summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorJan Friesse <jfriesse@redhat.com>2017-03-30 16:33:51 +0200
committerJan Friesse <jfriesse@redhat.com>2017-06-16 14:44:34 +0200
commit005a4afa6b0e710e11b47b11cfc152b028c596fc (patch)
tree72ce6f80073decd8c7c7b9648a7fb6f353b6dfb2
parent37bc343b58914bee30455ffa70009e20ac9e909c (diff)
res_corosync: Change thread stack size
In Corosync 2.x libraries were changed to use LibQB IPC. Sadly LibQB IPC doesn't support copy-free access to received buffer, so Corosync libraries were rewritten to use stack as buffer. Mostly the needed stack size is quite small, but for all *_dispatch functions, 1MiB is needed. Asterisk function ast_pthread_create_background set stack size for new thread to much smaller AST_BACKGROUND_STACKSIZE (~500KiB). This results in Asterisk crash when running with Corosync 2.x. Patch solves this issue by creating it's own version of ast_pthread_create_background which sets stack size to much higher value (actually it's AST_BACKGROUND_STACKSIZE + 3MiB). Another problem may appear when "corosync show members" netconsole command is executed. It is also executed in thread and also has only 500KiB stack size. Sadly it calls corosync_cfg_get_node_addrs which again needs at least 1MiB stack. Solution is to use HAVE_COROSYNC_CFG_STATE_TRACK as a discriminator between Corosync 1.x and 2.x. If 1.x is found, nothing changes. If 2.x is found, NodeID is displayed instead of IP address. ASTERISK-25370 #close Reported by: mdu113 Change-Id: Id95b0d21ab6e708e7d74ad8786c587211676fa08
-rw-r--r--res/res_corosync.c29
1 files changed, 24 insertions, 5 deletions
diff --git a/res/res_corosync.c b/res/res_corosync.c
index 6bbbc34b9..ce94e4151 100644
--- a/res/res_corosync.c
+++ b/res/res_corosync.c
@@ -79,6 +79,15 @@ struct corosync_node {
struct ast_sockaddr addr;
};
+/*! \brief Corosync ipc dispatch/request and reply size */
+#define COROSYNC_IPC_BUFFER_SIZE (8192 * 128)
+
+/*! \brief Version of pthread_create to ensure stack is large enough */
+#define corosync_pthread_create_background(a, b, c, d) \
+ ast_pthread_create_stack(a, b, c, d, \
+ (AST_BACKGROUND_STACKSIZE + (3 * COROSYNC_IPC_BUFFER_SIZE)), \
+ __FILE__, __FUNCTION__, __LINE__, #c)
+
static struct corosync_node *corosync_node_alloc(struct ast_event *event)
{
struct corosync_node *node;
@@ -810,10 +819,21 @@ static char *corosync_show_members(struct ast_cli_entry *e, int cmd, struct ast_
for (i = 1, cs_err = cpg_iteration_next(cpg_iter, &cpg_desc);
cs_err == CS_OK;
cs_err = cpg_iteration_next(cpg_iter, &cpg_desc), i++) {
+#ifdef HAVE_COROSYNC_CFG_STATE_TRACK
corosync_cfg_node_address_t addrs[8];
int num_addrs = 0;
unsigned int j;
+#endif
+
+ ast_cli(a->fd, "=== Node %u\n", i);
+ ast_cli(a->fd, "=== --> Group: %s\n", cpg_desc.group.value);
+#ifdef HAVE_COROSYNC_CFG_STATE_TRACK
+ /*
+ * Corosync 2.x cfg lib needs to allocate 1M on stack after calling
+ * corosync_cfg_get_node_addrs. netconsole thread has allocated only 0.5M
+ * resulting in crash.
+ */
cs_err = corosync_cfg_get_node_addrs(cfg_handle, cpg_desc.nodeid,
ARRAY_LEN(addrs), &num_addrs, addrs);
if (cs_err != CS_OK) {
@@ -821,9 +841,6 @@ static char *corosync_show_members(struct ast_cli_entry *e, int cmd, struct ast_
continue;
}
- ast_cli(a->fd, "=== Node %u\n", i);
- ast_cli(a->fd, "=== --> Group: %s\n", cpg_desc.group.value);
-
for (j = 0; j < num_addrs; j++) {
struct sockaddr *sa = (struct sockaddr *) addrs[j].address;
size_t sa_len = (size_t) addrs[j].address_length;
@@ -833,7 +850,9 @@ static char *corosync_show_members(struct ast_cli_entry *e, int cmd, struct ast_
ast_cli(a->fd, "=== --> Address %u: %s\n", j + 1, buf);
}
-
+#else
+ ast_cli(a->fd, "=== --> Nodeid: %"PRIu32"\n", cpg_desc.nodeid);
+#endif
}
ast_cli(a->fd, "===\n"
@@ -1159,7 +1178,7 @@ static int load_module(void)
goto failed;
}
- if (ast_pthread_create_background(&dispatch_thread.id, NULL,
+ if (corosync_pthread_create_background(&dispatch_thread.id, NULL,
dispatch_thread_handler, NULL)) {
ast_log(LOG_ERROR, "Error starting CPG dispatch thread.\n");
goto failed;