0) {
347 tl_assert(mallocs == NULL);
348 *pn_chunks = 0;
349 return NULL;
350 }
351 VG_(ssort)(mallocs, n_mallocs, sizeof(VgHashNode*), compare_MC_Chunks);
352
353 // Then we build an array containing a Bool for each malloc chunk,
354 // indicating whether it contains any mempools.
355 malloc_chunk_holds_a_pool_chunk = VG_(calloc)( "mc.fas.1",
356 n_mallocs, sizeof(Bool) );
357 n_chunks = n_mallocs;
358
359 // Then we loop over the mempool tables. For each chunk in each
360 // pool, we set the entry in the Bool array corresponding to the
361 // malloc chunk containing the mempool chunk.
362 VG_(HT_ResetIter)(MC_(mempool_list));
363 while ( (mp = VG_(HT_Next)(MC_(mempool_list))) ) {
364 VG_(HT_ResetIter)(mp->chunks);
365 while ( (mc = VG_(HT_Next)(mp->chunks)) ) {
366
367 // We'll need to record this chunk.
368 n_chunks++;
369
370 // Possibly invalidate the malloc holding the beginning of this chunk.
371 m = find_chunk_for(mc->data, mallocs, n_mallocs);
372 if (m != -1 && malloc_chunk_holds_a_pool_chunk[m] == False) {
373 tl_assert(n_chunks > 0);
374 n_chunks--;
375 malloc_chunk_holds_a_pool_chunk[m] = True;
376 }
377
378 // Possibly invalidate the malloc holding the end of this chunk.
379 if (mc->szB > 1) {
380 m = find_chunk_for(mc->data + (mc->szB - 1), mallocs, n_mallocs);
381 if (m != -1 && malloc_chunk_holds_a_pool_chunk[m] == False) {
382 tl_assert(n_chunks > 0);
383 n_chunks--;
384 malloc_chunk_holds_a_pool_chunk[m] = True;
385 }
386 }
387 }
388 }
389 tl_assert(n_chunks > 0);
390
391 // Create final chunk array.
392 chunks = VG_(malloc)("mc.fas.2", sizeof(VgHashNode*) * (n_chunks));
393 s = 0;
394
395 // Copy the mempool chunks and the non-marked malloc chunks into a
396 // combined array of chunks.
397 VG_(HT_ResetIter)(MC_(mempool_list));
398 while ( (mp = VG_(HT_Next)(MC_(mempool_list))) ) {
399 VG_(HT_ResetIter)(mp->chunks);
400 while ( (mc = VG_(HT_Next)(mp->chunks)) ) {
401 tl_assert(s < n_chunks);
402 chunks[s++] = mc;
403 }
404 }
405 for (m = 0; m < n_mallocs; ++m) {
406 if (!malloc_chunk_holds_a_pool_chunk[m]) {
407 tl_assert(s < n_chunks);
408 chunks[s++] = mallocs[m];
409 }
410 }
411 tl_assert(s == n_chunks);
412
413 // Free temporaries.
414 VG_(free)(mallocs);
415 VG_(free)(malloc_chunk_holds_a_pool_chunk);
416
417 *pn_chunks = n_chunks;
418
419 return chunks;
420}
421
422/*------------------------------------------------------------*/
423/*--- The leak detector proper. ---*/
424/*------------------------------------------------------------*/
425
426// Holds extra info about each block during leak checking.
427typedef
428 struct {
429 UInt state:2; // Reachedness.
philippeab1fce92013-09-29 13:47:32 +0000430 UInt pending:1; // Scan pending.
431 UInt heuristic: (sizeof(UInt)*8)-3;
432 // Heuristic with which this block was considered reachable.
433 // LchNone if state != Reachable or no heuristic needed to
434 // consider it reachable.
435
philippea22f59d2012-01-26 23:13:52 +0000436 union {
philippeab1fce92013-09-29 13:47:32 +0000437 SizeT indirect_szB;
438 // If Unreached, how many bytes are unreachable from here.
439 SizeT clique;
440 // if IndirectLeak, clique leader to which it belongs.
philippea22f59d2012-01-26 23:13:52 +0000441 } IorC;
njn8225cc02009-03-09 22:52:24 +0000442 }
443 LC_Extra;
444
445// An array holding pointers to every chunk we're checking. Sorted by address.
philippea22f59d2012-01-26 23:13:52 +0000446// lc_chunks is initialised during leak search. It is kept after leak search
447// to support printing the list of blocks belonging to a loss record.
448// lc_chunk array can only be used validly till the next "free" operation
449// (as a free operation potentially destroys one or more chunks).
450// To detect lc_chunk is valid, we store the nr of frees operations done
451// when lc_chunk was build : lc_chunks (and lc_extras) stays valid as
452// long as no free operations has been done since lc_chunks building.
njn8225cc02009-03-09 22:52:24 +0000453static MC_Chunk** lc_chunks;
454// How many chunks we're dealing with.
455static Int lc_n_chunks;
philippea22f59d2012-01-26 23:13:52 +0000456static SizeT lc_chunks_n_frees_marker;
457// This has the same number of entries as lc_chunks, and each entry
458// in lc_chunks corresponds with the entry here (ie. lc_chunks[i] and
459// lc_extras[i] describe the same block).
460static LC_Extra* lc_extras;
461
sewardjc8bd1df2011-06-26 12:41:33 +0000462// chunks will be converted and merged in loss record, maintained in lr_table
463// lr_table elements are kept from one leak_search to another to implement
464// the "print new/changed leaks" client request
465static OSet* lr_table;
philippea22f59d2012-01-26 23:13:52 +0000466// Array of sorted loss record (produced during last leak search).
467static LossRecord** lr_array;
468
philippeab1fce92013-09-29 13:47:32 +0000469// Value of the heuristics parameter used in the current (or last) leak check.
470static UInt detect_memory_leaks_last_heuristics;
sewardjc8bd1df2011-06-26 12:41:33 +0000471
472// DeltaMode used the last time we called detect_memory_leaks.
philippeab1fce92013-09-29 13:47:32 +0000473// The recorded leak errors are output using a logic based on this delta_mode.
sewardjc8bd1df2011-06-26 12:41:33 +0000474// The below avoids replicating the delta_mode in each LossRecord.
475LeakCheckDeltaMode MC_(detect_memory_leaks_last_delta_mode);
476
philippe4e32d672013-10-17 22:10:41 +0000477// Each leak search run increments the below generation counter.
478// A used suppression during a leak search will contain this
479// generation number.
480UInt MC_(leak_search_gen);
njn8225cc02009-03-09 22:52:24 +0000481
njn8225cc02009-03-09 22:52:24 +0000482// Records chunks that are currently being processed. Each element in the
483// stack is an index into lc_chunks and lc_extras. Its size is
484// 'lc_n_chunks' because in the worst case that's how many chunks could be
485// pushed onto it (actually I think the maximum is lc_n_chunks-1 but let's
486// be conservative).
487static Int* lc_markstack;
488// The index of the top element of the stack; -1 if the stack is empty, 0 if
489// the stack has one element, 1 if it has two, etc.
490static Int lc_markstack_top;
491
492// Keeps track of how many bytes of memory we've scanned, for printing.
493// (Nb: We don't keep track of how many register bytes we've scanned.)
494static SizeT lc_scanned_szB;
philippe7a76f4b2013-10-06 21:23:04 +0000495// Keeps track of how many bytes we have not scanned due to read errors that
496// caused a signal such as SIGSEGV.
497static SizeT lc_sig_skipped_szB;
njn8225cc02009-03-09 22:52:24 +0000498
499
500SizeT MC_(bytes_leaked) = 0;
501SizeT MC_(bytes_indirect) = 0;
502SizeT MC_(bytes_dubious) = 0;
503SizeT MC_(bytes_reachable) = 0;
504SizeT MC_(bytes_suppressed) = 0;
505
506SizeT MC_(blocks_leaked) = 0;
507SizeT MC_(blocks_indirect) = 0;
508SizeT MC_(blocks_dubious) = 0;
509SizeT MC_(blocks_reachable) = 0;
510SizeT MC_(blocks_suppressed) = 0;
511
philippeab1fce92013-09-29 13:47:32 +0000512// Subset of MC_(bytes_reachable) and MC_(blocks_reachable) which
513// are considered reachable due to the corresponding heuristic.
514static SizeT MC_(bytes_heuristically_reachable)[N_LEAK_CHECK_HEURISTICS]
515 = {0,0,0,0};
516static SizeT MC_(blocks_heuristically_reachable)[N_LEAK_CHECK_HEURISTICS]
517 = {0,0,0,0};
518
njn8225cc02009-03-09 22:52:24 +0000519// Determines if a pointer is to a chunk. Returns the chunk number et al
520// via call-by-reference.
521static Bool
522lc_is_a_chunk_ptr(Addr ptr, Int* pch_no, MC_Chunk** pch, LC_Extra** pex)
njn43c799e2003-04-08 00:08:52 +0000523{
njn8225cc02009-03-09 22:52:24 +0000524 Int ch_no;
525 MC_Chunk* ch;
526 LC_Extra* ex;
njn43c799e2003-04-08 00:08:52 +0000527
philippe57a16a22012-07-18 22:26:51 +0000528 // Quick filter. Note: implemented with am, not with get_vabits2
529 // as ptr might be random data pointing anywhere. On 64 bit
530 // platforms, getting va bits for random data can be quite costly
531 // due to the secondary map.
njn8225cc02009-03-09 22:52:24 +0000532 if (!VG_(am_is_valid_for_client)(ptr, 1, VKI_PROT_READ)) {
533 return False;
sewardjb5f6f512005-03-10 23:59:00 +0000534 } else {
njn8225cc02009-03-09 22:52:24 +0000535 ch_no = find_chunk_for(ptr, lc_chunks, lc_n_chunks);
536 tl_assert(ch_no >= -1 && ch_no < lc_n_chunks);
537
538 if (ch_no == -1) {
539 return False;
540 } else {
541 // Ok, we've found a pointer to a chunk. Get the MC_Chunk and its
542 // LC_Extra.
543 ch = lc_chunks[ch_no];
544 ex = &(lc_extras[ch_no]);
545
546 tl_assert(ptr >= ch->data);
547 tl_assert(ptr < ch->data + ch->szB + (ch->szB==0 ? 1 : 0));
548
549 if (VG_DEBUG_LEAKCHECK)
550 VG_(printf)("ptr=%#lx -> block %d\n", ptr, ch_no);
551
552 *pch_no = ch_no;
553 *pch = ch;
554 *pex = ex;
555
556 return True;
557 }
sewardjb5f6f512005-03-10 23:59:00 +0000558 }
559}
560
njn8225cc02009-03-09 22:52:24 +0000561// Push a chunk (well, just its index) onto the mark stack.
562static void lc_push(Int ch_no, MC_Chunk* ch)
sewardjb5f6f512005-03-10 23:59:00 +0000563{
tom1d0f3f62010-10-04 20:55:21 +0000564 if (!lc_extras[ch_no].pending) {
565 if (0) {
566 VG_(printf)("pushing %#lx-%#lx\n", ch->data, ch->data + ch->szB);
567 }
568 lc_markstack_top++;
569 tl_assert(lc_markstack_top < lc_n_chunks);
570 lc_markstack[lc_markstack_top] = ch_no;
571 tl_assert(!lc_extras[ch_no].pending);
572 lc_extras[ch_no].pending = True;
njn8225cc02009-03-09 22:52:24 +0000573 }
sewardjb5f6f512005-03-10 23:59:00 +0000574}
575
njn8225cc02009-03-09 22:52:24 +0000576// Return the index of the chunk on the top of the mark stack, or -1 if
577// there isn't one.
