Request for approval: 6929067: Stack guard pages should be removed when thread is detached
Coleen.Phillimore at Sun.COM
Fri Mar 5 09:25:19 PST 2010
This looks good to me, there are a couple of questions. Can you use
/proc/self/maps instead of getting the tid? Also there's a 'my' instead
of 'by' in the comments. Thank you for adding all the comments.
Keith McGuigan wrote:
> Andrew Haley wrote:
>> On 03/04/2010 10:42 PM, Coleen Phillimore wrote:
>>> I'm also reviewing this change but I also had the same question about
>>> iostream. I don't know the history either. We have a similar function
>>> in os_linux.cpp called find_vma() which uses standard C file io. You
>>> could rewrite get_stack_bounds() if there's a reason to.
>>> The other thing, || *jaxws/jaxws.properties *is checked into a
>>> repository (not the hotspot one).
>>> Also, I believe you need to add comments from the discussion in the
>>> emails in the new os_linux.cpp functions, because I for one will lose
>>> the email connection but it would be useful to preserve the reason for
>>> this change.
>>> Lastly, Andrew, do you need a sun employee to check this in? I will do
>>> this once we resolve the issues above.
>> Done. I also made the patch a little more robust: at the moment Linux
>> pthreads doesn't use growable mappings for thread stacks so there's no
>> need to munmap() the guard pages. If ever pthreads does switch over to
>> using growable mappings I'm pretty sure this code will still work.
> Looks good to me. Thanks!
> - Keith
More information about the hotspot-dev