Project

General

Profile

Actions

Bug #6257

open

pcap: hang at shutdown for some interfaces

Added by Victor Julien 11 months ago. Updated 11 months ago.

Status:
New
Priority:
Normal
Assignee:
Target version:
Affected Versions:
Effort:
Difficulty:
Label:

Description

On an inactive openvpn dev:

# ip l show tun0
10: tun0: <POINTOPOINT,MULTICAST,NOARP,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UNKNOWN mode DEFAULT group default qlen 100
    link/none

Getting Suricata to shutdown fails:

Info: runmodes: Using 2 live device(s). [RunModeSetLiveCaptureAutoFp:util-runmodes.c:143]
Info: pcap: tun0: running in 'auto' checksum mode. Detection of interface state will require 1000 packets [ReceivePcapThreadInit:source-pcap.c:522]
Info: ioctl: tun0: MTU 1500 [GetIfaceMTU:util-ioctl.c:110]
Info: pcap: tun0: snaplen set to 1524 [PcapOpenInterface:source-pcap.c:248]
Perf: ioctl: tun0: NIC offloading: RX unset TX unset [GetIfaceOffloadingLinux:util-ioctl.c:340]
Perf: ioctl: tun0: NIC offloading: SG: unset, GRO: unset, LRO: unset, TSO: unset, GSO: unset [GetIfaceOffloadingLinux:util-ioctl.c:385]
Info: pcap: tun1: running in 'auto' checksum mode. Detection of interface state will require 1000 packets [ReceivePcapThreadInit:source-pcap.c:522]
Info: ioctl: tun1: MTU 1500 [GetIfaceMTU:util-ioctl.c:110]
Info: pcap: tun1: snaplen set to 1524 [PcapOpenInterface:source-pcap.c:248]
Perf: ioctl: tun1: NIC offloading: RX unset TX unset [GetIfaceOffloadingLinux:util-ioctl.c:340]
Perf: ioctl: tun1: NIC offloading: SG: unset, GRO: unset, LRO: unset, TSO: unset, GSO: unset [GetIfaceOffloadingLinux:util-ioctl.c:385]
...
Notice: suricata: Signature(s) loaded, Detect thread(s) activated. [PostRunStartedDetectSetup:suricata.c:2540]
Notice: suricata: Signal Received.  Stopping engine. [SuricataMainLoop:suricata.c:2828]
Error: threads: Engine unable to disable detect thread - "RX#01-tun0". Killing engine [TmThreadDisableReceiveThreads:tm-threads.c:1359]

gdb shows:

(gdb) bt
#0  0x00007ffff631199f in __GI___poll (fds=0x7fffe05dc210, nfds=1, timeout=-1) at ../sysdeps/unix/sysv/linux/poll.c:29
#1  0x00007ffff75c65ca in __interceptor_poll (fds=fds@entry=0x7fffe05dc210, nfds=nfds@entry=1, timeout=-1) at ../../../../src/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc:3844
#2  0x00007ffff6455c4a in poll (__timeout=<optimized out>, __nfds=1, __fds=0x7fffe05dc210) at /usr/include/x86_64-linux-gnu/bits/poll2.h:46
#3  pcap_wait_for_frames_mmap (handle=handle@entry=0x6180000fc080) at ./pcap-linux.c:4869
#4  0x00007ffff6455e75 in pcap_read_linux_mmap_v3 (handle=0x6180000fc080, max_packets=64, callback=0x555555b4cd36 <PcapCallbackLoop>, user=0x60e000e86fc0 "\200\300\017") at ./pcap-linux.c:5367
#5  0x0000555555b4d827 in ReceivePcapLoop (tv=0x6120000880c0, data=0x60e000e86fc0, slot=0x6060001067c0) at source-pcap.c:424
#6  0x0000555555832724 in TmThreadsSlotPktAcqLoop (td=0x6120000880c0) at tm-threads.c:316
#7  0x00007ffff70e7609 in start_thread (arg=<optimized out>) at pthread_create.c:477
#8  0x00007ffff631e133 in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95

A big more debugging shows that pcap_dispatch is called, but we never return from it.

If the pcap handle is set to nonblocking with pcap_setnonblocking things do work, but the rest of the code isn't ready for this (100% cpu spin).

Same happens on a dummy device, so it should be easy to test.

It can be seen that pcap_breakloop is called on the handle, but it is not resulting in breaking the loop.


Related issues 1 (0 open1 closed)

Related to Suricata - Bug #2751: Engine unable to disable detect thread, Killing engine. (in libpcap mode)Closedjingyu YANGActions
Actions #1

Updated by Victor Julien 11 months ago

  • Related to Bug #2751: Engine unable to disable detect thread, Killing engine. (in libpcap mode) added
Actions #2

Updated by Victor Julien 11 months ago

  • Description updated (diff)
Actions #3

Updated by Victor Julien 11 months ago

Btw output above shows 2 interfaces, but problem happens as well with just one.

Actions

Also available in: Atom PDF