578static Bool lc_pop(Int* ret)
sewardjb5f6f512005-03-10 23:59:00 +0000579{
njn8225cc02009-03-09 22:52:24 +0000580 if (-1 == lc_markstack_top) {
581 return False;
582 } else {
583 tl_assert(0 <= lc_markstack_top && lc_markstack_top < lc_n_chunks);
584 *ret = lc_markstack[lc_markstack_top];
585 lc_markstack_top--;
tom1d0f3f62010-10-04 20:55:21 +0000586 tl_assert(lc_extras[*ret].pending);
587 lc_extras[*ret].pending = False;
njn8225cc02009-03-09 22:52:24 +0000588 return True;
589 }
590}
sewardjb5f6f512005-03-10 23:59:00 +0000591
philippeab1fce92013-09-29 13:47:32 +0000592static const HChar* pp_heuristic(LeakCheckHeuristic h)
593{
594 switch(h) {
595 case LchNone: return "none";
596 case LchStdString: return "stdstring";
philippe7c69a3e2014-07-21 19:55:11 +0000597 case LchLength64: return "length64";
philippeab1fce92013-09-29 13:47:32 +0000598 case LchNewArray: return "newarray";
599 case LchMultipleInheritance: return "multipleinheritance";
600 default: return "???invalid heuristic???";
601 }
602}
603
604// True if ptr looks like the address of a vtable, i.e. if ptr
605// points to an array of pointers to functions.
606// It is assumed the only caller of this function is heuristic_reachedness
607// which must check that ptr is aligned and above page 0.
608// Checking that ptr is above page 0 is an optimisation : it is assumed
609// that no vtable is located in the page 0. So, all small integer values
610// encountered during the scan will not incur the cost of calling this
611// function.
612static Bool aligned_ptr_above_page0_is_vtable_addr(Addr ptr)
613{
614 // ??? If performance problem:
615 // ??? maybe implement a cache (array indexed by ptr % primenr)
616 // ??? of "I am a vtable ptr" ???
617
618 // ??? Maybe the debug info could (efficiently?) be used to detect vtables ?
619
620 // We consider ptr as a vtable ptr if it points to a table
621 // where we find only NULL pointers or pointers pointing at an
622 // executable region. We must find at least 2 non NULL pointers
623 // before considering ptr as a vtable pointer.
624 // We scan a maximum of VTABLE_MAX_CHECK words for these 2 non NULL
625 // pointers.
626#define VTABLE_MAX_CHECK 20
627
628 NSegment const *seg;
629 UInt nr_fn_ptrs = 0;
630 Addr scan;
631 Addr scan_max;
632
633 // First verify ptr points inside a client mapped file section.
634 // ??? is a vtable always in a file mapped readable section ?
635 seg = VG_(am_find_nsegment) (ptr);
636 if (seg == NULL
637 || seg->kind != SkFileC
638 || !seg->hasR)
639 return False;
640
641 // Check potential function pointers, up to a maximum of VTABLE_MAX_CHECK.
642 scan_max = ptr + VTABLE_MAX_CHECK*sizeof(Addr);
643 // If ptr is near the end of seg, avoid scan_max exceeding the end of seg:
644 if (scan_max > seg->end - sizeof(Addr))
645 scan_max = seg->end - sizeof(Addr);
646 for (scan = ptr; scan <= scan_max; scan+=sizeof(Addr)) {
647 Addr pot_fn = *((Addr *)scan);
648 if (pot_fn == 0)
649 continue; // NULL fn pointer. Seems it can happen in vtable.
650 seg = VG_(am_find_nsegment) (pot_fn);
philippe05334052014-08-28 21:02:11 +0000651#if defined(VGA_ppc64be) || defined(VGA_ppc64le)
philippeab1fce92013-09-29 13:47:32 +0000652 // ppc64 use a thunk table. So, we have one more level of indirection
653 // to follow.
654 if (seg == NULL
655 || seg->kind != SkFileC
656 || !seg->hasR
657 || !seg->hasW)
658 return False; // ptr to nowhere, or not a ptr to thunks.
659 pot_fn = *((Addr *)pot_fn);
660 if (pot_fn == 0)
661 continue; // NULL fn pointer. Seems it can happen in vtable.
662 seg = VG_(am_find_nsegment) (pot_fn);
663#endif
664 if (seg == NULL
665 || seg->kind != SkFileC
666 || !seg->hasT)
667 return False; // ptr to nowhere, or not a fn ptr.
668 nr_fn_ptrs++;
669 if (nr_fn_ptrs == 2)
670 return True;
671 }
672
673 return False;
674}
675
philippe7c69a3e2014-07-21 19:55:11 +0000676// true if a is properly aligned and points to 64bits of valid memory
677static Bool is_valid_aligned_ULong ( Addr a )
678{
679 if (sizeof(Word) == 8)
680 return MC_(is_valid_aligned_word)(a);
681
682 return MC_(is_valid_aligned_word)(a)
683 && MC_(is_valid_aligned_word)(a + 4);
684}
685
philippeab1fce92013-09-29 13:47:32 +0000686// If ch is heuristically reachable via an heuristic member of heur_set,
687// returns this heuristic.
688// If ch cannot be considered reachable using one of these heuristics,
689// return LchNone.
690// This should only be called when ptr is an interior ptr to ch.
691// The StdString/NewArray/MultipleInheritance heuristics are directly
692// inspired from DrMemory:
693// see http://www.burningcutlery.com/derek/docs/drmem-CGO11.pdf [section VI,C]
694// and bug 280271.
695static LeakCheckHeuristic heuristic_reachedness (Addr ptr,
696 MC_Chunk *ch, LC_Extra *ex,
697 UInt heur_set)
698{
699 if (HiS(LchStdString, heur_set)) {
700 // Detects inner pointers to Std::String for layout being
701 // length capacity refcount char_array[] \0
702 // where ptr points to the beginning of the char_array.
philippe078ab862013-10-13 18:38:30 +0000703 // Note: we check definedness for length and capacity but
704 // not for refcount, as refcount size might be smaller than
705 // a SizeT, giving a uninitialised hole in the first 3 SizeT.
706 if ( ptr == ch->data + 3 * sizeof(SizeT)
707 && MC_(is_valid_aligned_word)(ch->data + sizeof(SizeT))) {
708 const SizeT capacity = *((SizeT*)(ch->data + sizeof(SizeT)));
709 if (3 * sizeof(SizeT) + capacity + 1 == ch->szB
710 && MC_(is_valid_aligned_word)(ch->data)) {
711 const SizeT length = *((SizeT*)ch->data);
712 if (length <= capacity) {
713 // ??? could check there is no null byte from ptr to ptr+length-1
714 // ??? and that there is a null byte at ptr+length.
715 // ???
716 // ??? could check that ch->allockind is MC_AllocNew ???
717 // ??? probably not a good idea, as I guess stdstring
718 // ??? allocator can be done via custom allocator
719 // ??? or even a call to malloc ????
720 return LchStdString;
721 }
philippeab1fce92013-09-29 13:47:32 +0000722 }
723 }
724 }
725
philippe7c69a3e2014-07-21 19:55:11 +0000726 if (HiS(LchLength64, heur_set)) {
727 // Detects inner pointers that point at 64bit offset (8 bytes) into a
728 // block following the length of the remaining as 64bit number
729 // (=total block size - 8).
730 // This is used e.g. by sqlite for tracking the total size of allocated
731 // memory.
732 // Note that on 64bit platforms, a block matching LchLength64 will
733 // also be matched by LchNewArray.
734 if ( ptr == ch->data + sizeof(ULong)
735 && is_valid_aligned_ULong(ch->data)) {
736 const ULong size = *((ULong*)ch->data);
737 if (size > 0 && (ch->szB - sizeof(ULong)) == size) {
738 return LchLength64;
739 }
740 }
741 }
742
philippeab1fce92013-09-29 13:47:32 +0000743 if (HiS(LchNewArray, heur_set)) {
744 // Detects inner pointers at second word of new[] array, following
745 // a plausible nr of elements.
746 // Such inner pointers are used for arrays of elements
747 // having a destructor, as the delete[] of the array must know
748 // how many elements to destroy.
749 //
750 // We have a strange/wrong case for 'ptr = new MyClass[0];' :
751 // For such a case, the returned ptr points just outside the
752 // allocated chunk. This chunk is then seen as a definite
753 // leak by Valgrind, as it is not considered an interior pointer.
754 // It is the c++ equivalent of bug 99923 (malloc(0) wrongly considered
755 // as definitely leaked). See the trick in find_chunk_for handling
756 // 0-sized block. This trick does not work for 'new MyClass[0]'
757 // because a chunk "word-sized" is allocated to store the (0) nr
758 // of elements.
philippe078ab862013-10-13 18:38:30 +0000759 if ( ptr == ch->data + sizeof(SizeT)
760 && MC_(is_valid_aligned_word)(ch->data)) {
philippeab1fce92013-09-29 13:47:32 +0000761 const SizeT nr_elts = *((SizeT*)ch->data);
762 if (nr_elts > 0 && (ch->szB - sizeof(SizeT)) % nr_elts == 0) {
763 // ??? could check that ch->allockind is MC_AllocNewVec ???
764 return LchNewArray;
765 }
766 }
767 }
768
769 if (HiS(LchMultipleInheritance, heur_set)) {
770 // Detect inner pointer used for multiple inheritance.
771 // Assumption is that the vtable pointers are before the object.
philippe078ab862013-10-13 18:38:30 +0000772 if (VG_IS_WORD_ALIGNED(ptr)
773 && MC_(is_valid_aligned_word)(ptr)) {
philippeab1fce92013-09-29 13:47:32 +0000774 Addr first_addr;
775 Addr inner_addr;
776
777 // Avoid the call to is_vtable_addr when the addr is not
778 // aligned or points in the page0, as it is unlikely
779 // a vtable is located in this page. This last optimisation
780 // avoids to call aligned_ptr_above_page0_is_vtable_addr
781 // for all small integers.
782 // Note: we could possibly also avoid calling this function
783 // for small negative integers, as no vtable should be located
784 // in the last page.
785 inner_addr = *((Addr*)ptr);
786 if (VG_IS_WORD_ALIGNED(inner_addr)
philippe078ab862013-10-13 18:38:30 +0000787 && inner_addr >= (Addr)VKI_PAGE_SIZE
788 && MC_(is_valid_aligned_word)(ch->data)) {
philippeab1fce92013-09-29 13:47:32 +0000789 first_addr = *((Addr*)ch->data);
790 if (VG_IS_WORD_ALIGNED(first_addr)
791 && first_addr >= (Addr)VKI_PAGE_SIZE
792 && aligned_ptr_above_page0_is_vtable_addr(inner_addr)
793 && aligned_ptr_above_page0_is_vtable_addr(first_addr)) {
794 // ??? could check that ch->allockind is MC_AllocNew ???
795 return LchMultipleInheritance;
796 }
797 }
798 }
799 }
800
801 return LchNone;
802}
803
njn8225cc02009-03-09 22:52:24 +0000804
805// If 'ptr' is pointing to a heap-allocated block which hasn't been seen
806// before, push it onto the mark stack.
