fsotdld crashes on changing networkinterface

[jake42]

fsotdld crashes when changing networkinterfaces, e.g ppp0 goes down. Here is a backtrace with gdb with SHR (e.g. StagingTest 140) on gta04. Also crashes with gta02, thougn no backtrace as it got too slow with gdb.

This is with srv rev: 12843f0
Issue in SHR: http://shr-project.org/trac/ticket/2110

#0  0xb6d48764 in __GI_raise (sig=sig@entry=6) at ../nptl/sysdeps/unix/sysv/linux/raise.c:56
#1  0xb6d4c23c in __GI_abort () at abort.c:89
#2  0xb6d80320 in __libc_message (do_abort=<optimized out>, fmt=0xb6e3686c "*** Error in `%s': %s: 0x%s ***\n") at ../sysdeps/posix/libc_fatal.c:175
#3  0xb6d8ad28 in malloc_printerr (action=3, str=0xb6e36980 "munmap_chunk(): invalid pointer", ptr=<optimized out>) at malloc.c:4895
#4  0xb6e9bc00 in g_free (mem=<optimized out>) at gmem.c:252
#5  0xb6e83754 in g_hash_table_remove_all_nodes (hash_table=0xb60024a8, notify=-1226610580) at ghash.c:500
#6  0xb6e84128 in g_hash_table_unref (hash_table=0xb60024a8) at ghash.c:1026
#7  0xb6fafbd4 in fso_framework_base_netlink_notifier_handleMessage (self=0x26738, type=<optimized out>, obj=<optimized out>) at netlinknotifier.c:532
#8  0xb6b3d286 in parse_cb (obj=<optimized out>, p=<optimized out>) at msg.c:721
#9  0xb6b53ab8 in addr_msg_parser (ops=<optimized out>, who=<optimized out>, nlh=<optimized out>, pp=0xbefff5a0) at route/addr.c:316
#10 0xb6b3b422 in nl_cache_parse (ops=ops@entry=0xb6b7119c <rtnl_addr_ops>, who=who@entry=0x0, nlh=0x35500, params=params@entry=0xbefff5a0) at cache.c:905
#11 0xb6b3e31e in nl_msg_parse (msg=msg@entry=0x35340, cb=<optimized out>, arg=arg@entry=0x13120) at msg.c:744
#12 0xb6faef5a in fso_framework_base_netlink_notifier_handleNetlinkMessage (self=0x26738, msg=0x35340) at netlinknotifier.c:387
#13 0xb6b3eed0 in nl_cb_call (msg=0x35340, type=0, cb=0x2ab00) at ../include/netlink-private/netlink.h:129
#14 recvmsgs (cb=0x2ab00, sk=0x2ab78) at nl.c:904
#15 nl_recvmsgs_report (sk=0x2ab78, cb=0x2ab00) at nl.c:955
#16 0xb6b3f126 in nl_recvmsgs (sk=<optimized out>, cb=<optimized out>) at nl.c:979
#17 0xb6b3f132 in nl_recvmsgs_default (sk=<optimized out>) at nl.c:993
#18 0xb6faee56 in fso_framework_base_netlink_notifier_onActionFromSocket (self=0x26738, source=0x2ac28, condition=<optimized out>) at netlinknotifier.c:325
#19 0xb6eda988 in g_io_unix_dispatch (source=0x31dc8, callback=0xb6faef11 <_fso_framework_base_netlink_notifier_onActionFromSocket_gio_func>, user_data=0x26738) at giounix.c:167
#20 0xb6e95d50 in g_main_dispatch (context=0x1de40) at gmain.c:3054
#21 g_main_context_dispatch (context=context@entry=0x1de40) at gmain.c:3630
#22 0xb6e960dc in g_main_context_iterate (context=0x1de40, block=block@entry=1, dispatch=dispatch@entry=1, self=<optimized out>) at gmain.c:3701
#23 0xb6e96598 in g_main_loop_run (loop=0x2f5a8) at gmain.c:3895
#24 0x00008df0 in _vala_main (args=0xbefff8a4, args_length1=1) at main.c:215
#25 0xb6d3327c in __libc_start_main (main=0xbefff8a4, argc=-1226522624, argv=0xb6d3327c <__libc_start_main+276>, init=<optimized out>, fini=0x8e7c <__libc_csu_fini>, rtld_fini=0xb6fe7c08 <_dl_fini>, stack_end=0xbefff8a4) at libc-start.c:269
#26 0x00008ae0 in _start () at ../ports/sysdeps/arm/start.S:124

[mickeyl]

This is quite strange. The crash actually occurs in libfsobasics, namely
protected void handleMessage( uint16 type, Netlink.Object obj )
From the look of it I would expect that all running fsodaemons should actually crash in that case, not just fsotdld. fsotdld not even registers a netlink notification handler with libfsobasics. Perhaps there is some new/unknown netlink message coming in due to the change in network interfaces. I can reproduce this quite easily and will take a look.

[mickeyl]

This is a problem in vala's libnl-3.0.vapi. There is at least one unowned missing, i.e. line 46 states:
public string get_label ();. The underlying API returns a char*, so it needs to be bound as unowned. I will fix this directly in vala and also commit a workaround for us.