BIRD 2.0.4 segfaulting on ARM

lorenz at irmhil.de lorenz at irmhil.de
Fri Apr 26 16:49:21 CEST 2019


Hello again,

I narrowed the bug down to the kernel protocol. The config for debugging
is as follows:


--- snip ---

log syslog all;
router id 10.33.0.0;

protocol device {
        scan time 15;
}


filter myrange {
        if net ~ fd22:9c28:6cf6::/48 then accept;
        reject;
};

protocol kernel {
#        scan time 10;
        ipv6 {
              export all;
        };
#        learn;
}


protocol ospf v3 MyOSPF {
        ipv6 {  
                import filter myrange;
                export filter myrange;
        };
        area 0 {
                networks {
                        fd22:9c28:6cf6::/48;
                };


                interface "eth0.1000" {
                        hello 2;
                        dead  5;
                        cost 10;
                };
        };
}

--- snap ---


If I remove the comment on "scan" or "learn" in the "protocol kernel"
section, I get the segfault.


Thanks for the code so far,

Lorenz



Am 26.04.19 um 13:39 schrieb Maria Jan Matejka:
> On 4/26/19 1:08 PM, lorenz at irmhil.de wrote:
>> Hello,
>>
>> after a "make clean", "./configure" and "make" I got this compile-time warning:
>>
>> --- snip ---
>>
>> sysdep/unix/io.c: In function ‘times_init’:
>> sysdep/unix/io.c:135:45: warning: comparison is always false due to limited range of data type [-Wtype-limits]
>>    if ((ts.tv_sec < 0) || (((s64) ts.tv_sec) > ((s64) 1 << 40)))
>>                                              ^
>> --- snap ---
>>
>>
>> But unfortunately the segmentation fault is still there. Is there anything I can do?
> Thank you for investigation; anyway I have no clue what may be happening.
> I'll try to install a local QEMU host to simulate this and then I'll return
> to you off-list if I don't happen to find any problem that may be related to this.
>
> Sadly, this seems too much to be some strange use-after-free (which may be caused
> by some architecture-specific misbehaviour) which I'm probably unable to debug
> only from core.
>
> Thank you
> Maria
>
>>
>> --- snip ---
>>
>> Core was generated by `bird -c bird.conf'.
>> Program terminated with signal SIGSEGV, Segmentation fault.
>> #0  ea__find (id=1554, e=0x81000601, e at entry=0x0) at nest/rt-attr.c:389
>> 389	      if (e->flags & EALF_BISECT)
>> (gdb) #0  ea__find (id=1554, e=0x81000601, e at entry=0x0) at nest/rt-attr.c:389
>>         a = <optimized out>
>>         l = <optimized out>
>>         r = <optimized out>
>>         m = <optimized out>
>>         a = <optimized out>
>>         l = <optimized out>
>>         r = <optimized out>
>>         m = <optimized out>
>> #1  ea_find (e=e at entry=0xf1ede200, id=id at entry=1554) at nest/rt-attr.c:426
>>         a = <optimized out>
>> #2  0x005267ba in nl_send_route (p=p at entry=0x575748, e=e at entry=0x587170, op=op at entry=1536, dest=<optimized out>, nh=<optimized out>, nh at entry=0x59c664) at sysdep/linux/netlink.c:1269
>>         ea = <optimized out>
>>         net = 0x588174
>>         a = 0x59c630
>>         eattrs = <optimized out>
>>         bufsize = 284
>>         priority = <optimized out>
>>         r = 0xbee4f180
>>         rsize = 312
>>         metrics = {16, 0, 0, 3202675588, 12, 5681780, 5796832, 5800932, 3202675764, 5783224, 5414941, 2147483648, 5797092, 5800932, 3202675764, 0}
>>         ews = {eattrs = 0x58e174, ea = 0x52aafb <krt_got_route+574>, visited = {5585652, 5797092, 5796884, 5585808}}
>> #3  0x005271b8 in nl_add_rte (e=0x587170, p=0x575748) at sysdep/linux/netlink.c:1351
>>         a = 0x59c630
>>         err = 0
>>         a = <optimized out>
>>         err = <optimized out>
>>         nh = <optimized out>
>> #4  krt_replace_rte (p=p at entry=0x575748, n=n at entry=0x588174, new=new at entry=0x587170, old=old at entry=0x5874b0) at sysdep/linux/netlink.c:1387
>>         err = 0
>> #5  0x0052a4d2 in krt_prune (p=0x575748) at sysdep/unix/krt.c:751
>>         verdict = 2
>>         new = <optimized out>
>>         old = 0x5874b0
>>         rt_free = 0x0
>>         fn_ = 0x58817c
>>         ff_ = 0x574694
>>         count_ = <optimized out>
>>         n = <optimized out>
>>         t = 0x574330
>>         t = <optimized out>
>>         fn_ = <optimized out>
>>         ff_ = <optimized out>
>>         count_ = <optimized out>
>>         n = <optimized out>
>>         verdict = <optimized out>
>>         new = <optimized out>
>>         old = <optimized out>
>>         rt_free = <optimized out>
>> #6  krt_scan (t=<optimized out>) at sysdep/unix/krt.c:838
>>         p = 0x575748
>>         q = 0x5758a8
>> #7  0x004f2b86 in timers_fire (loop=loop at entry=0x56b7f0 <main_timeloop>) at lib/timer.c:235
>> ---Type <return> to continue, or q <return> to quit---        base_time = 74049570330
>>         t = <optimized out>
>> #8  0x0052976e in io_loop () at sysdep/unix/io.c:2193
>>         poll_tout = <optimized out>
>>         timeout = <optimized out>
>>         nfds = <optimized out>
>>         events = 1
>>         pout = <optimized out>
>>         t = <optimized out>
>>         s = <optimized out>
>>         n = <optimized out>
>>         fdmax = 256
>>         pfd = 0x585df0
>> #9  0x004dabc6 in main (argc=<optimized out>, argv=<optimized out>) at sysdep/unix/main.c:884
>>         use_uid = <optimized out>
>>         use_gid = <optimized out>
>>         conf = <optimized out>
>> (gdb) quit
>>
>> --- snap ---
>>
>>
>>
>> Am 26.04.19 um 08:09 schrieb lorenz at irmhil.de:
>>> Hello again!
>>>
>>> I'm new to gdb - thank you for your quick advice.
>>>
>>> I ran bird again, about 10 seconds later it segfaulted again and dumped core.
>>>
>>> Looks like some strange metrics?
>>>
>>> I tried running bird on another ARM v7-box (Odroid XU4, nearly the same hardware as the Odroid HC-2) on the same network with a similar config. That bird doesn't crash. Perhaps something happend on compiling or installing bird, I'll try recompiling and reinstalling it.
>>>
>>> Thanks for any support!
>>>
>>> Lorenz
>>>
>>>
>>> The backtrace is:
>>>
>>> --- snip ---
>>>  ...
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://trubka.network.cz/pipermail/bird-users/attachments/20190426/6c6df942/attachment.html>


More information about the Bird-users mailing list