807static void
808lc_push_without_clique_if_a_chunk_ptr(Addr ptr, Bool is_prior_definite)
809{
810 Int ch_no;
811 MC_Chunk* ch;
812 LC_Extra* ex;
philippeab1fce92013-09-29 13:47:32 +0000813 Reachedness ch_via_ptr; // Is ch reachable via ptr, and how ?
njn8225cc02009-03-09 22:52:24 +0000814
815 if ( ! lc_is_a_chunk_ptr(ptr, &ch_no, &ch, &ex) )
816 return;
philippeab1fce92013-09-29 13:47:32 +0000817
818 if (ex->state == Reachable) {
philippe078ab862013-10-13 18:38:30 +0000819 if (ex->heuristic && ptr == ch->data)
820 // If block was considered reachable via an heuristic, and it is now
821 // directly reachable via ptr, clear the heuristic field.
philippeab1fce92013-09-29 13:47:32 +0000822 ex->heuristic = LchNone;
philippeab1fce92013-09-29 13:47:32 +0000823 return;
824 }
tom1d0f3f62010-10-04 20:55:21 +0000825
njn8225cc02009-03-09 22:52:24 +0000826 // Possibly upgrade the state, ie. one of:
827 // - Unreached --> Possible
828 // - Unreached --> Reachable
829 // - Possible --> Reachable
philippeab1fce92013-09-29 13:47:32 +0000830
831 if (ptr == ch->data)
832 ch_via_ptr = Reachable;
833 else if (detect_memory_leaks_last_heuristics) {
834 ex->heuristic
835 = heuristic_reachedness (ptr, ch, ex,
836 detect_memory_leaks_last_heuristics);
837 if (ex->heuristic)
838 ch_via_ptr = Reachable;
839 else
840 ch_via_ptr = Possible;
841 } else
842 ch_via_ptr = Possible;
843
844 if (ch_via_ptr == Reachable && is_prior_definite) {
845 // 'ptr' points to the start of the block or is to be considered as
846 // pointing to the start of the block, and the prior node is
njn8225cc02009-03-09 22:52:24 +0000847 // definite, which means that this block is definitely reachable.
848 ex->state = Reachable;
849
tom1d0f3f62010-10-04 20:55:21 +0000850 // State has changed to Reachable so (re)scan the block to make
851 // sure any blocks it points to are correctly marked.
852 lc_push(ch_no, ch);
853
njn8225cc02009-03-09 22:52:24 +0000854 } else if (ex->state == Unreached) {
855 // Either 'ptr' is a interior-pointer, or the prior node isn't definite,
856 // which means that we can only mark this block as possibly reachable.
857 ex->state = Possible;
tom1d0f3f62010-10-04 20:55:21 +0000858
859 // State has changed to Possible so (re)scan the block to make
860 // sure any blocks it points to are correctly marked.
861 lc_push(ch_no, ch);
njn8225cc02009-03-09 22:52:24 +0000862 }
863}
864
865static void
florian6bd9dc12012-11-23 16:17:43 +0000866lc_push_if_a_chunk_ptr_register(ThreadId tid, const HChar* regname, Addr ptr)
njn8225cc02009-03-09 22:52:24 +0000867{
868 lc_push_without_clique_if_a_chunk_ptr(ptr, /*is_prior_definite*/True);
869}
870
871// If ptr is pointing to a heap-allocated block which hasn't been seen
872// before, push it onto the mark stack. Clique is the index of the
873// clique leader.
874static void
philippea22f59d2012-01-26 23:13:52 +0000875lc_push_with_clique_if_a_chunk_ptr(Addr ptr, Int clique, Int cur_clique)
njn8225cc02009-03-09 22:52:24 +0000876{
877 Int ch_no;
878 MC_Chunk* ch;
879 LC_Extra* ex;
880
881 tl_assert(0 <= clique && clique < lc_n_chunks);
882
883 if ( ! lc_is_a_chunk_ptr(ptr, &ch_no, &ch, &ex) )
884 return;
885
886 // If it's not Unreached, it's already been handled so ignore it.
887 // If ch_no==clique, it's the clique leader, which means this is a cyclic
888 // structure; again ignore it because it's already been handled.
889 if (ex->state == Unreached && ch_no != clique) {
890 // Note that, unlike reachable blocks, we currently don't distinguish
891 // between start-pointers and interior-pointers here. We probably
892 // should, though.
njn8225cc02009-03-09 22:52:24 +0000893 lc_push(ch_no, ch);
894
895 // Add the block to the clique, and add its size to the
896 // clique-leader's indirect size. Also, if the new block was
897 // itself a clique leader, it isn't any more, so add its
898 // indirect_szB to the new clique leader.
899 if (VG_DEBUG_CLIQUE) {
philippea22f59d2012-01-26 23:13:52 +0000900 if (ex->IorC.indirect_szB > 0)
njn8225cc02009-03-09 22:52:24 +0000901 VG_(printf)(" clique %d joining clique %d adding %lu+%lu\n",
bartdc429d12011-07-29 14:24:07 +0000902 ch_no, clique, (unsigned long)ch->szB,
philippea22f59d2012-01-26 23:13:52 +0000903 (unsigned long)ex->IorC.indirect_szB);
njn8225cc02009-03-09 22:52:24 +0000904 else
905 VG_(printf)(" block %d joining clique %d adding %lu\n",
bartdc429d12011-07-29 14:24:07 +0000906 ch_no, clique, (unsigned long)ch->szB);
njn8225cc02009-03-09 22:52:24 +0000907 }
908
philippea22f59d2012-01-26 23:13:52 +0000909 lc_extras[clique].IorC.indirect_szB += ch->szB;
910 lc_extras[clique].IorC.indirect_szB += ex->IorC.indirect_szB;
911 ex->state = IndirectLeak;
912 ex->IorC.clique = (SizeT) cur_clique;
njn8225cc02009-03-09 22:52:24 +0000913 }
914}
915
916static void
philippeab1fce92013-09-29 13:47:32 +0000917lc_push_if_a_chunk_ptr(Addr ptr,
918 Int clique, Int cur_clique, Bool is_prior_definite)
njn8225cc02009-03-09 22:52:24 +0000919{
920 if (-1 == clique)
921 lc_push_without_clique_if_a_chunk_ptr(ptr, is_prior_definite);
922 else
philippea22f59d2012-01-26 23:13:52 +0000923 lc_push_with_clique_if_a_chunk_ptr(ptr, clique, cur_clique);
sewardjb5f6f512005-03-10 23:59:00 +0000924}
925
sewardj45d94cc2005-04-20 14:44:11 +0000926
sewardj97d3ebb2011-04-11 18:36:34 +0000927static VG_MINIMAL_JMP_BUF(memscan_jmpbuf);
philippe7a76f4b2013-10-06 21:23:04 +0000928static volatile Addr bad_scanned_addr;
sewardjb5f6f512005-03-10 23:59:00 +0000929
njn8225cc02009-03-09 22:52:24 +0000930static
931void scan_all_valid_memory_catcher ( Int sigNo, Addr addr )
sewardjb5f6f512005-03-10 23:59:00 +0000932{
njn8225cc02009-03-09 22:52:24 +0000933 if (0)
934 VG_(printf)("OUCH! sig=%d addr=%#lx\n", sigNo, addr);
philippe7a76f4b2013-10-06 21:23:04 +0000935 if (sigNo == VKI_SIGSEGV || sigNo == VKI_SIGBUS) {
936 bad_scanned_addr = addr;
sewardj6c591e12011-04-11 16:17:51 +0000937 VG_MINIMAL_LONGJMP(memscan_jmpbuf);
philippe7a76f4b2013-10-06 21:23:04 +0000938 }
njn8225cc02009-03-09 22:52:24 +0000939}
940
philippea22f59d2012-01-26 23:13:52 +0000941// lc_scan_memory has 2 modes:
942//
943// 1. Leak check mode (searched == 0).
944// -----------------------------------
njn8225cc02009-03-09 22:52:24 +0000945// Scan a block of memory between [start, start+len). This range may
946// be bogus, inaccessable, or otherwise strange; we deal with it. For each
947// valid aligned word we assume it's a pointer to a chunk a push the chunk
948// onto the mark stack if so.
philippea22f59d2012-01-26 23:13:52 +0000949// clique is the "highest level clique" in which indirectly leaked blocks have
950// to be collected. cur_clique is the current "lower" level clique through which
951// the memory to be scanned has been found.
952// Example: in the below tree if A is leaked, the top level clique will
953// be A, while lower level cliques will be B and C.
954/*
955 A
florianf5300ff2014-12-28 16:46:14 +0000956 / \
philippea22f59d2012-01-26 23:13:52 +0000957 B C
florianf5300ff2014-12-28 16:46:14 +0000958 / \ / \
philippea22f59d2012-01-26 23:13:52 +0000959 D E F G
960*/
961// Proper handling of top and lowest level clique allows block_list of a loss
962// record to describe the hierarchy of indirectly leaked blocks.
963//
964// 2. Search ptr mode (searched != 0).
965// -----------------------------------
966// In this mode, searches for pointers to a specific address range
philippeab1fce92013-09-29 13:47:32 +0000967// In such a case, lc_scan_memory just scans [start..start+len[ for pointers
968// to searched and outputs the places where searched is found.
969// It does not recursively scans the found memory.
njn8225cc02009-03-09 22:52:24 +0000970static void
philippeab1fce92013-09-29 13:47:32 +0000971lc_scan_memory(Addr start, SizeT len, Bool is_prior_definite,
972 Int clique, Int cur_clique,
philippea22f59d2012-01-26 23:13:52 +0000973 Addr searched, SizeT szB)
njn8225cc02009-03-09 22:52:24 +0000974{
philippe57a16a22012-07-18 22:26:51 +0000975 /* memory scan is based on the assumption that valid pointers are aligned
976 on a multiple of sizeof(Addr). So, we can (and must) skip the begin and
977 end portions of the block if they are not aligned on sizeof(Addr):
978 These cannot be a valid pointer, and calls to MC_(is_valid_aligned_word)
979 will assert for a non aligned address. */
philippe110c77e2013-10-15 21:04:56 +0000980#if defined(VGA_s390x)
981 // Define ptr as volatile, as on this platform, the value of ptr
982 // is read in code executed via a longjmp.
983 volatile
984#endif
philippe7a76f4b2013-10-06 21:23:04 +0000985 Addr ptr = VG_ROUNDUP(start, sizeof(Addr));
986 const Addr end = VG_ROUNDDN(start+len, sizeof(Addr));
sewardjb5f6f512005-03-10 23:59:00 +0000987 vki_sigset_t sigmask;
988
989 if (VG_DEBUG_LEAKCHECK)
njn8225cc02009-03-09 22:52:24 +0000990 VG_(printf)("scan %#lx-%#lx (%lu)\n", start, end, len);
991
sewardjb5f6f512005-03-10 23:59:00 +0000992 VG_(sigprocmask)(VKI_SIG_SETMASK, NULL, &sigmask);
njn695c16e2005-03-27 03:40:28 +0000993 VG_(set_fault_catcher)(scan_all_valid_memory_catcher);
sewardjb5f6f512005-03-10 23:59:00 +0000994
philippe57a16a22012-07-18 22:26:51 +0000995 /* Optimisation: the loop below will check for each begin
996 of SM chunk if the chunk is fully unaddressable. The idea is to
997 skip efficiently such fully unaddressable SM chunks.
998 So, we preferrably start the loop on a chunk boundary.
999 If the chunk is not fully unaddressable, we might be in
1000 an unaddressable page. Again, the idea is to skip efficiently
1001 such unaddressable page : this is the "else" part.
1002 We use an "else" so that two consecutive fully unaddressable
1003 SM chunks will be skipped efficiently: first one is skipped
1004 by this piece of code. The next SM chunk will be skipped inside
1005 the loop. */
1006 if ( ! MC_(is_within_valid_secondary)(ptr) ) {
1007 // Skip an invalid SM chunk till the beginning of the next SM Chunk.
1008 ptr = VG_ROUNDUP(ptr+1, SM_SIZE);
1009 } else if (!VG_(am_is_valid_for_client)(ptr, sizeof(Addr), VKI_PROT_READ)) {
1010 // else we are in a (at least partially) valid SM chunk.
1011 // We might be in the middle of an unreadable page.
