mochi-hpc / mochi-ssg

Scalable Service Groups (SSG), a group membership service for Mochi
Other
1 stars 1 forks source link

ssg_group_join segfault #64

Closed mdorier closed 2 years ago

mdorier commented 2 years ago

ssg_group_join segfaults if SWIM has been disabled instead of returning a proper error (an error does show up on the process that the "joiner" is trying to contact).

Here is the backtrace from gdb:

(gdb) bt
#0  0x00007ffff79fc769 in __GI___libc_free (mem=0x7ffff1851010) at malloc.c:3288
#1  0x00007ffff7918a41 in ssg_group_join_send (g_id=g_id@entry=13614397414369239985, target_addr=<optimized out>, mid_state=<optimized out>,
    group_size=group_size@entry=0x7fffffffc2ec, group_config=group_config@entry=0x7fffffffc320, view_buf=view_buf@entry=0x7fffffffc308) at src/ssg-rpc.c:250
#2  0x00007ffff7913c78 in ssg_group_join (mid=mid@entry=0x5555559ed450, group_id=<optimized out>,
    update_cb=update_cb@entry=0x7ffff7ee6ad0 <bedrock::SSGUpdateHandler::membershipUpdate(void*, unsigned long, ssg_member_update_type)>, update_cb_dat=0x5555555871c0)
    at src/ssg.c:910
shanedsnyder commented 2 years ago

I was able to reproduce this easily, I'll see if I can get a quick fix committed.

shanedsnyder commented 2 years ago

This should be fixed on main for now, thanks again for the report.