зеркало из https://github.com/mozilla/gecko-dev.git
439 строки
15 KiB
C++
439 строки
15 KiB
C++
/* ***** BEGIN LICENSE BLOCK *****
|
|
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
|
|
*
|
|
* The contents of this file are subject to the Mozilla Public License Version
|
|
* 1.1 (the "License"); you may not use this file except in compliance with
|
|
* the License. You may obtain a copy of the License at
|
|
* http://www.mozilla.org/MPL/
|
|
*
|
|
* Software distributed under the License is distributed on an "AS IS" basis,
|
|
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
|
|
* for the specific language governing rights and limitations under the
|
|
* License.
|
|
*
|
|
* The Original Code is leaksoup3, a memory graph analysis tool that
|
|
* finds roots based on strongly connected components (SCCs).
|
|
*
|
|
* The Initial Developer of the Original Code is L. David Baron.
|
|
* Portions created by the Initial Developer are Copyright (C) 2002
|
|
* the Initial Developer. All Rights Reserved.
|
|
*
|
|
* Contributor(s):
|
|
* L. David Baron <dbaron@fas.harvard.edu> (original author)
|
|
*
|
|
* Alternatively, the contents of this file may be used under the terms of
|
|
* either the GNU General Public License Version 2 or later (the "GPL"), or
|
|
* the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
|
|
* in which case the provisions of the GPL or the LGPL are applicable instead
|
|
* of those above. If you wish to allow use of your version of this file only
|
|
* under the terms of either the GPL or the LGPL, and not to allow others to
|
|
* use your version of this file under the terms of the MPL, indicate your
|
|
* decision by deleting the provisions above and replace them with the notice
|
|
* and other provisions required by the GPL or the LGPL. If you do not delete
|
|
* the provisions above, a recipient may use your version of this file under
|
|
* the terms of any one of the MPL, the GPL or the LGPL.
|
|
*
|
|
* ***** END LICENSE BLOCK ***** */
|
|
|
|
#include "adreader.h"
|
|
|
|
#include <stdio.h>
|
|
#include "plhash.h"
|
|
|
|
#include "nsVoidArray.h"
|
|
#include "nsQuickSort.h"
|
|
|
|
/*
|
|
* Read in an allocation dump, presumably one taken at shutdown (using
|
|
* the --shutdown-leaks=file option, which must be used along with
|
|
* --trace-malloc=tmlog), and treat the memory in the dump as leaks.
|
|
* Find the leak roots, including cycles that are roots, by finding the
|
|
* strongly connected components in the graph. Print output to stdout
|
|
* as HTML.
|
|
*/
|
|
|
|
struct AllocationNode {
|
|
const ADLog::Entry *entry;
|
|
|
|
// Other |AllocationNode| objects whose memory has a pointer to
|
|
// this object.
|
|
nsAutoVoidArray pointers_to;
|
|
|
|
// The reverse.
|
|
nsAutoVoidArray pointers_from;
|
|
|
|
// Early on in the algorithm, the pre-order index from a DFS.
|
|
// Later on, set to the index of the strongly connected component to
|
|
// which this node belongs.
|
|
PRUint32 index;
|
|
|
|
PRPackedBool reached;
|
|
PRPackedBool is_root;
|
|
};
|
|
|
|
static PLHashNumber hash_pointer(const void *key)
|
|
{
|
|
return (PLHashNumber) NS_PTR_TO_INT32(key);
|
|
}
|
|
|
|
static int sort_by_index(const void* e1, const void* e2, void*)
|
|
{
|
|
const AllocationNode *n1 = *NS_STATIC_CAST(const AllocationNode*const*, e1);
|
|
const AllocationNode *n2 = *NS_STATIC_CAST(const AllocationNode*const*, e2);
|
|
return n1->index - n2->index;
|
|
}
|
|
|
|
static int sort_by_reverse_index(const void* e1, const void* e2, void*)
|
|
{
|
|