1012 // Do a cheap check to see if it's valid;
1013 // if not, skip onto the next page.
njn8225cc02009-03-09 22:52:24 +00001014 ptr = VG_PGROUNDUP(ptr+1); // First page is bad - skip it.
philippe57a16a22012-07-18 22:26:51 +00001015 }
philippe7a76f4b2013-10-06 21:23:04 +00001016 /* The above optimisation and below loop is based on some relationships
1017 between VKI_PAGE_SIZE, SM_SIZE and sizeof(Addr) which are asserted in
philippe57a16a22012-07-18 22:26:51 +00001018 MC_(detect_memory_leaks). */
sewardjb5f6f512005-03-10 23:59:00 +00001019
philippe7a76f4b2013-10-06 21:23:04 +00001020 // During scan, we check with aspacemgr that each page is readable and
1021 // belongs to client.
1022 // We still protect against SIGSEGV and SIGBUS e.g. in case aspacemgr is
1023 // desynchronised with the real page mappings.
1024 // Such a desynchronisation could happen due to an aspacemgr bug.
1025 // Note that if the application is using mprotect(NONE), then
1026 // a page can be unreadable but have addressable and defined
1027 // VA bits (see mc_main.c function mc_new_mem_mprotect).
1028 if (VG_MINIMAL_SETJMP(memscan_jmpbuf) != 0) {
1029 // Catch read error ...
1030 // We need to restore the signal mask, because we were
1031 // longjmped out of a signal handler.
1032 VG_(sigprocmask)(VKI_SIG_SETMASK, &sigmask, NULL);
philippe110c77e2013-10-15 21:04:56 +00001033# if defined(VGA_s390x)
1034 // For a SIGSEGV, s390 delivers the page address of the bad address.
1035 // For a SIGBUS, old s390 kernels deliver a NULL address.
1036 // bad_scanned_addr can thus not be used.
1037 // So, on this platform, we always skip a full page from ptr.
1038 // The below implies to mark ptr as volatile, as we read the value
1039 // after a longjmp to here.
1040 lc_sig_skipped_szB += VKI_PAGE_SIZE;
1041 ptr = ptr + VKI_PAGE_SIZE; // Unaddressable, - skip it.
1042# else
1043 // On other platforms, just skip one Addr.
philippe7a76f4b2013-10-06 21:23:04 +00001044 lc_sig_skipped_szB += sizeof(Addr);
1045 tl_assert(bad_scanned_addr >= VG_ROUNDUP(start, sizeof(Addr)));
1046 tl_assert(bad_scanned_addr < VG_ROUNDDN(start+len, sizeof(Addr)));
1047 ptr = bad_scanned_addr + sizeof(Addr); // Unaddressable, - skip it.
philippe110c77e2013-10-15 21:04:56 +00001048#endif
philippe7a76f4b2013-10-06 21:23:04 +00001049 }
sewardj05fe85e2005-04-27 22:46:36 +00001050 while (ptr < end) {
sewardjb5f6f512005-03-10 23:59:00 +00001051 Addr addr;
1052
njn8225cc02009-03-09 22:52:24 +00001053 // Skip invalid chunks.
philippe57a16a22012-07-18 22:26:51 +00001054 if (UNLIKELY((ptr % SM_SIZE) == 0)) {
1055 if (! MC_(is_within_valid_secondary)(ptr) ) {
1056 ptr = VG_ROUNDUP(ptr+1, SM_SIZE);
1057 continue;
1058 }
sewardjb5f6f512005-03-10 23:59:00 +00001059 }
1060
njn8225cc02009-03-09 22:52:24 +00001061 // Look to see if this page seems reasonable.
philippe57a16a22012-07-18 22:26:51 +00001062 if (UNLIKELY((ptr % VKI_PAGE_SIZE) == 0)) {
njn8225cc02009-03-09 22:52:24 +00001063 if (!VG_(am_is_valid_for_client)(ptr, sizeof(Addr), VKI_PROT_READ)) {
1064 ptr += VKI_PAGE_SIZE; // Bad page - skip it.
1065 continue;
1066 }
sewardjb5f6f512005-03-10 23:59:00 +00001067 }
1068
philippe57a16a22012-07-18 22:26:51 +00001069 if ( MC_(is_valid_aligned_word)(ptr) ) {
1070 lc_scanned_szB += sizeof(Addr);
philippe7a76f4b2013-10-06 21:23:04 +00001071 // If the below read fails, we will longjmp to the loop begin.
philippe57a16a22012-07-18 22:26:51 +00001072 addr = *(Addr *)ptr;
1073 // If we get here, the scanned word is in valid memory. Now
1074 // let's see if its contents point to a chunk.
1075 if (UNLIKELY(searched)) {
1076 if (addr >= searched && addr < searched + szB) {
philippeab1fce92013-09-29 13:47:32 +00001077 if (addr == searched) {
philippe57a16a22012-07-18 22:26:51 +00001078 VG_(umsg)("*%#lx points at %#lx\n", ptr, searched);
philippeab1fce92013-09-29 13:47:32 +00001079 MC_(pp_describe_addr) (ptr);
1080 } else {
1081 Int ch_no;
1082 MC_Chunk *ch;
1083 LC_Extra *ex;
philippe57a16a22012-07-18 22:26:51 +00001084 VG_(umsg)("*%#lx interior points at %lu bytes inside %#lx\n",
1085 ptr, (long unsigned) addr - searched, searched);
philippeab1fce92013-09-29 13:47:32 +00001086 MC_(pp_describe_addr) (ptr);
1087 if (lc_is_a_chunk_ptr(addr, &ch_no, &ch, &ex) ) {
1088 Int h;
philippe7c69a3e2014-07-21 19:55:11 +00001089 for (h = LchStdString; h < N_LEAK_CHECK_HEURISTICS; h++) {
philippeab1fce92013-09-29 13:47:32 +00001090 if (heuristic_reachedness(addr, ch, ex, H2S(h)) == h) {
1091 VG_(umsg)("block at %#lx considered reachable "
1092 "by ptr %#lx using %s heuristic\n",
1093 ch->data, addr, pp_heuristic(h));
1094 }
1095 }
philippe7a76f4b2013-10-06 21:23:04 +00001096 // Verify the loop above has properly scanned all
1097 // heuristics. If the below fails, it probably means the
1098 // LeakCheckHeuristic enum is not in sync anymore with the
1099 // above loop and/or with N_LEAK_CHECK_HEURISTICS.
philippe5bd40602013-10-02 20:59:05 +00001100 tl_assert (h == N_LEAK_CHECK_HEURISTICS);
philippeab1fce92013-09-29 13:47:32 +00001101 }
1102 }
philippea22f59d2012-01-26 23:13:52 +00001103 }
philippe57a16a22012-07-18 22:26:51 +00001104 } else {
1105 lc_push_if_a_chunk_ptr(addr, clique, cur_clique, is_prior_definite);
njn8225cc02009-03-09 22:52:24 +00001106 }
philippe57a16a22012-07-18 22:26:51 +00001107 } else if (0 && VG_DEBUG_LEAKCHECK) {
1108 VG_(printf)("%#lx not valid\n", ptr);
sewardjb5f6f512005-03-10 23:59:00 +00001109 }
philippe57a16a22012-07-18 22:26:51 +00001110 ptr += sizeof(Addr);
sewardjb5f6f512005-03-10 23:59:00 +00001111 }
1112
1113 VG_(sigprocmask)(VKI_SIG_SETMASK, &sigmask, NULL);
1114 VG_(set_fault_catcher)(NULL);
1115}
1116
sewardj45d94cc2005-04-20 14:44:11 +00001117
njn8225cc02009-03-09 22:52:24 +00001118// Process the mark stack until empty.
1119static void lc_process_markstack(Int clique)
sewardjb5f6f512005-03-10 23:59:00 +00001120{
njne3675d62009-05-19 02:08:25 +00001121 Int top = -1; // shut gcc up
njn8225cc02009-03-09 22:52:24 +00001122 Bool is_prior_definite;
sewardjb5f6f512005-03-10 23:59:00 +00001123
njn8225cc02009-03-09 22:52:24 +00001124 while (lc_pop(&top)) {
tom1d0f3f62010-10-04 20:55:21 +00001125 tl_assert(top >= 0 && top < lc_n_chunks);
sewardjb5f6f512005-03-10 23:59:00 +00001126
njn8225cc02009-03-09 22:52:24 +00001127 // See comment about 'is_prior_definite' at the top to understand this.
1128 is_prior_definite = ( Possible != lc_extras[top].state );
sewardjb5f6f512005-03-10 23:59:00 +00001129
njn8225cc02009-03-09 22:52:24 +00001130 lc_scan_memory(lc_chunks[top]->data, lc_chunks[top]->szB,
philippea22f59d2012-01-26 23:13:52 +00001131 is_prior_definite, clique, (clique == -1 ? -1 : top),
1132 /*searched*/ 0, 0);
sewardjb5f6f512005-03-10 23:59:00 +00001133 }
1134}
1135
njn29a5c012009-05-06 06:15:55 +00001136static Word cmp_LossRecordKey_LossRecord(const void* key, const void* elem)
1137{
florian3e798632012-11-24 19:41:54 +00001138 const LossRecordKey* a = key;
1139 const LossRecordKey* b = &(((const LossRecord*)elem)->key);
njn29a5c012009-05-06 06:15:55 +00001140
1141 // Compare on states first because that's fast.
1142 if (a->state < b->state) return -1;
1143 if (a->state > b->state) return 1;
1144 // Ok, the states are equal. Now compare the locations, which is slower.
1145 if (VG_(eq_ExeContext)(
1146 MC_(clo_leak_resolution), a->allocated_at, b->allocated_at))
1147 return 0;
1148 // Different locations. Ordering is arbitrary, just use the ec pointer.
1149 if (a->allocated_at < b->allocated_at) return -1;
1150 if (a->allocated_at > b->allocated_at) return 1;
1151 VG_(tool_panic)("bad LossRecord comparison");
1152}
1153
florian6bd9dc12012-11-23 16:17:43 +00001154static Int cmp_LossRecords(const void* va, const void* vb)
njn29a5c012009-05-06 06:15:55 +00001155{
florian3e798632012-11-24 19:41:54 +00001156 const LossRecord* lr_a = *(const LossRecord *const *)va;
1157 const LossRecord* lr_b = *(const LossRecord *const *)vb;
njn29a5c012009-05-06 06:15:55 +00001158 SizeT total_szB_a = lr_a->szB + lr_a->indirect_szB;
1159 SizeT total_szB_b = lr_b->szB + lr_b->indirect_szB;
1160
1161 // First compare by sizes.
1162 if (total_szB_a < total_szB_b) return -1;
1163 if (total_szB_a > total_szB_b) return 1;
1164 // If size are equal, compare by states.
1165 if (lr_a->key.state < lr_b->key.state) return -1;
1166 if (lr_a->key.state > lr_b->key.state) return 1;
njne10c7f82009-05-06 06:52:47 +00001167 // If they're still equal here, it doesn't matter that much, but we keep
1168 // comparing other things so that regtests are as deterministic as
1169 // possible. So: compare num_blocks.
1170 if (lr_a->num_blocks < lr_b->num_blocks) return -1;
1171 if (lr_a->num_blocks > lr_b->num_blocks) return 1;
1172 // Finally, compare ExeContext addresses... older ones are likely to have
1173 // lower addresses.
1174 if (lr_a->key.allocated_at < lr_b->key.allocated_at) return -1;
1175 if (lr_a->key.allocated_at > lr_b->key.allocated_at) return 1;
njn29a5c012009-05-06 06:15:55 +00001176 return 0;
1177}
1178
philippea22f59d2012-01-26 23:13:52 +00001179// allocates or reallocates lr_array, and set its elements to the loss records
1180// contains in lr_table.
