From: Jeff Dike <jdike@ka...> - 2002-03-14 18:16:52
> In fact, there is definately a race for UML SMP.
There's probably lots of them.
If you want to get UML booting SMP reliably, you could audit the UML arch
code for SMP safety.
> Does this mean you are going to fix this?
It's a bug, therefore I'm going to fix it.
From: Jeff Dike <jdike@ka...> - 2002-03-14 19:45:17
> However, in the name of preventing base changes, I suspect all the
> post-fork logic in kernel_thread1() has to be moved into
Yes, that's definitely the fix. kernel_thread1 was bogus from the start.
Factor out the stuff a kernel thread has in common with a process fork, make
that the main line of copy_thread, then a branch into the stuff they don't
have in common.
> If Jeff doesn't have the time to fix it right now, I'd be interested
> in doing it.
That would be great.