Re: [Lse-tech] scalability of signal delivery for Posix Threads

[Ray Bryant <raybry@sgi.com>]

Andi Kleen wrote:
> On Mon, Nov 22, 2004 at 05:51:59PM +0100, Andreas Schwab wrote:
> 
>>Andi Kleen <ak@suse.de> writes:
>>
>>
>>>At least in traditional signal semantics you have to call sigaction
>>>or signal in each signal handler to reset the signal. So that 
>>>assumption is not necessarily true.
>>
>>If you use sigaction then you get POSIX semantics, which don't have this
>>problem.
> 
> 
> It's just a common case where Ray's assumption is not true.
> 
> -Andi
> 

True enough.  And in that case the design that I was describing wouldn't
make sigaction() that much more expensive since if you are not in the POSIX
thread environment (more precisely, the thread was not created with
CLONE_SIGHAND) each thread has its own sighand structure and the "global" 
locking mechanisum I had proposed would only require the taking of one 
additional lock.

However, special casing ITIMER_PROF is also a reasonable avenue of approach.
The performance monitor code can also deliver signals to user space when
a sampling buffer overflows, and this can have the same kind of scaling
problem as ITIMER_PROF.  I'll have to do a little research to figure out
how exactly that works, but that signal (SIGIO?) would also be a candidate
for special casing on our platform.

-- 
Best Regards,
Ray
-----------------------------------------------
                   Ray Bryant
512-453-9679 (work)         512-507-7807 (cell)
raybry@sgi.com             raybry@austin.rr.com
The box said: "Requires Windows 98 or better",
            so I installed Linux.
-----------------------------------------------