login
Header Space

 
 

Re: [feature] automatically detect hung TASK_UNINTERRUPTIBLE tasks

Score:
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
To: Ingo Molnar <mingo@...>
Cc: <linux-kernel@...>, Andrew Morton <akpm@...>, Arjan van de Ven <arjan@...>, Thomas Gleixner <tglx@...>
Date: Saturday, December 1, 2007 - 2:42 pm

On Sat, 1 Dec 2007, Ingo Molnar wrote:


Wouldn't a natural extension of this feature be to mark these hung 
TASK_UNINTERRUPTIBLE tasks with a new thread flag such as TIF_HUNG for the 
purposes of the OOM killer?

Right now, the OOM killer will become a no-op when any candidate task that 
it scans through is found to have the TIF_MEMDIE flag when selecting a 
task to kill.  So any hung task in this state could cause the OOM killer 
to infinitely loop.

If lockdep could set_tsk_thread_flag(g, TIF_HUNG), this could be detected 
in the OOM killer and not only could we prevent the infinite looping but 
we could also clear TIF_MEMDIE and reduce the increased timeslice that the 
OOM killer gives to the tasks it kills.

		David
--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
Re: [feature] automatically detect hung TASK_UNINTERRUPTIBLE..., Radoslaw Szkodzinski, (Mon Dec 3, 6:15 am)
Re: [feature] automatically detect hung TASK_UNINTERRUPTIBLE..., David Rientjes, (Sat Dec 1, 2:42 pm)
speck-geostationary