I'm looking to find a way how to identify a non responding (not zombie) process programmatically. I find some information to check TH_STATE_UNINTERRUPTIBLE status but there was some discussion that it's not the开发者_开发百科 right way.
I'm assuming you mean a spinning wheel application hang? There are many ways to freeze. The particular cause is important. If it's a Cocoa app, you could try sending your main thread / Window an event... or scripting up Spin Control.
A random answer … I'm not a programmer but I stumbled across something of possible interest whilst working through stuff in Ask Different …
sched_prim.c (Scheduling primitives) in relatively old xnu-124.7 includes:
#define MAX_STUCK_THREADS 128
/*
* do_thread_scan: scan for stuck threads. A thread is stuck if
* it is runnable but its priority is so low that it has not
* run for several seconds. Its priority should be higher, but
* won't be until it runs and calls update_priority. The scanner
* finds these threads and does the updates.
*
* Scanner runs in two passes. Pass one squirrels likely
* thread ids away in an array (takes out references for them).
* Pass two does the priority updates. This is necessary because
* the run queue lock is required for the candidate scan, but
* cannot be held during updates [set_pri will deadlock].
*
* Array length should be enough so that restart isn't necessary,
* but restart logic is included. Does not scan processor runqs.
*
*/
thread_t stuck_threads[MAX_STUCK_THREADS];
int stuck_count = 0;
/*
* do_runq_scan is the guts of pass 1. It scans a runq for
* stuck threads. A boolean is returned indicating whether
* a retry is needed.
*/
– is that, about stuck threads, food for thought?
Or too way off from the question about processes?
At a glance, no comparable block of code in sched_prim.c in xnu-1699.26.8 source for Mac OS X 10.7.4.
精彩评论