On Fri, May 04, 2018 at 05:26:40PM +0100, Al Viro wrote:

On Fri, May 04, 2018 at 06:21:02PM +0200, Peter Zijlstra wrote:

quoted

On Fri, May 04, 2018 at 06:07:26PM +0200, Sebastian Andrzej Siewior wrote:

quoted

do you intend to kill refcount_dec_and_lock() in the longterm?

You meant to say atomic_dec_and_lock() ? Dunno if we ever get there, but
typically dec_and_lock is fairly refcounty, but I suppose it is possible
to have !refcount users, in which case we're eternally stuck with it.

Yes, there are - consider e.g.

void iput(struct inode *inode)
{ 
        if (!inode)
                return;
        BUG_ON(inode->i_state & I_CLEAR);
retry:
        if (atomic_dec_and_lock(&inode->i_count, &inode->i_lock)) {

inode->i_count sure as hell isn't refcount_t fodder...

Yeah, I should've remembered, I tried to convert that once ;-) i_count is
a usage count, not a refcount.