I fired up syslog-ng on my alpha OpenBSD 2.6 loghost and got this. I have no what to make of it. I'm running syslog-ng 1.4.3 and libol.0.2.16. Is this a problem, perhaps similar to ger@tinet.ie's post earlier this month?
pid 4179 (syslog-ng): unaligned access: va=0x140017a5c pc=0x120013be4 ra=0x120013be4 op=stq pid 4179 (syslog-ng): unaligned access: va=0x140017a4c pc=0x120013a38 ra=0x120013a38 op=stq pid 4179 (syslog-ng): unaligned access: va=0x140017a54 pc=0x120013a40 ra=0x120013a40 op=stq pid 4179 (syslog-ng): unaligned access: va=0x140017a4c pc=0x120013a48 ra=0x120013a48 op=ldq pid 4179 (syslog-ng): unaligned access: va=0x140017adc pc=0x120013be4 ra=0x120013be4 op=stq pid 4179 (syslog-ng): unaligned access: va=0x140017acc pc=0x120013a38 ra=0x120013a38 op=stq pid 4179 (syslog-ng): unaligned access: va=0x140017ad4 pc=0x120013a40 ra=0x120013a40 op=stq pid 4179 (syslog-ng): unaligned access: va=0x140017acc pc=0x120013a48 ra=0x120013a48 op=ldq
It's the same... Somewhere in syslog-ng there's an unaligned dword access, but without a core or something it's quite difficult to trace. Try to find out where pc= points. For example: gdb syslog-ng disassemble $pc it'll result in something like this: Dump of assembler code for function main: 0x8049f94 <main>: push %ebp 0x8049f95 <main+1>: mov %esp,%ebp 0x8049f97 <main+3>: sub $0x9c,%esp 0x8049f9d <main+9>: push %edi 0x8049f9e <main+10>: push %esi 0x8049f9f <main+11>: push %ebx 0x8049fa0 <main+12>: movl $0x1,0xfffffffc(%ebp) 0x8049fa7 <main+19>: add $0xfffffff4,%esp 0x8049faa <main+22>: push $0x805d2c0 0x8049faf <main+27>: call 0x80567ec <ol_object_alloc> 0x8049fb4 <main+32>: add $0x10,%esp 0x8049fb7 <main+35>: mov %eax,0xfffffff4(%ebp) 0x8049fba <main+38>: lea 0xffffff74(%ebp),%eax Of course listing Alpha instructions instead of x86. If you can locate the function for me, I'll try to find the unaligned access. -- Bazsi PGP info: KeyID 9AF8D0A9 Fingerprint CD27 CFB0 802C 0944 9CFD 804E C82C 8EB1 url: http://www.balabit.hu/pgpkey.txt