------------------- > Hi, > > > I'd like to know how a modularized PCI subsystem can be loaded _AFTER_ > > a PCI dependent FB driver. > > Well, duh, obviously you can't do that. :-/
Heh, I has a feeling that wouldn't be that easy ;) > > What then if the PCI buswalk kills the kernel? > > You write these parts of the kernel defensively so that this doesn't > happen, and/or you test it extensively so that you're sure it won't > happen. "this doesn't happen"/"you're sure it won't happen" ? Are you kidding, or are we talking about "Alice in kerneland"? 0% failure does not exist, that's a RULE. > You boot without framebuffer options, and look at the console text. Such a console text doesn't exist on ppc. Nor on mips, iirc. There are issues on hppa as well, etc. > > This is a gross example, still it proves that kind of idea can't be > > implemented. > > > It is the direction the kernel people want to move further in. Though that may be a "General Direction", some arch dependent facts have to be taken into account. Last time I checked, the Linux kernel wasn't an x86-only kernel ;) Greetings, Thibaut VARENE PA/Linux ESIEE Team http://www.pateam.org/
