Torsten Landschoff wrote:
>
> On Mon, Mar 19, 2001 at 02:05:34PM +0300, Michael Tokarev wrote:
[]
> > I did some investigations -- yes, machine in question is the only one
> > that has /etc/ethers file with one line in it. If I remove this file,
> > tcpdump works fine. I looked to source -- it crashes in libc's nss
> > stuff trying to resolve all-ones ethernet address. This looks like
> > glibc bug/issue, tcpdump's side is ok.
>
> In that case it is another bug. Can you instruct me how to reproduce it?
It's easy here. Just creating /etc/ethers file is enouth for this.
I use glibc-2.2-12.i[36]86 (RedHat's 7.0 errata), compiled by myself
using again redhat's gcc-2.96-69 (should not matter).
tcpdump-3.6.3# ./tcpdump
Segmentation fault (core dumped)
# _
gdb> bt
#0 0x4022d643 in _nss_files_getntohost_r (addr=0x8096330, result=0xbffff7c0,
buffer=0xbffff3c0 "`\003", buflen=1024, errnop=0x40216fb9)
at nss_files/files-XXX.c:194
#1 0x401ef98a in ether_ntohost (hostname=0xbffff810 "", addr=0x8096330)
at ether_ntoh.c:68
#2 0x805f4ff in init_etherarray () at ./addrtoname.c:697
#3 0x805f5d4 in init_addrtoname (localnet=108736, mask=16777215)
at ./addrtoname.c:757
#4 0x804a1f8 in main (argc=1, argv=0xbffffb04) at ./tcpdump.c:404
#5 0x40116ee1 in __libc_start_main (main=0x8049b94 <main>, argc=1,
ubp_av=0xbffffb04, init=0x8049454 <_init>, fini=0x8077f3c <_fini>,
rtld_fini=0x4000e244 <_dl_fini>, stack_end=0xbffffafc)
at ../sysdeps/generic/libc-start.c:129
gdb> _
addrtoname.c near line 697:
#ifdef HAVE_ETHER_NTOHOST
/* Use yp/nis version of name if available */
if (ether_ntohost(name, (struct ether_addr *)el->addr) == 0) {
tp->e_name = savestr(name);
continue;
}
#endif
el-addr here contains all-ones ether_addr_octet (ff:ff:ff:ff:ff:ff),
name is an automatic char[256] variable.
I don't know what's the ether_ntohost() (guess: it's like gethostbyaddr()
but for ethernet address instead of ip one?), and found no references to
it in man and glibc's info.
But wait... unpacking glibc...done... looking again... Aha --
found it. It's definitely a glibc bug:
gdb> frame 1
#1 0x401ef98a in ether_ntohost (hostname=0xbffff810 "", addr=0x8096330)
at ether_ntoh.c:68
68 status = (*fct) (addr, ðerent, buffer, sizeof buffer);
gdb> frame 0
#0 0x4022d643 in _nss_files_getntohost_r (addr=0x8096330, result=0xbffff7c0,
buffer=0xbffff3c0 "`\003", buflen=1024, errnop=0x40216fb9)
at nss_files/files-XXX.c:194
194 *errnop = ENOENT;
gdb> _
See? _nss_files_getntohost_r expects 5 arguments, but caller (ether_ntohost)
calls it via function pointer with only 4 arguments, omitting last errnop.
In glibc sources:
inet/ether_ntoh.c: ether_ntohost
BTW, before submiting bug report (I never did this before for glibc),
it should be a good idea to check 2.2.1 version (if memory server me
right). Bugzilla database at redhat currently down (well, it gives
"software error" at query page), so I can't check this there either...
Regards,
Michael.
P.S. I set Reply-To to list - i'm on the list.
-
This is the TCPDUMP workers list. It is archived at
http://www.tcpdump.org/lists/workers/index.html
To unsubscribe use mailto:[EMAIL PROTECTED]?body=unsubscribe
