[utah-c] Re: SIGBUS that doesn't go away even after application restart, only after reboot

[AJ ONeal]

I tried switching it over to use write() but now it leaks memory very quickly.

/dev/shm fills up to 50% and then reports out-of-memory
This isn't too unexpected being that the system may be using that memory.


However, df -h shows that it ISN'T actually using the memory.
unmounting and remounting /dev/shm fixes the problem.

This might be an okay hack-fix, except that it only takes about 5
minutes before it leaks to full.

AJ ONeal



On Mon, Dec 20, 2010 at 2:45 PM, Shawn Willden <shawn at willden.org> wrote:
> On Mon, Dec 20, 2010 at 2:39 PM, AJ ONeal <coolaj86 at gmail.com> wrote:
>>
>> Well thanks for the advice, I'll be posting to the LKML shortly, perhaps
>> with a link to a gist of the source code.
>
> Assuming you are munmap'ing the memory, that would be among my next steps.
> Another option might be to try write()ing the data, rather than mmapping and
> memcpying.  Performance should be identical.
> --
> Shawn
>