Linux emulation on FreeBSD AMD64

[Valery V.Chikalov]

Roman Divacky wrote:
>>> Just for the record: this make mprottect return success,
>>> but oracle coredumps nevertheless. (Sorry for rhyme :-)
>>>
>> But despite of this is a step forward!
>> Oracle server itself is running!
>> Only sqlplus(oracle control utility) and connected client oracle process 
>> are coredumping. So now I can start oracle server and ever generate new 
>> database.
>>
>> Now I am trying to understand the reasons of this coredumps. They are 
>> occurring rather accidental, and for example I must several times try to 
>> start oracle to achieve the result. Moreover after some time successful 
>> running and serving queries oracle server dump core and stop working too.
>  
> I think its quite obvious...  oracle expects some memory is executable
> and tries to execute something in it but freebsd has non-exec bit set
> on that memory hence core dump...
> 
> do you see any obvious flaw in this thought?

One and half(IMHO).

This occurs accidentally, as I have said I can sometimes start server, 
ask him about something and only after some time find him dead. 
Sometimes I ever cant start it. I think in your scenario it must be 
absolutely deterministic.

There are no AMD64 specific in your reasoning. Its working in i386 mode. 
Do you have some ideas about this?.

>>
>> And "bt" are not looking reasonable.
> 
> try to "disas" memory before the IP when the coredump happened. you
> might disas memory in the middle of instruction so try to increase/decrease
> by 1 the memory location...

I will try.

Thank you.
Valery.

> 
> roman