[haiku-development] Re: Writing Device Drivers - CRAM - Problem uninit_driver()
- From: Axel Dörfler <axeld@xxxxxxxxxxxxxxxx>
- To: haiku-development@xxxxxxxxxxxxx
- Date: Sat, 20 Aug 2011 09:27:29 +0200
On 08/20/2011 04:41 AM, Earl Pottinger wrote:
One minor problem with my driver is flushing the contents of the RAMDISK
to storage. Presently, I solve problem by calling my flush() every three
seconds or so.
As Ingo already pointed out in an earlier mail, it really sounds as if
your RAMDISK is just a poor man's cache, and as such completely
superfluous, and extremely counter productive on Haiku.
What I really want to do is to flush the contents when the driver is
unloaded/shutdown. The uninit_driver() should serve this purpose, but in
my tests this call is taken least than 1-in-20 times whenever I select
Haiku's shutdown command.
This causes a problem because while the BFS ioctl() passes my code a
flush request, some writes() will occur that flush request. If I do not
That should really not happen. Can you identify what exactly is still
trying to write?
force a flush() after those final writes() the resulting drive image is
not a valid one or is missing the last files written to it when I reboot
and reload the resulting drive image to memory.
Any suggestions in how to insure uninit_driver() is always called after
Haiku-OS starts a shutdown?
The word you were looking for is probably "ensure", BTW. Besides that,
there currently is no such mechanism for this. The shutdown mechanism is
currently very (read: too) simplistic: it just kills all teams left, and
then sync()s. It doesn't even wait for all teams to be dead, and only
even kills anything when rebooting.
Patches welcome.
Bye,
Axel.
Other related posts:
