Re: [Help-gnunet] [BUG] gnunetd crash (0.8.0pre0)

[Christian Grothoff]

You wouldn't happen to have a line like

dstore = sqstore_mysql

in your gnunetd.conf by any chance?  That would be quite, well, fatal -- not 
to mention perfectly explain this...

Christian

On Saturday 12 April 2008 06:45:15 am David Kuehling wrote:
> Hi,
>
> after noticing that gnunetd 0.8.0pre0 uptime did not reach more than a
> few days, I ran it from gdb this time.  After almost a week of uptime it
> crashed again:
>
> Program received signal SIGSEGV, Segmentation fault.
> [Switching to Thread -1314849872 (LWP 25232)]
> put (key=0xb03f7144, value=0x0) at mysql.c:788
> 788       if (((ntohl (value->size) < sizeof (GNUNET_DatastoreValue))) ||
>
> (gdb) bt
> #0  put (key=0xb03f7144, value=0x0) at mysql.c:788
> #1  0xb4543a1c in handleGet (sender=0xb090548c, msg=0xb03f7134)
>     at routing.c:525
> #2  0xb7f6dd59 in GNUNET_CORE_p2p_inject_message (sender=0xb090548c,
>     msg=0xb03f7134 "", size=80, wasEncrypted=1, session=0xb0907e58)
>     at handler.c:499
> #3  0xb7f6e011 in threadMain (cls=0xbf940508) at handler.c:614
> #4  0xb7e380bd in start_thread () from /lib/tls/libpthread.so.0
> #5  0xb7dcd01e in clone () from /lib/tls/libc.so.6
>
> (gdb) frame 1
> #1  0xb4543a1c in handleGet (sender=0xb090548c, msg=0xb03f7134)
>     at routing.c:525
> 525       total = dstore->get (&get->key, ntohl (get->type), &routeResult,
> NULL); (gdb) print *dstore
> $5 = {put = 0xb454a550 <getSize>, get = 0xb454c980 <put>}
>
> print *(GNUNET_SQstore_ServiceAPI*)dstore
> $8 = {getSize = 0xb454a550 <getSize>, put = 0xb454c980 <put>,
>   get = 0xb454bde0 <get>, update = 0xb454a660 <update>,
>   iterateLowPriority = 0xb454bda0 <iterateLowPriority>,
>   iterateNonAnonymous = 0xb454bd60 <iterateNonAnonymous>,
>   iterateExpirationTime = 0xb454bd20 <iterateExpirationTime>,
>   iterateMigrationOrder = 0xb454bce0 <iterateMigrationOrder>,
>   iterateAllNow = 0xb454bca0 <iterateAllNow>, drop = 0xb454a340 <drop>}
>
>
> I don't quite understand what happened here.  If the debugger is right,
> then the 'dstore' pointer got overridden by a pointer to a
> GNUNET_SQstore_ServiceAPI object.  But maybe that's just artefacts from
> debugging a program compiled with -O2?  Disassembling shows, that the
> crash really occured on a null-pointer dereference:
>
> (gdb) info reg
> eax            0x0      0
> ecx            0x0      0
> edx            0xb03f7144       -1338019516
> ebx            0xb454f53c       -1269500612
> esp            0xb1a0ee20       0xb1a0ee20
> ebp            0xb1a0f0f8       0xb1a0f0f8
> esi            0xb090548c       -1332718452
> edi            0x1      1
> eip            0xb454c99a       0xb454c99a <put+26>
>
> (gdb) disas $pc
> Dump of assembler code for function put:
> 0xb454c980 <put+0>:     push   %ebp
> 0xb454c981 <put+1>:     mov    %esp,%ebp
> 0xb454c983 <put+3>:     push   %edi
> 0xb454c984 <put+4>:     push   %esi
> 0xb454c985 <put+5>:     push   %ebx
> 0xb454c986 <put+6>:     sub    $0x2cc,%esp
> 0xb454c98c <put+12>:    mov    0xc(%ebp),%eax
> 0xb454c98f <put+15>:    call   0xb4548465 <__i686.get_pc_thunk.bx>
> 0xb454c994 <put+20>:    add    $0x2ba8,%ebx
> 0xb454c99a <put+26>:    mov    (%eax),%edx
> 0xb454c99c <put+28>:    bswap  %edx
> ...
>
>
> David