N:M-multiplexing threads library (Was: SawMill Multiserver...)

[Farid Hajji]

> > People interested in porting the Hurd to L4 are invited to
> > join the address@hidden mailing list. l4-hurd was put on hold
> > some time ago, since we didn't find volunteers with expertise
> > in thread libraries (we need a n:m-multiplexing thread-library
> > for L4), but we should IMHO have a new try at it. Please note
> [snip]
> 
> I have no idea how useful/useless this is going to be, but IBM
> is developing an n:m-multiplexing thread library for Linux.  I
> have no experience with L4, but I do realise it is nothing like
> Linux, so please ignore this message if there's no way that a
> Linux n:m POSIX thread lib would help hurd on L4 :)
L4's tasks come with a fixed number (typically 64 or 128) of threads.
Every thread except the first one is disabled on task creation time.
There's a system call to activate a thread by passing it the address
of a routine in the instruction pointer. This kind of behaviour could
be wrapped in a custom library call, say:
  int l4_native_thread_create(function_pointer_to_thread_service_routine);
Of course, we could call it whatever we like, probably copying
Linux's names to retain compatibility to whatever Linux' based
thread library we choose.

> http://oss.software.ibm.com/developerworks/opensource/pthreads/
Thanks, I'll have a look at it.

Meanwhile, Linux/Glibc/LinuxThread experts (sorry, I'm not one
of them) could summarize the API of Linux' native threads (_not_
pthreads) and please post them here.

Regards,

-Farid.

-- 
Farid Hajji -- Unix Systems and Network Admin | Phone: +49-2131-67-555
Broicherdorfstr. 83, D-41564 Kaarst, Germany  | address@hidden
- - - - - - - - - - - - - - - - - - - - - - - + - - - - - - - - - - - -
One OS To Rule Them All And In The Darkness Bind Them... --Bill Gates.