Re: envsys/gttwsi deadlock

[Taylor R Campbell <campbell+netbsd-tech-kern%mumble.net@localhost>]

   Date: Sat, 24 Oct 2015 14:33:50 +0200
   From: Manuel Bouyer <bouyer%antioche.eu.org@localhost>

   I looked a bit, and it's not that easy.
   sysmon_envsys_refresh_sensor() is called from various places with sme_mtx
   held; sometimes the function calling sysmon_envsys_refresh_sensor() is
   entered with the lock held.

Yes...  The locking was very badly done to begin with.  However, I am
pretty confident that the structure I quoted is the right thing, even
if it requires more changes beyond the ioctl in question.

                               sme_mtx protects various states of sme, and
   reworking this requires quite a bit of work (dropping it before calling
   sysmon_envsys_refresh_sensor() and reaquiring after is not an option).
   More work than I can put in this right now ...

Please document any lock requirements that are not written down, if
you've found any!