1181static Int get_lr_array_from_lr_table(void) {
1182 Int i, n_lossrecords;
1183 LossRecord* lr;
1184
1185 n_lossrecords = VG_(OSetGen_Size)(lr_table);
1186
1187 // (re-)create the array of pointers to the loss records.
1188 // lr_array is kept to allow producing the block list from gdbserver.
1189 if (lr_array != NULL)
1190 VG_(free)(lr_array);
1191 lr_array = VG_(malloc)("mc.pr.2", n_lossrecords * sizeof(LossRecord*));
1192 i = 0;
1193 VG_(OSetGen_ResetIter)(lr_table);
1194 while ( (lr = VG_(OSetGen_Next)(lr_table)) ) {
1195 lr_array[i++] = lr;
1196 }
1197 tl_assert(i == n_lossrecords);
1198 return n_lossrecords;
1199}
1200
philippe84234902012-01-14 13:53:13 +00001201
1202static void get_printing_rules(LeakCheckParams* lcp,
1203 LossRecord* lr,
1204 Bool* count_as_error,
1205 Bool* print_record)
sewardjb5f6f512005-03-10 23:59:00 +00001206{
philippe84234902012-01-14 13:53:13 +00001207 // Rules for printing:
1208 // - We don't show suppressed loss records ever (and that's controlled
1209 // within the error manager).
philippe2193a7c2012-12-08 17:54:16 +00001210 // - We show non-suppressed loss records that are specified in
1211 // --show-leak-kinds=... if --leak-check=yes.
philippe84234902012-01-14 13:53:13 +00001212
1213 Bool delta_considered;
1214
1215 switch (lcp->deltamode) {
1216 case LCD_Any:
1217 delta_considered = lr->num_blocks > 0;
1218 break;
1219 case LCD_Increased:
1220 delta_considered
1221 = lr->szB > lr->old_szB
1222 || lr->indirect_szB > lr->old_indirect_szB
1223 || lr->num_blocks > lr->old_num_blocks;
1224 break;
1225 case LCD_Changed:
1226 delta_considered = lr->szB != lr->old_szB
1227 || lr->indirect_szB != lr->old_indirect_szB
1228 || lr->num_blocks != lr->old_num_blocks;
1229 break;
1230 default:
1231 tl_assert(0);
1232 }
1233
philippe2193a7c2012-12-08 17:54:16 +00001234 *print_record = lcp->mode == LC_Full && delta_considered
1235 && RiS(lr->key.state,lcp->show_leak_kinds);
philippe84234902012-01-14 13:53:13 +00001236 // We don't count a leaks as errors with lcp->mode==LC_Summary.
1237 // Otherwise you can get high error counts with few or no error
philippe2193a7c2012-12-08 17:54:16 +00001238 // messages, which can be confusing. Otherwise, we count as errors
1239 // the leak kinds requested by --errors-for-leak-kinds=...
1240 *count_as_error = lcp->mode == LC_Full && delta_considered
1241 && RiS(lr->key.state,lcp->errors_for_leak_kinds);
philippe84234902012-01-14 13:53:13 +00001242}
1243
1244static void print_results(ThreadId tid, LeakCheckParams* lcp)
1245{
1246 Int i, n_lossrecords, start_lr_output_scan;
njn29a5c012009-05-06 06:15:55 +00001247 LossRecord* lr;
1248 Bool is_suppressed;
philippeab1fce92013-09-29 13:47:32 +00001249 /* old_* variables are used to report delta in summary. */
1250 SizeT old_bytes_leaked = MC_(bytes_leaked);
sewardjc8bd1df2011-06-26 12:41:33 +00001251 SizeT old_bytes_indirect = MC_(bytes_indirect);
1252 SizeT old_bytes_dubious = MC_(bytes_dubious);
1253 SizeT old_bytes_reachable = MC_(bytes_reachable);
1254 SizeT old_bytes_suppressed = MC_(bytes_suppressed);
1255 SizeT old_blocks_leaked = MC_(blocks_leaked);
1256 SizeT old_blocks_indirect = MC_(blocks_indirect);
1257 SizeT old_blocks_dubious = MC_(blocks_dubious);
1258 SizeT old_blocks_reachable = MC_(blocks_reachable);
1259 SizeT old_blocks_suppressed = MC_(blocks_suppressed);
sewardjb5f6f512005-03-10 23:59:00 +00001260
philippeab1fce92013-09-29 13:47:32 +00001261 SizeT old_bytes_heuristically_reachable[N_LEAK_CHECK_HEURISTICS];
1262 SizeT old_blocks_heuristically_reachable[N_LEAK_CHECK_HEURISTICS];
1263
1264 for (i = 0; i < N_LEAK_CHECK_HEURISTICS; i++) {
1265 old_bytes_heuristically_reachable[i]
1266 = MC_(bytes_heuristically_reachable)[i];
1267 MC_(bytes_heuristically_reachable)[i] = 0;
1268 old_blocks_heuristically_reachable[i]
1269 = MC_(blocks_heuristically_reachable)[i];
1270 MC_(blocks_heuristically_reachable)[i] = 0;
1271 }
1272
sewardjc8bd1df2011-06-26 12:41:33 +00001273 if (lr_table == NULL)
1274 // Create the lr_table, which holds the loss records.
1275 // If the lr_table already exists, it means it contains
1276 // loss_records from the previous leak search. The old_*
1277 // values in these records are used to implement the
1278 // leak check delta mode
1279 lr_table =
1280 VG_(OSetGen_Create)(offsetof(LossRecord, key),
1281 cmp_LossRecordKey_LossRecord,
1282 VG_(malloc), "mc.pr.1",
1283 VG_(free));
1284
philippea22f59d2012-01-26 23:13:52 +00001285 // If we have loss records from a previous search, reset values to have
1286 // proper printing of the deltas between previous search and this search.
1287 n_lossrecords = get_lr_array_from_lr_table();
1288 for (i = 0; i < n_lossrecords; i++) {
philippe4bbfc5f2012-02-27 21:52:45 +00001289 if (lr_array[i]->num_blocks == 0) {
philippea22f59d2012-01-26 23:13:52 +00001290 // remove from lr_table the old loss_records with 0 bytes found
1291 VG_(OSetGen_Remove) (lr_table, &lr_array[i]->key);
philippe4bbfc5f2012-02-27 21:52:45 +00001292 VG_(OSetGen_FreeNode)(lr_table, lr_array[i]);
1293 } else {
philippea22f59d2012-01-26 23:13:52 +00001294 // move the leak sizes to old_* and zero the current sizes
1295 // for next leak search
1296 lr_array[i]->old_szB = lr_array[i]->szB;
1297 lr_array[i]->old_indirect_szB = lr_array[i]->indirect_szB;
1298 lr_array[i]->old_num_blocks = lr_array[i]->num_blocks;
1299 lr_array[i]->szB = 0;
1300 lr_array[i]->indirect_szB = 0;
1301 lr_array[i]->num_blocks = 0;
1302 }
1303 }
1304 // lr_array now contains "invalid" loss records => free it.
1305 // lr_array will be re-created below with the kept and new loss records.
1306 VG_(free) (lr_array);
1307 lr_array = NULL;
njn29a5c012009-05-06 06:15:55 +00001308
1309 // Convert the chunks into loss records, merging them where appropriate.
njn8225cc02009-03-09 22:52:24 +00001310 for (i = 0; i < lc_n_chunks; i++) {
njn29a5c012009-05-06 06:15:55 +00001311 MC_Chunk* ch = lc_chunks[i];
1312 LC_Extra* ex = &(lc_extras)[i];
1313 LossRecord* old_lr;
1314 LossRecordKey lrkey;
1315 lrkey.state = ex->state;
philippe8617b5b2013-01-12 19:53:08 +00001316 lrkey.allocated_at = MC_(allocated_at)(ch);
sewardjb5f6f512005-03-10 23:59:00 +00001317
philippeab1fce92013-09-29 13:47:32 +00001318 if (ex->heuristic) {
1319 MC_(bytes_heuristically_reachable)[ex->heuristic] += ch->szB;
1320 MC_(blocks_heuristically_reachable)[ex->heuristic]++;
1321 if (VG_DEBUG_LEAKCHECK)
1322 VG_(printf)("heuristic %s %#lx len %lu\n",
1323 pp_heuristic(ex->heuristic),
1324 ch->data, (unsigned long)ch->szB);
1325 }
1326
njn29a5c012009-05-06 06:15:55 +00001327 old_lr = VG_(OSetGen_Lookup)(lr_table, &lrkey);
1328 if (old_lr) {
1329 // We found an existing loss record matching this chunk. Update the
1330 // loss record's details in-situ. This is safe because we don't
1331 // change the elements used as the OSet key.
1332 old_lr->szB += ch->szB;
philippea22f59d2012-01-26 23:13:52 +00001333 if (ex->state == Unreached)
1334 old_lr->indirect_szB += ex->IorC.indirect_szB;
njn29a5c012009-05-06 06:15:55 +00001335 old_lr->num_blocks++;
sewardjb5f6f512005-03-10 23:59:00 +00001336 } else {
njn29a5c012009-05-06 06:15:55 +00001337 // No existing loss record matches this chunk. Create a new loss
1338 // record, initialise it from the chunk, and insert it into lr_table.
1339 lr = VG_(OSetGen_AllocNode)(lr_table, sizeof(LossRecord));
1340 lr->key = lrkey;
1341 lr->szB = ch->szB;
philippea22f59d2012-01-26 23:13:52 +00001342 if (ex->state == Unreached)
1343 lr->indirect_szB = ex->IorC.indirect_szB;
1344 else
1345 lr->indirect_szB = 0;
njn29a5c012009-05-06 06:15:55 +00001346 lr->num_blocks = 1;
sewardjc8bd1df2011-06-26 12:41:33 +00001347 lr->old_szB = 0;
1348 lr->old_indirect_szB = 0;
1349 lr->old_num_blocks = 0;
njn29a5c012009-05-06 06:15:55 +00001350 VG_(OSetGen_Insert)(lr_table, lr);
sewardjb5f6f512005-03-10 23:59:00 +00001351 }
1352 }
1353
philippea22f59d2012-01-26 23:13:52 +00001354 // (re-)create the array of pointers to the (new) loss records.
1355 n_lossrecords = get_lr_array_from_lr_table ();
1356 tl_assert(VG_(OSetGen_Size)(lr_table) == n_lossrecords);
njn29a5c012009-05-06 06:15:55 +00001357
1358 // Sort the array by loss record sizes.
1359 VG_(ssort)(lr_array, n_lossrecords, sizeof(LossRecord*),
1360 cmp_LossRecords);
1361
1362 // Zero totals.
njn8225cc02009-03-09 22:52:24 +00001363 MC_(blocks_leaked) = MC_(bytes_leaked) = 0;
1364 MC_(blocks_indirect) = MC_(bytes_indirect) = 0;
1365 MC_(blocks_dubious) = MC_(bytes_dubious) = 0;
1366 MC_(blocks_reachable) = MC_(bytes_reachable) = 0;
1367 MC_(blocks_suppressed) = MC_(bytes_suppressed) = 0;
1368
philippe84234902012-01-14 13:53:13 +00001369 // If there is a maximum nr of loss records we can output, then first
1370 // compute from where the output scan has to start.
1371 // By default, start from the first loss record. Compute a higher
1372 // value if there is a maximum to respect. We need to print the last
1373 // records, as the one with the biggest sizes are more interesting.
