Re: thread suspension when dumping core

On Wed, Jun 08, 2016 at 07:30:55AM +0300, Konstantin Belousov wrote:
> On Tue, Jun 07, 2016 at 11:19:19PM +0200, Jilles Tjoelker wrote:
> > In this case it is clear which sleep(9) calls should be affected so it
> > may be better to avoid more hidden state.
> In this case yes, but apparently some out-of-tree users exist. And,
> the marking of the single sx_sleep() call depends on knowing our
> implementation. I remember that as the arguments to change from PBDRY to
> the current state setter, in NFS it is not too hard to try to enumerate
> interruptible sleeps. 
> 
> > 
> > I also wonder whether we may be overengineering things here. Perhaps
> > the advlock sleep can simply turn off TDF_SBDRY.
> Well, this was the very first patch suggested.  I would be fine with that,
> but again, out-of-tree code seems to be not quite fine with that local
> solution.

In our particular case, we could possibly use a similar approach. In
general, it seems incorrect to clear TDF_SBDRY if the thread calling
sx_sleep() has any locks held. It is easy to verify that all callers of
lf_advlock() are safe in this respect, but this kind of auditing is
generally hard. In fact, I believe the sx_sleep that led to the problem
described in D2612 is the same as the one in my case. That is, the
sleeping thread may or may not hold a vnode lock depending on context.