Fail if interface specified is actually a file or directory

[fklassen]

As reported...

The problem, in a nutshell:

# tcpreplay -t -i p1p1 big.pcap

Fatal Error:
%s is not a valid Tcpreplay character device

Two issues.
(1) This seems to have intended to be routed through a printf function,
but doesn¹t seem to be (see missing args below).
(2) It doesn¹t work, whereas old tcpreplay 3.4.4 works fine.


This problem seems independent of the pcap file specified.
It has been reproduced on a different CentOS 6.5 system, but with 똪o¹,
and with a more recent kernel.

Yes, I am root (uid=0). No, SELinux is not running.


Any thoughts? Details follow ...

Thanks for any help you can offer.

kind regards,

Seth

. . . .

CentOS 6.3, 2.6.32-279.19.1.el6.x86_64 kernel.

tcpreplay version: 4.0.3 (build git:v4.0.3)
Copyright 2013-2014 by Fred Klassen <tcpreplay at appneta dot com> -
AppNeta Inc.
Copyright 2000-2012 by Aaron Turner <aturner at synfin dot net>
The entire Tcpreplay Suite is licensed under the GPLv3
Cache file supported: 04
Not compiled with libdnet.
Compiled against libpcap: 1.4.0
64 bit packet counters: enabled
Verbose printing via tcpdump: enabled
Packet editing: disabled
Fragroute engine: disabled
Injection method: PF_PACKET send()

driver: ixgbe
version: 3.12.6
firmware-version: 0x61bd0001
bus-info: 0000:01:00.0

# tcpreplay 땗istnics
Available network interfaces:
p1p1
p1p2
em1
em2
any

#0  0x000000370e6acea0 in _exit () from /lib64/libc.so.6
#1  0x000000370e635d52 in exit () from /lib64/libc.so.6
#2  0x000000000040ab7c in sendpacket_open (device=0x832a50 "p1p1",
    errbuf=0x7fffffffdcb0 "", direction=TCPR_DIR_C2S,
    sendpacket_type=<value optimized out>) at sendpacket.c:555
#3  0x0000000000407258 in tcpreplay_post_args (ctx=0x827010,
    argc=<value optimized out>) at tcpreplay_api.c:289
#4  0x00000000004054f5 in main (argc=1, argv=0x7fffffffe610) at
tcpreplay.c:73

#2  0x000000000040ab7c in sendpacket_open (device=0x832a50 "p1p1",
    errbuf=0x7fffffffdcb0 "", direction=TCPR_DIR_C2S,
    sendpacket_type=<value optimized out>) at sendpacket.c:555
555             err(1, "%s is not a valid Tcpreplay character device²);

(gdb) list
550         if (((sdata.st_mode & S_IFMT) == S_IFCHR)) {
551
552             sp = sendpacket_open_khial(device, errbuf);
553
554         } else {
555             err(1, "%s is not a valid Tcpreplay character device");
556         }
557     } else {
558 #ifdef HAVE_NETMAP
559         if (sendpacket_type == SP_TYPE_NETMAP)
(gdb) p /x sdata.st_mode
$2 = 0x41ed

Also ...

Based on this comment and running strace on tcpreplay and stat, I am 90%
sure that the problem is that some other user created a directory called
Œp1p1¹. It didn¹t occur to me to look for this until you said that.
You can probably reproduce my issue by doing a Œmkdir <ifacename>¹ and
then trying to invoke 'tcpreplay -i <ifacename>¹ from that directory.

When I run tcpreplay from /tmp, the problem goes away. So you could say I
have a workaround.

[fklassen]

Added following commit...
e590431

Unable to reproduce on my ixgbe Ubuntu machine. Notice the stack trace during --listnics with assertion on ebuf. Suspect stack overflow, so moved ebuf from stack to heap.

Will close for now and reopen if this bug is recreated.