From: Jeff Dike <jdike@ka...> - 2003-02-10 17:11:21
> Here comes a brand new question : Has the TT mode more/as/less
> developpement "Man-Months" as SKAS ? [ since SKAS seems much more
> advocated these days ;-) ]
tt mode has a lot more development behind it. A lot of that is complexity
due to the limits of the host support it gets. A number of things which are
huge, horrible, and (formerly, hopefully :-) bug-ridden in tt mode are
completely trivial (and obviously correct) in skas mode.
See _switch_to for a good example of this.
From: Jeff Dike <jdike@ka...> - 2003-02-10 17:11:23
> We hope this can be of use to others who need to limit resource usage
> in their UMLs. Jeff, do you think this is something worthy of
> inclusion into the UML streams?
Can you use resource limits on the host to accomplish the same thing?
ulimit -u looks like it does what you want.
On the patch itself, a mixture of a UML-specific option and a generic
implementation isn't going to fly. Either make it entirely generic so
everyone can limit the number of processes on their boxes, or make it entirely
UML-specific, and have copy_thread() implement it and return -ENOMEM or
something if it hits the limit.