Re: [COUNTERPATCH] mm: avoid overflowing preempt_count() in mmu_take_all_locks()

Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]
From: Andrea Arcangeli
Date: Thursday, April 1, 2010 - 9:02 am

On Thu, Apr 01, 2010 at 06:51:17PM +0300, Avi Kivity wrote:

We can always switch to srcu. Switching to srcu is not a noop for all
mmu notifier invalidates only after these locks can schedule. At that
point, so with srcu + mutex in the rmap locks, all the mmu notifier
invalidates can schedule, allowing XPMEM to be synchronous its
invalidates, and making it safe.
--
Previous message: [thread] [date] [author]
Next message: [thread] [date] [author]

Messages in current thread:
Re: [COUNTERPATCH] mm: avoid overflowing preempt_count() i ..., Andrea Arcangeli, (Thu Apr 1, 9:02 am)