RE: +AFs-Linux-ia64+AF0- reader-writer livelock proble

RE: +AFs-Linux-ia64+AF0- reader-writer livelock proble

[David Mosberger]

>>>>> On Mon, 11 Nov 2002 14:36:38 -0600, "Van Maren, Kevin" <kevin.vanmaren@unisys.com> said:

  Van> I have not looked at this, but I don't believe it is the right
  Van> way to solve the problem: users who +AF8-need+AF8- to use all
  Van> the CPUs for computation would be punished just to work around
  Van> a kernel implementation issue: that's like saying don't allow
  Van> processes to allocate virtual memory because if the VM is
  Van> over-committed by X amount the kernel deadlocks.

  Van> It would be a bad hack to limit the system-call rate just to
  Van> prevent livelock.

Certainly.  I didn't suggest PRM as a way to _solve_ the livelock
problem, but since Mario asked for a method to cap CPU utilization, I
mentiond it.

	--david

Re: [Linux-ia64] RE: +AFs-Linux-ia64+AF0- reader-writer livelock proble

[Grant Grundler]

"Van Maren, Kevin" wrote:
> It is also possible that a processor can get stuck +ACI-forever+ACI-
> spinning in the kernel with interrupts disabled trying to
> acquire a lock, and never succeed, without the rest of the
> kernel going south.  If that happens, and application will
> be livelocked, but the rest of the system will function.

Probably not. ia64 systems (and x86 systems with IO xapic) direct
IO interrupts to specific CPUs. Devices would not get serviced
in the above case and IO to/from those devices would come to a
grinding halt. It would look more like "dead" lock than "live" lock.


> It really depends on the particular circumstances.

yes. But it sounds very likely to me.

grant