const AllocationNode *n1 = *NS_STATIC_CAST(const AllocationNode*const*, e1);
|
|
const AllocationNode *n2 = *NS_STATIC_CAST(const AllocationNode*const*, e2);
|
|
return n2->index - n1->index;
|
|
}
|
|
|
|
static void print_escaped(FILE *aStream, const char* aData)
|
|
{
|
|
char c;
|
|
char buf[1000];
|
|
char *p = buf;
|
|
while ((c = *aData++)) {
|
|
switch (c) {
|
|
#define CH(char) *p++ = char
|
|
case '<':
|
|
CH('&'); CH('l'); CH('t'); CH(';');
|
|
break;
|
|
case '>':
|
|
CH('&'); CH('g'); CH('t'); CH(';');
|
|
break;
|
|
case '&':
|
|
CH('&'); CH('a'); CH('m'); CH('p'); CH(';');
|
|
break;
|
|
default:
|
|
CH(c);
|
|
break;
|
|
#undef CH
|
|
}
|
|
if (p + 10 > buf + sizeof(buf)) {
|
|
*p = '\0';
|
|
fputs(buf, aStream);
|
|
p = buf;
|
|
}
|
|
}
|
|
*p = '\0';
|
|
fputs(buf, aStream);
|
|
}
|
|
|
|
int main(int argc, char **argv)
|
|
{
|
|
if (argc != 2) {
|
|
fprintf(stderr,
|
|
"Expected usage: %s <sd-leak-file>\n"
|
|
" sd-leak-file: Output of --shutdown-leaks=<file> option.\n",
|
|
argv[0]);
|
|
return 1;
|
|
}
|
|
|
|
ADLog log;
|
|
if (!log.Read(argv[1])) {
|
|
fprintf(stderr,
|
|
"%s: Error reading input file %s.\n", argv[0], argv[1]);
|
|
}
|
|
|
|
const size_t count = log.count();
|
|
|
|
PLHashTable *memory_map =
|
|
PL_NewHashTable(count * 8, hash_pointer, PL_CompareValues,
|
|
PL_CompareValues, 0, 0);
|
|
if (!memory_map) {
|
|
fprintf(stderr, "%s: Out of memory.\n", argv[0]);
|
|
return 1;
|
|
}
|
|
|
|
// Create one |AllocationNode| object for each log entry, and create
|
|
// entries in the hashtable pointing to it for each byte it occupies.
|
|
AllocationNode *nodes = new AllocationNode[count];
|
|
if (!nodes) {
|
|
fprintf(stderr, "%s: Out of memory.\n", argv[0]);
|
|
return 1;
|
|
}
|
|
|
|
{
|
|
AllocationNode *cur_node = nodes;
|
|
for (ADLog::const_iterator entry = log.begin(), entry_end = log.end();
|
|
entry != entry_end; ++entry, ++cur_node) {
|
|
const ADLog::Entry *e = cur_node->entry = *entry;
|
|
cur_node->reached = PR_FALSE;
|
|
|
|
for (ADLog::Pointer p = e->address,
|
|
p_end = e->address + e->datasize;
|
|
p != p_end; ++p) {
|
|
PLHashEntry *e = PL_HashTableAdd(memory_map, p, cur_node);
|
|
if (!e) {
|
|
fprintf(stderr, "%s: Out of memory.\n", argv[0]);
|
|
return 1;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// Construct graph based on pointers.
|
|
for (AllocationNode *node = nodes, *node_end = nodes + count;
|
|
node != node_end; ++node) {
|
|
const ADLog::Entry *e = node->entry;
|
|
for (const char *d = e->data, *d_end = e->data + e->datasize -
|
|
e->datasize % sizeof(ADLog::Pointer);
|
|
d != d_end; d += sizeof(ADLog::Pointer)) {
|
|
AllocationNode *target = (AllocationNode*)
|
|
PL_HashTableLookup(memory_map, *(void**)d);
|
|
if (target) {
|
|
target->pointers_from.AppendElement(node);
|
|
node->pointers_to.AppendElement(target);
|
|
}
|
|
}
|
|
}
|
|
|
|
// Do a depth-first search on the graph (i.e., by following
|
|
// |pointers_to|) and assign the post-order index to |index|.
|
|
{
|
|
PRUint32 dfs_index = 0;
|
|
nsVoidArray stack;
|
|
|
|
for (AllocationNode *n = nodes, *n_end = nodes+count; n != n_end; ++n) {
|
|
if (n->reached) {
|
|
continue;
|
|
}
|
|
stack.AppendElement(n);
|
|
|
|
do {
|
|
PRUint32 pos = stack.Count() - 1;
|
|
AllocationNode *n =
|
|
NS_STATIC_CAST(AllocationNode*, stack[pos]);
|
|
if (n->reached) {
|
|
n->index = dfs_index++;
|
|
stack.RemoveElementAt(pos);
|
|
} else {
|
|
n->reached = PR_TRUE;
|
|
|
|
// When doing post-order processing, we have to be
|
|
// careful not to put reached nodes into the stack.
