cvs commit: src/sys/dev/io iodev.c

[Poul-Henning Kamp]

In message <20080813014505.H1310 at besplex.bde.org>, Bruce Evans writes:

>I once thought that using D_TRACKCLOSE was the right fix for this in
>serial drivers.

D_TRACKCLOSE Was created for drivers that need per open state
reclaimation, forced closes is an entirely different kettle of fish.

Writing a driver to support forced closes is significantly more work
than writing a "normal" driver, but it can be eased a lot if we 
add infrastructure to keep track of and wake up all sleeping threads
in a particular driver.

Either way, it is significant architectural work.

Poul-Henning

-- 
Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.