Re: [RFC] get_write_access()/deny_write_access() without inode->i_lock
From: Al Viro <viro@ZenIV.linux.org.uk>
Date: 2011-06-20 17:03:05
Also in:
linux-fsdevel, lkml
On Mon, Jun 20, 2011 at 05:42:11PM +0100, Al Viro wrote:
BTW, atomic_add_unless/atomic_inc_not_zero is done as read/cmpxchg on everything I've checked, including alpha and sparc. I suspect that it's seriously suboptimal there...
Actually, on sparc it's probably fine. alpha, OTOH, probably ought to switch to something similar to what ppc is doing...