|
|
nsVoidArray &pt = n->pointers_to;
|
|
for (PRInt32 i = pt.Count() - 1; i >= 0; --i) {
|
|
if (!NS_STATIC_CAST(AllocationNode*, pt[i])->reached) {
|
|
stack.AppendElement(pt[i]);
|
|
}
|
|
}
|
|
}
|
|
} while (stack.Count() > 0);
|
|
}
|
|
}
|
|
|
|
// Sort the nodes by their DFS index, in reverse, so that the first
|
|
// node is guaranteed to be in a root SCC.
|
|
AllocationNode **sorted_nodes = new AllocationNode*[count];
|
|
if (!sorted_nodes) {
|
|
fprintf(stderr, "%s: Out of memory.\n", argv[0]);
|
|
return 1;
|
|
}
|
|
|
|
{
|
|
for (size_t i = 0; i < count; ++i) {
|
|
sorted_nodes[i] = nodes + i;
|
|
}
|
|
NS_QuickSort(sorted_nodes, count, sizeof(AllocationNode*),
|
|
sort_by_reverse_index, 0);
|
|
}
|
|
|
|
// Put the nodes into their strongly-connected components.
|
|
PRUint32 num_sccs = 0;
|
|
{
|
|
for (size_t i = 0; i < count; ++i) {
|
|
nodes[i].reached = PR_FALSE;
|
|
}
|
|
nsVoidArray stack;
|
|
for (AllocationNode **sn = sorted_nodes,
|
|
**sn_end = sorted_nodes + count; sn != sn_end; ++sn) {
|
|
if ((*sn)->reached) {
|
|
continue;
|
|
}
|
|
|
|
// We found a new strongly connected index.
|
|
stack.AppendElement(*sn);
|
|
do {
|
|
PRUint32 pos = stack.Count() - 1;
|
|
AllocationNode *n =
|
|
NS_STATIC_CAST(AllocationNode*, stack[pos]);
|
|
stack.RemoveElementAt(pos);
|
|
|
|
if (!n->reached) {
|
|
n->reached = PR_TRUE;
|
|
n->index = num_sccs;
|
|
stack.AppendElements(n->pointers_from);
|
|
}
|
|
} while (stack.Count() > 0);
|
|
++num_sccs;
|
|
}
|
|
}
|
|
|
|
// Identify which nodes are leak roots by using DFS, and watching
|
|
// for component transitions.
|
|
PRUint32 num_root_nodes = count;
|
|
{
|
|
for (size_t i = 0; i < count; ++i) {
|
|
nodes[i].is_root = PR_TRUE;
|
|
}
|
|
|
|
nsVoidArray stack;
|
|
for (AllocationNode *n = nodes, *n_end = nodes+count; n != n_end; ++n) {
|
|
if (!n->is_root) {
|
|
continue;
|
|
}
|
|
|
|
// Loop through pointers_to, and add any that are in a
|
|
// different SCC to stack:
|
|
for (int i = n->pointers_to.Count() - 1; i >= 0; --i) {
|
|
AllocationNode *target =
|
|
NS_STATIC_CAST(AllocationNode*, n->pointers_to[i]);
|
|
if (n->index != target->index) {
|
|
stack.AppendElement(target);
|
|
}
|
|
}
|
|
|
|
while (stack.Count() > 0) {
|
|
PRUint32 pos = stack.Count() - 1;
|
|
AllocationNode *n =
|
|
NS_STATIC_CAST(AllocationNode*, stack[pos]);
|
|
stack.RemoveElementAt(pos);
|
|
|
|
if (n->is_root) {
|
|
n->is_root = PR_FALSE;
|
|
--num_root_nodes;
|
|
stack.AppendElements(n->pointers_to);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
// Sort the nodes by their SCC index.
|
|
NS_QuickSort(sorted_nodes, count, sizeof(AllocationNode*),
|
|
sort_by_index, 0);
|
|
|
|
// Print output.