1374 start_lr_output_scan = 0;
1375 if (lcp->mode == LC_Full && lcp->max_loss_records_output < n_lossrecords) {
1376 Int nr_printable_records = 0;
1377 for (i = n_lossrecords - 1; i >= 0 && start_lr_output_scan == 0; i--) {
1378 Bool count_as_error, print_record;
1379 lr = lr_array[i];
1380 get_printing_rules (lcp, lr, &count_as_error, &print_record);
1381 // Do not use get_printing_rules results for is_suppressed, as we
1382 // only want to check if the record would be suppressed.
1383 is_suppressed =
1384 MC_(record_leak_error) ( tid, i+1, n_lossrecords, lr,
1385 False /* print_record */,
1386 False /* count_as_error */);
1387 if (print_record && !is_suppressed) {
1388 nr_printable_records++;
1389 if (nr_printable_records == lcp->max_loss_records_output)
1390 start_lr_output_scan = i;
1391 }
sewardjc8bd1df2011-06-26 12:41:33 +00001392 }
philippe84234902012-01-14 13:53:13 +00001393 }
sewardjc8bd1df2011-06-26 12:41:33 +00001394
philippe84234902012-01-14 13:53:13 +00001395 // Print the loss records (in size order) and collect summary stats.
1396 for (i = start_lr_output_scan; i < n_lossrecords; i++) {
1397 Bool count_as_error, print_record;
1398 lr = lr_array[i];
1399 get_printing_rules(lcp, lr, &count_as_error, &print_record);
sewardjb5f6f512005-03-10 23:59:00 +00001400 is_suppressed =
njn18afe5d2009-08-10 08:25:39 +00001401 MC_(record_leak_error) ( tid, i+1, n_lossrecords, lr, print_record,
1402 count_as_error );
sewardjb5f6f512005-03-10 23:59:00 +00001403
1404 if (is_suppressed) {
njn29a5c012009-05-06 06:15:55 +00001405 MC_(blocks_suppressed) += lr->num_blocks;
1406 MC_(bytes_suppressed) += lr->szB;
sewardjb5f6f512005-03-10 23:59:00 +00001407
njn29a5c012009-05-06 06:15:55 +00001408 } else if (Unreached == lr->key.state) {
1409 MC_(blocks_leaked) += lr->num_blocks;
1410 MC_(bytes_leaked) += lr->szB;
sewardjb5f6f512005-03-10 23:59:00 +00001411
njn29a5c012009-05-06 06:15:55 +00001412 } else if (IndirectLeak == lr->key.state) {
1413 MC_(blocks_indirect) += lr->num_blocks;
1414 MC_(bytes_indirect) += lr->szB;
sewardjb5f6f512005-03-10 23:59:00 +00001415
njn29a5c012009-05-06 06:15:55 +00001416 } else if (Possible == lr->key.state) {
1417 MC_(blocks_dubious) += lr->num_blocks;
1418 MC_(bytes_dubious) += lr->szB;
sewardjb5f6f512005-03-10 23:59:00 +00001419
njn29a5c012009-05-06 06:15:55 +00001420 } else if (Reachable == lr->key.state) {
1421 MC_(blocks_reachable) += lr->num_blocks;
1422 MC_(bytes_reachable) += lr->szB;
sewardjb5f6f512005-03-10 23:59:00 +00001423
1424 } else {
njn8225cc02009-03-09 22:52:24 +00001425 VG_(tool_panic)("unknown loss mode");
sewardjb5f6f512005-03-10 23:59:00 +00001426 }
sewardjb5f6f512005-03-10 23:59:00 +00001427 }
sewardjb5f6f512005-03-10 23:59:00 +00001428
njn8225cc02009-03-09 22:52:24 +00001429 if (VG_(clo_verbosity) > 0 && !VG_(clo_xml)) {
floriancf6e7342014-09-28 13:29:06 +00001430 HChar d_bytes[31];
1431 HChar d_blocks[31];
philippeab1fce92013-09-29 13:47:32 +00001432# define DBY(new,old) \
floriancf6e7342014-09-28 13:29:06 +00001433 MC_(snprintf_delta) (d_bytes, sizeof(d_bytes), (new), (old), \
1434 lcp->deltamode)
philippeab1fce92013-09-29 13:47:32 +00001435# define DBL(new,old) \
floriancf6e7342014-09-28 13:29:06 +00001436 MC_(snprintf_delta) (d_blocks, sizeof(d_blocks), (new), (old), \
1437 lcp->deltamode)
sewardjc8bd1df2011-06-26 12:41:33 +00001438
sewardj6b523cd2009-07-15 14:49:40 +00001439 VG_(umsg)("LEAK SUMMARY:\n");
sewardjc8bd1df2011-06-26 12:41:33 +00001440 VG_(umsg)(" definitely lost: %'lu%s bytes in %'lu%s blocks\n",
1441 MC_(bytes_leaked),
philippeab1fce92013-09-29 13:47:32 +00001442 DBY (MC_(bytes_leaked), old_bytes_leaked),
sewardjc8bd1df2011-06-26 12:41:33 +00001443 MC_(blocks_leaked),
philippeab1fce92013-09-29 13:47:32 +00001444 DBL (MC_(blocks_leaked), old_blocks_leaked));
sewardjc8bd1df2011-06-26 12:41:33 +00001445 VG_(umsg)(" indirectly lost: %'lu%s bytes in %'lu%s blocks\n",
1446 MC_(bytes_indirect),
philippeab1fce92013-09-29 13:47:32 +00001447 DBY (MC_(bytes_indirect), old_bytes_indirect),
sewardjc8bd1df2011-06-26 12:41:33 +00001448 MC_(blocks_indirect),
philippeab1fce92013-09-29 13:47:32 +00001449 DBL (MC_(blocks_indirect), old_blocks_indirect));
sewardjc8bd1df2011-06-26 12:41:33 +00001450 VG_(umsg)(" possibly lost: %'lu%s bytes in %'lu%s blocks\n",
1451 MC_(bytes_dubious),
philippeab1fce92013-09-29 13:47:32 +00001452 DBY (MC_(bytes_dubious), old_bytes_dubious),
sewardjc8bd1df2011-06-26 12:41:33 +00001453 MC_(blocks_dubious),
philippeab1fce92013-09-29 13:47:32 +00001454 DBL (MC_(blocks_dubious), old_blocks_dubious));
sewardjc8bd1df2011-06-26 12:41:33 +00001455 VG_(umsg)(" still reachable: %'lu%s bytes in %'lu%s blocks\n",
1456 MC_(bytes_reachable),
philippeab1fce92013-09-29 13:47:32 +00001457 DBY (MC_(bytes_reachable), old_bytes_reachable),
sewardjc8bd1df2011-06-26 12:41:33 +00001458 MC_(blocks_reachable),
philippeab1fce92013-09-29 13:47:32 +00001459 DBL (MC_(blocks_reachable), old_blocks_reachable));
1460 for (i = 0; i < N_LEAK_CHECK_HEURISTICS; i++)
1461 if (old_blocks_heuristically_reachable[i] > 0
1462 || MC_(blocks_heuristically_reachable)[i] > 0) {
1463 VG_(umsg)(" of which "
1464 "reachable via heuristic:\n");
1465 break;
1466 }
1467 for (i = 0; i < N_LEAK_CHECK_HEURISTICS; i++)
1468 if (old_blocks_heuristically_reachable[i] > 0
1469 || MC_(blocks_heuristically_reachable)[i] > 0)
florian866862a2014-12-13 18:35:00 +00001470 VG_(umsg)(" %-19s: "
philippeab1fce92013-09-29 13:47:32 +00001471 "%'lu%s bytes in %'lu%s blocks\n",
1472 pp_heuristic(i),
1473 MC_(bytes_heuristically_reachable)[i],
1474 DBY (MC_(bytes_heuristically_reachable)[i],
1475 old_bytes_heuristically_reachable[i]),
1476 MC_(blocks_heuristically_reachable)[i],
1477 DBL (MC_(blocks_heuristically_reachable)[i],
1478 old_blocks_heuristically_reachable[i]));
sewardjc8bd1df2011-06-26 12:41:33 +00001479 VG_(umsg)(" suppressed: %'lu%s bytes in %'lu%s blocks\n",
1480 MC_(bytes_suppressed),
philippeab1fce92013-09-29 13:47:32 +00001481 DBY (MC_(bytes_suppressed), old_bytes_suppressed),
sewardjc8bd1df2011-06-26 12:41:33 +00001482 MC_(blocks_suppressed),
philippeab1fce92013-09-29 13:47:32 +00001483 DBL (MC_(blocks_suppressed), old_blocks_suppressed));
philippe84234902012-01-14 13:53:13 +00001484 if (lcp->mode != LC_Full &&
njn8225cc02009-03-09 22:52:24 +00001485 (MC_(blocks_leaked) + MC_(blocks_indirect) +
1486 MC_(blocks_dubious) + MC_(blocks_reachable)) > 0) {
philippe84234902012-01-14 13:53:13 +00001487 if (lcp->requested_by_monitor_command)
philippeab1fce92013-09-29 13:47:32 +00001488 VG_(umsg)("To see details of leaked memory, "
1489 "give 'full' arg to leak_check\n");
sewardjc8bd1df2011-06-26 12:41:33 +00001490 else
1491 VG_(umsg)("Rerun with --leak-check=full to see details "
1492 "of leaked memory\n");
njn8225cc02009-03-09 22:52:24 +00001493 }
philippe84234902012-01-14 13:53:13 +00001494 if (lcp->mode == LC_Full &&
philippeab1fce92013-09-29 13:47:32 +00001495 MC_(blocks_reachable) > 0 && !RiS(Reachable,lcp->show_leak_kinds)) {
sewardj6b523cd2009-07-15 14:49:40 +00001496 VG_(umsg)("Reachable blocks (those to which a pointer "
1497 "was found) are not shown.\n");
philippe84234902012-01-14 13:53:13 +00001498 if (lcp->requested_by_monitor_command)
sewardj30b3eca2011-06-28 08:20:39 +00001499 VG_(umsg)("To see them, add 'reachable any' args to leak_check\n");
sewardjc8bd1df2011-06-26 12:41:33 +00001500 else
1501 VG_(umsg)("To see them, rerun with: --leak-check=full "
philippe2193a7c2012-12-08 17:54:16 +00001502 "--show-leak-kinds=all\n");
sewardjb5f6f512005-03-10 23:59:00 +00001503 }
njnb6267bd2009-08-12 00:14:16 +00001504 VG_(umsg)("\n");
philippeab1fce92013-09-29 13:47:32 +00001505 #undef DBL
1506 #undef DBY
sewardjb5f6f512005-03-10 23:59:00 +00001507 }
1508}
1509
philippea22f59d2012-01-26 23:13:52 +00001510// print recursively all indirectly leaked blocks collected in clique.
1511static void print_clique (Int clique, UInt level)
1512{
1513 Int ind;
1514 Int i, n_lossrecords;;
1515
1516 n_lossrecords = VG_(OSetGen_Size)(lr_table);
1517
1518 for (ind = 0; ind < lc_n_chunks; ind++) {
1519 LC_Extra* ind_ex = &(lc_extras)[ind];
philippeab1fce92013-09-29 13:47:32 +00001520 if (ind_ex->state == IndirectLeak
1521 && ind_ex->IorC.clique == (SizeT) clique) {
philippea22f59d2012-01-26 23:13:52 +00001522 MC_Chunk* ind_ch = lc_chunks[ind];
1523 LossRecord* ind_lr;
1524 LossRecordKey ind_lrkey;
1525 Int lr_i;
1526 ind_lrkey.state = ind_ex->state;
philippe8617b5b2013-01-12 19:53:08 +00001527 ind_lrkey.allocated_at = MC_(allocated_at)(ind_ch);
philippea22f59d2012-01-26 23:13:52 +00001528 ind_lr = VG_(OSetGen_Lookup)(lr_table, &ind_lrkey);
1529 for (lr_i = 0; lr_i < n_lossrecords; lr_i++)
1530 if (ind_lr == lr_array[lr_i])
1531 break;
1532 for (i = 0; i < level; i++)
1533 VG_(umsg)(" ");
1534 VG_(umsg)("%p[%lu] indirect loss record %d\n",
1535 (void *)ind_ch->data, (unsigned long)ind_ch->szB,
1536 lr_i+1); // lr_i+1 for user numbering.
