Re: kernel fault -current of 23/24 Apr 2015

[Paul Goyette <paul%vps1.whooppee.com@localhost>]

On Sat, 25 Apr 2015, Masao Uebayashi wrote:

> Can't you defer attachment of your failing device?

That was my initial thought.  But there are lots of potential failing 
devices.

A quick scan shows that there are about 125-130 sources files which 
attempt to register with sysmon_{pswitch,wdog,envsys}_register.  Some of 
those are already modularized, so we could simply refuse to attach 
during autoconfig, since the modules' modcmd(INIT) routines would be 
called later, during module initialization, and in an appropriate order 
to obey inter-module dependencies.

But there are a lot of others which are not yet modularized, so they 
won't get "called later during module initialization".


An alternative might be to modify all of these 125-130 files to use 
config_defer() (or some new, similar, config_defer_module() routine) to 
delay the sysmon_*_register() calls until after module initialization.


The big problem with either of these approaches is the scope of the 
work.  There's a lot of work to do.

And of course, the other big problem with either of these approaches is 
the amount of testing (including some for non-tier-1 platforms).

:)


I'm willing to entertain other suggestions!  :)


-------------------------------------------------------------------------
| Paul Goyette     | PGP Key fingerprint:     | E-mail addresses:       |
| (Retired)        | FA29 0E3B 35AF E8AE 6651 | paul at whooppee.com    |
| Kernel Developer | 0786 F758 55DE 53BA 7731 | pgoyette at netbsd.org  |
-------------------------------------------------------------------------