svn commit: r212964 - head/sys/kern

[Gavin Atkinson]

On Wed, 2010-09-22 at 17:43 +0300, Andriy Gapon wrote:
> on 22/09/2010 16:46 Gavin Atkinson said the following:
> > Ignoring the rest of the discussion about locking, I think this is a
> > step in the right direction.  However, what I feel we should be strongly
> > considering is for textdumps to be enabled by default on release media.
> > 
> > As more groups choose to use the kernels from the release media (PC-BSD,
> > FreeNAS, etc) and put them into places where end users are never
> > expected to recompile kernels, having textdumps enabled by default in
> > RELEASE kernels becomes a bigger and bigger priority.  Groups like
> > PC-BSD don't necessarily have the time or skills to do the needed kernel
> > debugging (and nor should they have to, it's not their purpose), so
> > anything we can do in releases to make sure we have enough info to
> > resolve panics seen by their users is a big bonus.
> > 
> > Is there any real reason why we shouldn't go down this route?
> 
> textdumps need DDB.
> textdumps also need dumpdev which is not enabled by default in
> /etc/defaults/rc.conf, but that's easier to fix for any individual user or a
> FreeBSD "distribution".

Indeed, I was happy to see dumpdev enabled on 7.x, and disappointed to
see it disabled on 8.x.

> I originally suggested to include DDB in the stable GENERICs.
> But that has a number of consequences:
> 1. With DDB a panic results in a system going to ddb prompt instead of rebooting,
> unless KDB_UNATTENDED is also specified.  System not rebooting on panic may be a
> POLA violation and an inconvenience.  However, a user can change this using
> debug.debugger_on_panic=0 sysctl.  Although, sysctl time might be too late in some
> cases, not sure.

Indeed, if we went down this route we would have to make sure that the
existing behaviour of rebooting is preserved.  We'd also have to ensure
Ctrl-Alt-Escape doesn't drop to the debugger on release media.

> For me personally, the significant benefits outweigh possible minor/temporary
> inconveniences and surprises, but this opinion is not shared by all developers.

I (and I suspect any of the other people on the PR database frontline)
would probably support this :)

> Really besides the point, but such FreeBSD "distributions" like FreeNAS and PC-BSD
> can really use their own kernel config, they don't have to use the same GENERICs.

Indeed, although in some ways they shouldn't have to.  Having textdumps
enabled by default would also be a huge win for us, of course.

> P.S. I think that it would have been better if you followed up to my proposal on
> arch at .

I'm not subscribed to -arch at the moment, I probably should be.  I
guess you're talking about the thread in June?  I wish I'd seen that at
the time, I'm 100% in favour of it.

Gavin
-- 
Gavin Atkinson
FreeBSD committer and bugmeister
GPG: A093262B (313A A79F 697D 3A5C 216A  EDF5 935D EF44 A093 262B)