1537 if (lr_i >= n_lossrecords)
1538 VG_(umsg)
1539 ("error: no indirect loss record found for %p[%lu]?????\n",
1540 (void *)ind_ch->data, (unsigned long)ind_ch->szB);
1541 print_clique(ind, level+1);
1542 }
1543 }
1544 }
1545
1546Bool MC_(print_block_list) ( UInt loss_record_nr)
1547{
1548 Int i, n_lossrecords;
1549 LossRecord* lr;
1550
1551 if (lr_table == NULL || lc_chunks == NULL || lc_extras == NULL) {
1552 VG_(umsg)("Can't print block list : no valid leak search result\n");
1553 return False;
1554 }
1555
1556 if (lc_chunks_n_frees_marker != MC_(get_cmalloc_n_frees)()) {
1557 VG_(umsg)("Can't print obsolete block list : redo a leak search first\n");
1558 return False;
1559 }
1560
1561 n_lossrecords = VG_(OSetGen_Size)(lr_table);
1562 if (loss_record_nr >= n_lossrecords)
1563 return False; // Invalid loss record nr.
1564
1565 tl_assert (lr_array);
1566 lr = lr_array[loss_record_nr];
1567
1568 // (re-)print the loss record details.
1569 // (+1 on loss_record_nr as user numbering for loss records starts at 1).
1570 MC_(pp_LossRecord)(loss_record_nr+1, n_lossrecords, lr);
1571
1572 // Match the chunks with loss records.
1573 for (i = 0; i < lc_n_chunks; i++) {
1574 MC_Chunk* ch = lc_chunks[i];
1575 LC_Extra* ex = &(lc_extras)[i];
1576 LossRecord* old_lr;
1577 LossRecordKey lrkey;
1578 lrkey.state = ex->state;
philippe8617b5b2013-01-12 19:53:08 +00001579 lrkey.allocated_at = MC_(allocated_at)(ch);
philippea22f59d2012-01-26 23:13:52 +00001580
1581 old_lr = VG_(OSetGen_Lookup)(lr_table, &lrkey);
1582 if (old_lr) {
1583 // We found an existing loss record matching this chunk.
philippeab1fce92013-09-29 13:47:32 +00001584 // If this is the loss record we are looking for, output the pointer.
philippea22f59d2012-01-26 23:13:52 +00001585 if (old_lr == lr_array[loss_record_nr]) {
1586 VG_(umsg)("%p[%lu]\n",
1587 (void *)ch->data, (unsigned long) ch->szB);
1588 if (ex->state != Reachable) {
1589 // We can print the clique in all states, except Reachable.
1590 // In Unreached state, lc_chunk[i] is the clique leader.
1591 // In IndirectLeak, lc_chunk[i] might have been a clique leader
1592 // which was later collected in another clique.
1593 // For Possible, lc_chunk[i] might be the top of a clique
1594 // or an intermediate clique.
1595 print_clique(i, 1);
1596 }
1597 }
1598 } else {
1599 // No existing loss record matches this chunk ???
1600 VG_(umsg)("error: no loss record found for %p[%lu]?????\n",
1601 (void *)ch->data, (unsigned long) ch->szB);
1602 }
1603 }
1604 return True;
1605}
1606
1607// If searched = 0, scan memory root set, pushing onto the mark stack the blocks
1608// encountered.
philippeab1fce92013-09-29 13:47:32 +00001609// Otherwise (searched != 0), scan the memory root set searching for ptr
1610// pointing inside [searched, searched+szB[.
philippea22f59d2012-01-26 23:13:52 +00001611static void scan_memory_root_set(Addr searched, SizeT szB)
1612{
1613 Int i;
1614 Int n_seg_starts;
1615 Addr* seg_starts = VG_(get_segment_starts)( &n_seg_starts );
1616
1617 tl_assert(seg_starts && n_seg_starts > 0);
1618
1619 lc_scanned_szB = 0;
philippe7a76f4b2013-10-06 21:23:04 +00001620 lc_sig_skipped_szB = 0;
philippea22f59d2012-01-26 23:13:52 +00001621
1622 // VG_(am_show_nsegments)( 0, "leakcheck");
1623 for (i = 0; i < n_seg_starts; i++) {
1624 SizeT seg_size;
1625 NSegment const* seg = VG_(am_find_nsegment)( seg_starts[i] );
1626 tl_assert(seg);
1627
1628 if (seg->kind != SkFileC && seg->kind != SkAnonC) continue;
1629 if (!(seg->hasR && seg->hasW)) continue;
1630 if (seg->isCH) continue;
1631
1632 // Don't poke around in device segments as this may cause
1633 // hangs. Exclude /dev/zero just in case someone allocated
1634 // memory by explicitly mapping /dev/zero.
1635 if (seg->kind == SkFileC
1636 && (VKI_S_ISCHR(seg->mode) || VKI_S_ISBLK(seg->mode))) {
floriand3166c42015-01-24 00:02:19 +00001637 const HChar* dev_name = VG_(am_get_filename)( seg );
philippea22f59d2012-01-26 23:13:52 +00001638 if (dev_name && 0 == VG_(strcmp)(dev_name, "/dev/zero")) {
1639 // Don't skip /dev/zero.
1640 } else {
1641 // Skip this device mapping.
1642 continue;
1643 }
1644 }
1645
1646 if (0)
1647 VG_(printf)("ACCEPT %2d %#lx %#lx\n", i, seg->start, seg->end);
1648
1649 // Scan the segment. We use -1 for the clique number, because this
1650 // is a root-set.
1651 seg_size = seg->end - seg->start + 1;
1652 if (VG_(clo_verbosity) > 2) {
1653 VG_(message)(Vg_DebugMsg,
1654 " Scanning root segment: %#lx..%#lx (%lu)\n",
1655 seg->start, seg->end, seg_size);
1656 }
1657 lc_scan_memory(seg->start, seg_size, /*is_prior_definite*/True,
1658 /*clique*/-1, /*cur_clique*/-1,
1659 searched, szB);
1660 }
philippe7d69fd92012-02-26 21:26:00 +00001661 VG_(free)(seg_starts);
philippea22f59d2012-01-26 23:13:52 +00001662}
1663
njn8225cc02009-03-09 22:52:24 +00001664/*------------------------------------------------------------*/
1665/*--- Top-level entry point. ---*/
1666/*------------------------------------------------------------*/
sewardj3cf26a52006-07-27 23:48:53 +00001667
philippe84234902012-01-14 13:53:13 +00001668void MC_(detect_memory_leaks) ( ThreadId tid, LeakCheckParams* lcp)
njn43c799e2003-04-08 00:08:52 +00001669{
njnb965efb2009-08-10 07:36:54 +00001670 Int i, j;
njn43c799e2003-04-08 00:08:52 +00001671
philippe84234902012-01-14 13:53:13 +00001672 tl_assert(lcp->mode != LC_Off);
sewardjc8bd1df2011-06-26 12:41:33 +00001673
philippe57a16a22012-07-18 22:26:51 +00001674 // Verify some assertions which are used in lc_scan_memory.
1675 tl_assert((VKI_PAGE_SIZE % sizeof(Addr)) == 0);
1676 tl_assert((SM_SIZE % sizeof(Addr)) == 0);
1677 // Above two assertions are critical, while below assertion
1678 // ensures that the optimisation in the loop is done in the
1679 // correct order : the loop checks for (big) SM chunk skipping
1680 // before checking for (smaller) page skipping.
1681 tl_assert((SM_SIZE % VKI_PAGE_SIZE) == 0);
1682
philippe4e32d672013-10-17 22:10:41 +00001683 MC_(leak_search_gen)++;
philippe84234902012-01-14 13:53:13 +00001684 MC_(detect_memory_leaks_last_delta_mode) = lcp->deltamode;
philippeab1fce92013-09-29 13:47:32 +00001685 detect_memory_leaks_last_heuristics = lcp->heuristics;
njn43c799e2003-04-08 00:08:52 +00001686
njn8225cc02009-03-09 22:52:24 +00001687 // Get the chunks, stop if there were none.
philippea22f59d2012-01-26 23:13:52 +00001688 if (lc_chunks) {
1689 VG_(free)(lc_chunks);
1690 lc_chunks = NULL;
1691 }
njn8225cc02009-03-09 22:52:24 +00001692 lc_chunks = find_active_chunks(&lc_n_chunks);
philippea22f59d2012-01-26 23:13:52 +00001693 lc_chunks_n_frees_marker = MC_(get_cmalloc_n_frees)();
njn8225cc02009-03-09 22:52:24 +00001694 if (lc_n_chunks == 0) {
1695 tl_assert(lc_chunks == NULL);
sewardjc8bd1df2011-06-26 12:41:33 +00001696 if (lr_table != NULL) {
philippea22f59d2012-01-26 23:13:52 +00001697 // forget the previous recorded LossRecords as next leak search
1698 // can in any case just create new leaks.
sewardjc8bd1df2011-06-26 12:41:33 +00001699 // Maybe it would be better to rather call print_result ?
philippea22f59d2012-01-26 23:13:52 +00001700 // (at least when leak decreases are requested)
sewardjc8bd1df2011-06-26 12:41:33 +00001701 // This will then output all LossRecords with a size decreasing to 0
1702 VG_(OSetGen_Destroy) (lr_table);
philippea22f59d2012-01-26 23:13:52 +00001703 lr_table = NULL;
sewardjc8bd1df2011-06-26 12:41:33 +00001704 }
sewardj71bc3cb2005-05-19 00:25:45 +00001705 if (VG_(clo_verbosity) >= 1 && !VG_(clo_xml)) {
njnb6267bd2009-08-12 00:14:16 +00001706 VG_(umsg)("All heap blocks were freed -- no leaks are possible\n");
sewardj2d9e8742009-08-07 15:46:56 +00001707 VG_(umsg)("\n");
sewardj37d06f22003-09-17 21:48:26 +00001708 }
njn43c799e2003-04-08 00:08:52 +00001709 return;
1710 }
1711
njn8225cc02009-03-09 22:52:24 +00001712 // Sort the array so blocks are in ascending order in memory.
1713 VG_(ssort)(lc_chunks, lc_n_chunks, sizeof(VgHashNode*), compare_MC_Chunks);
njn43c799e2003-04-08 00:08:52 +00001714
njn8225cc02009-03-09 22:52:24 +00001715 // Sanity check -- make sure they're in order.
1716 for (i = 0; i < lc_n_chunks-1; i++) {
1717 tl_assert( lc_chunks[i]->data <= lc_chunks[i+1]->data);
1718 }
njn43c799e2003-04-08 00:08:52 +00001719
njnb965efb2009-08-10 07:36:54 +00001720 // Sanity check -- make sure they don't overlap. The one exception is that
1721 // we allow a MALLOCLIKE block to sit entirely within a malloc() block.
1722 // This is for bug 100628. If this occurs, we ignore the malloc() block
1723 // for leak-checking purposes. This is a hack and probably should be done
1724 // better, but at least it's consistent with mempools (which are treated
1725 // like this in find_active_chunks). Mempools have a separate VgHashTable
1726 // for mempool chunks, but if custom-allocated blocks are put in a separate
1727 // table from normal heap blocks it makes free-mismatch checking more
1728 // difficult.
