Re: [patch] O(1) sys_exit(), threading, scalable-exit-2.5.31-A6

[Ingo Molnar]

On Mon, 19 Aug 2002, Dave McCracken wrote:

> In looking at the code I was wondering something.  What happens to the
> real parent of a ptraced task when it calls wait4()?  If that's its only
> child, won't it return ECHILD?

yes, this is ugly beyond belief.

eg. under bash start up some code that blocks, eg:

	# cat

the shell will not display a prompt, it wait4()s for the 'cat' process to 
exit.

strace the bash PID - it shows the wait4().

strace 'cat' PID (via strace -p) and keep the strace running - it will
show 'cat' blocked on console input. [some signal like this could happen
to a shell anyway, in a reasonably complex script.]

now do something that breaks bash out of its wait4 - eg. send a 'kill
-SIGCONT BASH_PID' signal - bash returns with a prompt! The 'cat' becomes
a 'background' task - and it gets majorly confused when doing the next
shell command in bash - it displays "cat: -: Input/output error" and goes
zombie.

ptrace is clearly broken - and i tested this with a stock 2.4 kernel.

	Ingo