Date: Wed, 27 Apr 2011 10:19:31 -0400 From: John Baldwin <jhb@freebsd.org> To: freebsd-hackers@freebsd.org Cc: Kostik Belousov <kostikbel@gmail.com>, Bartosz Fabianowski <freebsd@chillt.de>, Hans Petter Selasky <hselasky@c2i.net> Subject: Re: Is there some implicit locking of device methods? Message-ID: <201104271019.31844.jhb@freebsd.org> In-Reply-To: <4DB76085.4000402@chillt.de> References: <4DB695DB.1080505@chillt.de> <20110426124403.GQ48734@deviant.kiev.zoral.com.ua> <4DB76085.4000402@chillt.de>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tuesday, April 26, 2011 8:17:09 pm Bartosz Fabianowski wrote: > > If you needs per-file private data for cdev, you would be better served > > by cdevpriv(9) KPI. Cloning is too hard to use correctly for such task. > > Thanks, I just got that working. To help those going down a similar path > in the future, I would like to note quickly that the following must be > added to the cdevsw structure to ensure proper clean-up: > > .d_flags = D_TRACKCLOSE > > I just spent hours debugging panics until I realized only the last > close() was triggering a call to my .d_close method. Err, if you use cdevpriv you shouldn't even have a d_close method. All your d_close logic should be in the cdevpriv destructor, and the kernel will call your destructor when all references to an open file descriptor go away (i.e. it is closed). -- John Baldwin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?201104271019.31844.jhb>