vili at huwico.hu
Fri May 12 12:31:14 CEST 2006
On May 12, 2006, at 1:08 AM, Adam Lackorzynski wrote:
> On Thu May 11, 2006 at 10:09:10 +0200, Vilmos Nebehaj wrote:
>> I've compiled and installed Fiasco and L4Linux-2.6 and it works,
>> that when I start a simple test program that creates lots of
>> pthreads, the
>> whole L4Linux task freezes (simply pthread_create() a thousand
>> threads after
>> each other is enough to reproduce the problem). I can ping the
> As you already seem to have a small test program, can you just send it
> to me (possibly via PM) and I'll check?
Yes, of course. I'm not sure whether mailman strips attachments, so I've
put it up to http://broadbit.com/~vili/pthread_create.c
>> machine from
>> another host, but can't login through ssh, the console becomes
>> unresponsive, etc.
>> At the same time, the following message appears in the log:
>> simplets| ERROR: task #23 doesn't match: client 10.03, owner 0.0
>> l4lx | l4ts_create_task(): failed (server=7.00, ret=-3, exc=0)
>> Other L4 tasks continue running uncluttered.
>> Do you have an idea why this happens? Is this a bug in Fiasco, or
>> is it
>> a misconfiguration problem?
> Currently it's not possible to create thousands of threads not
> of tasks. I have something in the queue to get the first limit
> and we'll probably do something for the second too.
> I don't think it's a bug in Fiasco and probably not a misconfiguration
> but let me check first.
It doesn't create them at the same time, it just tests how much time it
takes to create a given number of detached pthreads exiting immediately.
More information about the l4-hackers