The large stack size in cpg_dispatch, resulting from the message buffer, is causing a segmentation fault in environments with a constrained stack size, namely, when calling into cpg_dispatch from Java using JNI (i.e. for http://github.com/jopereira/jgcs/tree/master/jgcs-corosync). --- lib/cpg.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-) diff --git a/lib/cpg.c b/lib/cpg.c index b96df4e..f3b2bff 100644 --- a/lib/cpg.c +++ b/lib/cpg.c @@ -346,7 +346,7 @@ cs_error_t cpg_dispatch ( struct cpg_ring_id ring_id; uint32_t totem_member_list[CPG_MEMBERS_MAX]; int32_t errno_res; - char dispatch_buf[IPC_DISPATCH_SIZE]; + char* dispatch_buf; error = hdb_error_to_cs (hdb_handle_get (&cpg_handle_t_db, handle, (void *)&cpg_inst)); if (error != CS_OK) { @@ -361,6 +361,7 @@ cs_error_t cpg_dispatch ( timeout = 0; } + dispatch_buf = malloc(IPC_DISPATCH_SIZE); dispatch_data = (struct qb_ipc_response_header *)dispatch_buf; do { errno_res = qb_ipcc_event_recv ( @@ -504,6 +505,7 @@ cs_error_t cpg_dispatch ( error_put: hdb_handle_put (&cpg_handle_t_db, handle); + free(dispatch_buf); return (error); } -- 1.8.1.4 -- Jose Orlando Pereira
Attachment:
smime.p7s
Description: S/MIME cryptographic signature
_______________________________________________ discuss mailing list discuss@xxxxxxxxxxxx http://lists.corosync.org/mailman/listinfo/discuss
