Hi,
The Apache version is 2.0.35 and OS is Solaris 8
with prefrok MPM variation. It works fine but gives a
problem only at shutdown. All the child process die
gracefully, but the parent process does not. It hangs
in ' apr_allocator_destroy' for the 'global_pool'
(defined in apr_pools.c) and takes too much of CPU.
The stack trace is :
#0 apr_allocator_destroy (allocator=0x1f5e38) at
apr_pools.c:136
#1 0xff3196a0 in apr_pool_terminate () at
apr_pools.c:519
#2 0x9f038 in destroy_and_exit_process
(process=0x1fa748,
process_exit_value=0) at main.c:253
#3 0x9fcd8 in main (argc=2764944, argv=0x1e4fec) at
main.c:657
The source for apr_allocator_destroy() is ( I have
added some fprintfs)
131 fprintf(stderr, " In apr_allocator_destroy
....\n");
132 for (index = 0; index < MAX_INDEX;
index++) {
133 ref = &allocator->free[index];
134 while ((node = *ref) != NULL) {
135 *ref = node->next;
(gdb)
136 fprintf(stderr, "Node is %p index=
%d\n",node, index);
137 free(node);
138 node = NULL;
139 fprintf(stderr, "After free Node is %p
\n",node);
140 }
141 }
--------------------------------------------------
GDB shows following information-
(gdb) p *allocator
$7 = {max_index = 2, mutex = 0x0, owner = 0x1f5eb0,
free = {0x0, 0x2653b8,
0x2a3ba8, 0x0 <repeats 17 times>}}
(gdb) p node
$8 = (apr_memnode_t *) 0x2653b8
(gdb) p *node
$9 = {next = 0x2633b0, index = 1, first_avail =
0x265960 "", endp = 0x2673b8 ""}
(gdb)p index
$10 = 1
------------------------------------------------
My observation is:
allocator->free[1] and node->next form a cycle in link
list. And hence the while loop does not break, causing
the hang.
I tried to set 'node' to NULL after free, but didn't
help.
Could someone give any clues the problem?
Thanks in advance
Rohan
__________________________________________________
Do you Yahoo!?
Yahoo! Tax Center - File online, calculators, forms, and more
http://tax.yahoo.com