1729 //
1730 // If this check fails, it probably means that the application
njn8225cc02009-03-09 22:52:24 +00001731 // has done something stupid with VALGRIND_MALLOCLIKE_BLOCK client
njnb965efb2009-08-10 07:36:54 +00001732 // requests, eg. has made overlapping requests (which are
1733 // nonsensical), or used VALGRIND_MALLOCLIKE_BLOCK for stack locations;
1734 // again nonsensical.
1735 //
njn8225cc02009-03-09 22:52:24 +00001736 for (i = 0; i < lc_n_chunks-1; i++) {
1737 MC_Chunk* ch1 = lc_chunks[i];
1738 MC_Chunk* ch2 = lc_chunks[i+1];
njnb965efb2009-08-10 07:36:54 +00001739
1740 Addr start1 = ch1->data;
1741 Addr start2 = ch2->data;
1742 Addr end1 = ch1->data + ch1->szB - 1;
1743 Addr end2 = ch2->data + ch2->szB - 1;
1744 Bool isCustom1 = ch1->allockind == MC_AllocCustom;
1745 Bool isCustom2 = ch2->allockind == MC_AllocCustom;
1746
1747 if (end1 < start2) {
1748 // Normal case - no overlap.
1749
1750 // We used to allow exact duplicates, I'm not sure why. --njn
1751 //} else if (start1 == start2 && end1 == end2) {
1752 // Degenerate case: exact duplicates.
1753
1754 } else if (start1 >= start2 && end1 <= end2 && isCustom1 && !isCustom2) {
1755 // Block i is MALLOCLIKE and entirely within block i+1.
1756 // Remove block i+1.
1757 for (j = i+1; j < lc_n_chunks-1; j++) {
1758 lc_chunks[j] = lc_chunks[j+1];
1759 }
1760 lc_n_chunks--;
1761
1762 } else if (start2 >= start1 && end2 <= end1 && isCustom2 && !isCustom1) {
1763 // Block i+1 is MALLOCLIKE and entirely within block i.
1764 // Remove block i.
1765 for (j = i; j < lc_n_chunks-1; j++) {
1766 lc_chunks[j] = lc_chunks[j+1];
1767 }
1768 lc_n_chunks--;
1769
1770 } else {
philippe09007e32012-03-01 22:00:36 +00001771 VG_(umsg)("Block 0x%lx..0x%lx overlaps with block 0x%lx..0x%lx\n",
bart3c4fa9f2011-05-09 10:46:55 +00001772 start1, end1, start2, end2);
philippe09007e32012-03-01 22:00:36 +00001773 VG_(umsg)("Blocks allocation contexts:\n"),
philippe8617b5b2013-01-12 19:53:08 +00001774 VG_(pp_ExeContext)( MC_(allocated_at)(ch1));
philippe09007e32012-03-01 22:00:36 +00001775 VG_(umsg)("\n"),
philippe8617b5b2013-01-12 19:53:08 +00001776 VG_(pp_ExeContext)( MC_(allocated_at)(ch2));
njnb965efb2009-08-10 07:36:54 +00001777 VG_(umsg)("This is usually caused by using VALGRIND_MALLOCLIKE_BLOCK");
philippe09007e32012-03-01 22:00:36 +00001778 VG_(umsg)("in an inappropriate way.\n");
njnb965efb2009-08-10 07:36:54 +00001779 tl_assert (0);
njn8225cc02009-03-09 22:52:24 +00001780 }
njn8225cc02009-03-09 22:52:24 +00001781 }
1782
1783 // Initialise lc_extras.
philippea22f59d2012-01-26 23:13:52 +00001784 if (lc_extras) {
1785 VG_(free)(lc_extras);
1786 lc_extras = NULL;
1787 }
njn8225cc02009-03-09 22:52:24 +00001788 lc_extras = VG_(malloc)( "mc.dml.2", lc_n_chunks * sizeof(LC_Extra) );
1789 for (i = 0; i < lc_n_chunks; i++) {
1790 lc_extras[i].state = Unreached;
tom1d0f3f62010-10-04 20:55:21 +00001791 lc_extras[i].pending = False;
philippeab1fce92013-09-29 13:47:32 +00001792 lc_extras[i].heuristic = LchNone;
philippea22f59d2012-01-26 23:13:52 +00001793 lc_extras[i].IorC.indirect_szB = 0;
njn8225cc02009-03-09 22:52:24 +00001794 }
1795
1796 // Initialise lc_markstack.
1797 lc_markstack = VG_(malloc)( "mc.dml.2", lc_n_chunks * sizeof(Int) );
1798 for (i = 0; i < lc_n_chunks; i++) {
1799 lc_markstack[i] = -1;
sewardjb5f6f512005-03-10 23:59:00 +00001800 }
1801 lc_markstack_top = -1;
njn43c799e2003-04-08 00:08:52 +00001802
njn8225cc02009-03-09 22:52:24 +00001803 // Verbosity.
sewardj2d9e8742009-08-07 15:46:56 +00001804 if (VG_(clo_verbosity) > 1 && !VG_(clo_xml)) {
njnb6267bd2009-08-12 00:14:16 +00001805 VG_(umsg)( "Searching for pointers to %'d not-freed blocks\n",
sewardj6b523cd2009-07-15 14:49:40 +00001806 lc_n_chunks );
sewardj2d9e8742009-08-07 15:46:56 +00001807 }
sewardjb5f6f512005-03-10 23:59:00 +00001808
njn8225cc02009-03-09 22:52:24 +00001809 // Scan the memory root-set, pushing onto the mark stack any blocks
1810 // pointed to.
philippea22f59d2012-01-26 23:13:52 +00001811 scan_memory_root_set(/*searched*/0, 0);
sewardjb5f6f512005-03-10 23:59:00 +00001812
njn8225cc02009-03-09 22:52:24 +00001813 // Scan GP registers for chunk pointers.
1814 VG_(apply_to_GP_regs)(lc_push_if_a_chunk_ptr_register);
sewardjb5f6f512005-03-10 23:59:00 +00001815
njn8225cc02009-03-09 22:52:24 +00001816 // Process the pushed blocks. After this, every block that is reachable
1817 // from the root-set has been traced.
1818 lc_process_markstack(/*clique*/-1);
njn43c799e2003-04-08 00:08:52 +00001819
njnb6267bd2009-08-12 00:14:16 +00001820 if (VG_(clo_verbosity) > 1 && !VG_(clo_xml)) {
1821 VG_(umsg)("Checked %'lu bytes\n", lc_scanned_szB);
philippe7a76f4b2013-10-06 21:23:04 +00001822 if (lc_sig_skipped_szB > 0)
1823 VG_(umsg)("Skipped %'lu bytes due to read errors\n",
1824 lc_sig_skipped_szB);
njnb6267bd2009-08-12 00:14:16 +00001825 VG_(umsg)( "\n" );
1826 }
njn43c799e2003-04-08 00:08:52 +00001827
njn8225cc02009-03-09 22:52:24 +00001828 // Trace all the leaked blocks to determine which are directly leaked and
1829 // which are indirectly leaked. For each Unreached block, push it onto
1830 // the mark stack, and find all the as-yet-Unreached blocks reachable
1831 // from it. These form a clique and are marked IndirectLeak, and their
1832 // size is added to the clique leader's indirect size. If one of the
1833 // found blocks was itself a clique leader (from a previous clique), then
1834 // the cliques are merged.
1835 for (i = 0; i < lc_n_chunks; i++) {
1836 MC_Chunk* ch = lc_chunks[i];
1837 LC_Extra* ex = &(lc_extras[i]);
njn43c799e2003-04-08 00:08:52 +00001838
njn8225cc02009-03-09 22:52:24 +00001839 if (VG_DEBUG_CLIQUE)
1840 VG_(printf)("cliques: %d at %#lx -> Loss state %d\n",
1841 i, ch->data, ex->state);
njn43c799e2003-04-08 00:08:52 +00001842
njn8225cc02009-03-09 22:52:24 +00001843 tl_assert(lc_markstack_top == -1);
1844
1845 if (ex->state == Unreached) {
1846 if (VG_DEBUG_CLIQUE)
1847 VG_(printf)("%d: gathering clique %#lx\n", i, ch->data);
1848
1849 // Push this Unreached block onto the stack and process it.
1850 lc_push(i, ch);
philippea22f59d2012-01-26 23:13:52 +00001851 lc_process_markstack(/*clique*/i);
njn8225cc02009-03-09 22:52:24 +00001852
1853 tl_assert(lc_markstack_top == -1);
1854 tl_assert(ex->state == Unreached);
nethercote0f19bce2003-12-02 10:17:44 +00001855 }
njn43c799e2003-04-08 00:08:52 +00001856 }
njn8225cc02009-03-09 22:52:24 +00001857
sewardjc8bd1df2011-06-26 12:41:33 +00001858 print_results( tid, lcp);
njn43c799e2003-04-08 00:08:52 +00001859
sewardjb5f6f512005-03-10 23:59:00 +00001860 VG_(free) ( lc_markstack );
philippea22f59d2012-01-26 23:13:52 +00001861 lc_markstack = NULL;
1862 // lc_chunks, lc_extras, lr_array and lr_table are kept (needed if user
1863 // calls MC_(print_block_list)). lr_table also used for delta leak reporting
1864 // between this leak search and the next leak search.
1865}
1866
1867static Addr searched_wpa;
1868static SizeT searched_szB;
1869static void
florian6bd9dc12012-11-23 16:17:43 +00001870search_address_in_GP_reg(ThreadId tid, const HChar* regname, Addr addr_in_reg)
philippea22f59d2012-01-26 23:13:52 +00001871{
1872 if (addr_in_reg >= searched_wpa
1873 && addr_in_reg < searched_wpa + searched_szB) {
1874 if (addr_in_reg == searched_wpa)
1875 VG_(umsg)
1876 ("tid %d register %s pointing at %#lx\n",
1877 tid, regname, searched_wpa);
1878 else
1879 VG_(umsg)
1880 ("tid %d register %s interior pointing %lu bytes inside %#lx\n",
1881 tid, regname, (long unsigned) addr_in_reg - searched_wpa,
1882 searched_wpa);
1883 }
1884}
1885
1886void MC_(who_points_at) ( Addr address, SizeT szB)
1887{
1888 MC_Chunk** chunks;
1889 Int n_chunks;
1890 Int i;
1891
1892 if (szB == 1)
1893 VG_(umsg) ("Searching for pointers to %#lx\n", address);
1894 else
1895 VG_(umsg) ("Searching for pointers pointing in %lu bytes from %#lx\n",
1896 szB, address);
1897
philippeab1fce92013-09-29 13:47:32 +00001898 chunks = find_active_chunks(&n_chunks);
1899
philippea22f59d2012-01-26 23:13:52 +00001900 // Scan memory root-set, searching for ptr pointing in address[szB]
1901 scan_memory_root_set(address, szB);
1902
1903 // Scan active malloc-ed chunks
philippea22f59d2012-01-26 23:13:52 +00001904 for (i = 0; i < n_chunks; i++) {
1905 lc_scan_memory(chunks[i]->data, chunks[i]->szB,
1906 /*is_prior_definite*/True,
1907 /*clique*/-1, /*cur_clique*/-1,
1908 address, szB);
1909 }
1910 VG_(free) ( chunks );
1911
1912 // Scan GP registers for pointers to address range.
1913 searched_wpa = address;
1914 searched_szB = szB;
1915 VG_(apply_to_GP_regs)(search_address_in_GP_reg);
1916
njn43c799e2003-04-08 00:08:52 +00001917}
1918
1919/*--------------------------------------------------------------------*/
njn1d0825f2006-03-27 11:37:07 +00001920/*--- end ---*/
njn43c799e2003-04-08 00:08:52 +00001921/*--------------------------------------------------------------------*/
1922