Re: Locking in disk(9) api

[haad <haaaad%gmail.com@localhost>]

On Tue, Dec 29, 2009 at 1:39 PM, Matthias Scheler 
<tron%zhadum.org.uk@localhost> wrote:
>
> On 29 Dec 2009, at 0:49, matthew green wrote:
>> after thinking about this during the afternoon, i'm more convinced
>> that the current method is OK.
>
> I agree with that for two reasons:
> 1.) It seems correct for an architecture point of view. The disk(9) data
> Â Âis IMHO just another bit of driver instance data which should be
> Â Âprotected by the driver.

Ok this way it makes much more sense, but we need to rewrite disk(9)
manual page then.


> The patch seems to do the job. My stress test (un-tarring a lot of NetBSD 
> sources
> to two LVM volumes at the same time) completed without problems with the
> modified "dm" driver.

Great, I will prepare better patch and commit it today. I will add
private mutex to dm_dev_t which will be used only for mutual exclusion
for disk routines.


-- 


Regards.

Adam