|
|
{
|
|
printf("<!DOCTYPE html PUBLIC \"-//W3C//DTD HTML 4.01//EN\">\n"
|
|
"<html>\n"
|
|
"<head>\n"
|
|
"<title>Leak analysis</title>\n"
|
|
"<style type=\"text/css\">\n"
|
|
" .root { background: white; color: black; }\n"
|
|
" .nonroot { background: #ccc; color: black; }\n"
|
|
"</style>\n"
|
|
"</head>\n");
|
|
printf("<body>\n\n"
|
|
"<p>Generated %d entries (%d in root SCCs) and %d SCCs.</p>\n\n",
|
|
count, num_root_nodes, num_sccs);
|
|
|
|
for (size_t i = 0; i < count; ++i) {
|
|
nodes[i].reached = PR_FALSE;
|
|
}
|
|
|
|
// Loop over the sorted nodes twice, first printing the roots
|
|
// and then the non-roots.
|
|
for (PRBool root_type = PR_TRUE;
|
|
root_type == PR_TRUE || root_type == PR_FALSE; --root_type) {
|
|
if (root_type) {
|
|
printf("\n\n"
|
|
"<div class=\"root\">\n"
|
|
"<h1 id=\"root\">Root components</h1>\n");
|
|
} else {
|
|
printf("\n\n"
|
|
"<div class=\"nonroot\">\n"
|
|
"<h1 id=\"nonroot\">Non-root components</h1>\n");
|
|
}
|
|
PRUint32 component = (PRUint32)-1;
|
|
PRBool one_object_component;
|
|
for (const AllocationNode *const* sn = sorted_nodes,
|
|
*const* sn_end = sorted_nodes + count;
|
|
sn != sn_end; ++sn) {
|
|
const AllocationNode *n = *sn;
|
|
if (n->is_root != root_type)
|
|
continue;
|
|
const ADLog::Entry *e = n->entry;
|
|
|
|
if (n->index != component) {
|
|
component = n->index;
|
|
one_object_component =
|
|
sn + 1 == sn_end || (*(sn+1))->index != component;
|
|
if (!one_object_component)
|
|
printf("\n\n<h2 id=\"c%d\">Component %d</h2>\n",
|
|
component, component);
|
|
}
|
|
|
|
if (one_object_component) {
|
|
printf("\n\n<div id=\"c%d\">\n", component);
|
|
printf("<h2 id=\"o%d\">Object %d "
|
|
"(single-object component %d)</h2>\n",
|
|
n-nodes, n-nodes, component);
|
|
} else {
|
|
printf("\n\n<h3 id=\"o%d\">Object %d</h3>\n",
|
|
n-nodes, n-nodes);
|
|
}
|
|
printf("<pre>\n");
|
|
printf("%p <%s> (%d)\n",
|
|
e->address, e->type, e->datasize);
|
|
for (size_t d = 0; d < e->datasize;
|
|
d += sizeof(ADLog::Pointer)) {
|
|
AllocationNode *target = (AllocationNode*)
|
|
PL_HashTableLookup(memory_map, *(void**)(e->data + d));
|
|
if (target) {
|
|
printf(" <a href=\"#o%d\">0x%08X</a> <%s>",
|
|
target - nodes,
|
|
*(unsigned int*)(e->data + d),
|
|
target->entry->type);
|
|
if (target->index != n->index) {
|
|
printf(", component %d", target->index);
|
|
}
|
|
printf("\n");
|
|
} else {
|
|
printf(" 0x%08X\n",
|
|
*(unsigned int*)(e->data + d));
|
|
}
|
|
}
|
|
|
|
if (n->pointers_from.Count()) {
|
|
printf("\nPointers from:\n");
|
|
for (PRUint32 i = 0, i_end = n->pointers_from.Count();
|
|
i != i_end; ++i) {
|
|
AllocationNode *t = NS_STATIC_CAST(AllocationNode*,
|
|
n->pointers_from[i]);
|
|
const ADLog::Entry *te = t->entry;
|
|
printf(" <a href=\"#o%d\">%s</a> (Object %d, ",
|
|
t - nodes, te->type, t - nodes);
|
|
if (t->index != n->index) {
|
|
printf("component %d, ", t->index);
|
|
}
|
|
if (t == n) {
|
|
printf("self)\n");
|
|
} else {
|
|
printf("%p)\n", te->address);
|
|
}
|
|
}
|
|
}
|
|
|
|
print_escaped(stdout, e->allocation_stack);
|
|
|
|
printf("</pre>\n");
|
|
if (one_object_component) {
|
|
printf("</div>\n");
|
|
}
|
|
}
|
|
printf("</div>\n");
|
|
}
|
|
printf("</body>\n"
|
|
"</html>\n");
|
|
}
|
|
|
|
delete [] sorted_nodes;
|
|
delete [] nodes;
|
|
|
|
return 0;
|
|
}
|