From: Chris W. <cw...@f0...> - 2004-11-03 20:10:12
|
On Wed, Nov 03, 2004 at 08:28:44PM +0100, Blaisorblade wrote: > I'm going to test this. please do > I thought that Gerd Knorr patch (which I sent cc'ing LKML and most > of you) already solved this (I actually modified that one, replacing > his SIGCONT kill()ing with a PTRACE_KILL, but I did this in the > places he identified). it might, i'm going to check soon what worries me is that two very different code paths might be fixing the same problem which makes me think the flow of execution here is very vague and needs cleaninng up also, check your return values for errors --- i bet you will see some. os_kill_process has this problem too --- many invocations of it are pointless and fail, especially those from relase_thread_tt (i need to check the details here, this is all from memory and im getting old) > I guess that old_pid should either already be dead there or going to > die after a little, but I'm going to check (after I get UML to run > in the current snapshot...) it should build pretty close to as-is, if not let me know and i'll sent you what i have |