Re: Design hole in nptl (and more general in POSIX threads)

  Jim,
  
  Thank you very much! Together with few pieces of Boehm GC, this looks
like enough information for me to finish this stack examining. In fact,
I only have to make hash table for low level mapping of pthread_t to
traced references of Modula-3 Thread.T object to (I hope :)) finish my
work.

  What bothers me now is... What procedures in pthread API are async
safe? Boehm GC uses pthread_self(), as probably you too are using. Is it
safe to use pthread_key_getspecific()?

  Thanks,
  dd

Дана сре, 17-12-2003 у 21:12, Jim Blandy је написао:
> You may find these helpful:
> 
> http://svn.red-bean.com/repos/minor/trunk/gc/threads.c
> http://svn.red-bean.com/repos/minor/trunk/gc/threads.h
> 
> They are a work in progress, and are completely untested; I haven't
> even compiled them yet.  (Hopefully I'll be able to do that in the
> next few weeks.)  But there are extensive comments explaining how it's
> supposed to work, which might give you some ideas.
> 
> The general approach to stopping threads is borrowed from the Boehm
> GC, but this is a precise GC, and I have arranged the user's API so
> that I don't need to intercept thread creation calls or do the other
> strange things Boehm does.
> 
> It does depend on architecture-specific memory barrier functions;
> POSIX doesn't include any functions that both synchronize memory and
> are async-safe.
> 
> The "Minor C API" referred to in the comments is here, but you should
> be able to make sense of threads.[ch] without reading it:
> 
> http://svn.red-bean.com/repos/minor/trunk/include/minor/minor.h
>