[FreeBSD-CVS] cvs commit: src/sys/dev/twa twa_cam.ctwa_globals.c twa_ioctl.h

[Vinod Kashyap]

> The same problem (3ware controller found, but device /dev/daX not
> discovered on booting and hence no root mount point) popped up for
> me yesterday under RELENG_5, too. The problem can be traced back to
> changes between 2005-02-03 (date of my old still working kernel) and
> 2005-02-10 (date of my new broken kernel). The whole diff to sys/
> of RELENG_5 is about 350 KB (I used "cvs diff -jRELENG_5:2005-02-03
> -jRELENG_5:2005-02-10") and I was at least able to reduce the 
> origin of
> the problem to the appended 29 KB change set.
> 
> With the latest RELENG_5 as of 2005-02-11 (today) but with 
> this appended
> patch again backed out, the twa(4) problem is solved, too. Similarily,
> if I do not back out those changes and just use plain RELENG_5 as of
> 2005-02-11 with your patch above applied it is also solved.
> 
> I guess your patch workarounds the problem, but the real origin of the
> problem is somewhere in the changes we can see in the appended patch.
> Unfortunately I do not understand the side-effects of those 
> changes so I
> was not able to figure out myself which part of the change is actually
> causing the trouble for the twa(4) driver.
> 
> Perhaps someone else can jump in and fix the origin of the problem...
> 

But the problem was first reported with RELENG_5 on Jan 12th, '05.
And I couldn't find any CAM related changes in the patch you sent.
The thing is, the occurrence of this seems to be pretty inconsistent.
I have never seen the problem myself at boot time, although I have
seen CAM not send down inquiries some times, on doing a kldload of
the driver (without my latest patch).