Date: Mon, 05 Jul 1999 15:42:13 -0700 From: Tom Pavel <pavel@SLAC.Stanford.EDU> To: freebsd-current@FreeBSD.ORG Subject: Kernel versions and kld modules [WAS: this of interest to anyone?] Message-ID: <0FEF008IZ5QDMA@mailbox.slac.stanford.edu> In-Reply-To: "Your message of Sat, 03 Jul 1999 00:40:08 %2B0800." <19990702164008.2619E64@overcee.netplex.com.au>
next in thread | previous in thread | raw e-mail | index | archive | help
>>>>> On Sat, 3 Jul 1999, Peter Wemm <peter@netplex.com.au> writes: > A few key suggestions for people still along for the ride: > > 1: When you've got a good running kernel that you're happy with, do yourself > a big favour and copy it from /kernel to /kernel.ok or something like that. > So, when you manage to get a bad /kernel and /kernel.old, you've still got > a fallback that doesn't mean resorting to a fixit disk boot. This reminds me of something I wanted to ask about. A little while back I made the typical junior mistake of building/installing a new kernel without doing the make install of the related modules. In my case, the crash came from kernfs and it was easy to boot single-user and comment out a line in fstab. But my question is with the trend towards more module-ization of the kernel, does saving a /kernel.old make sense without saving a /modules.old? Should the config-generated Makefile for the kernel have a target to build and install the new modules as well (as opposed to installing them with make world)? Should we install modules into some path identified by a kernel version (whatever that means exactly)? The extreme end of this line of logic is the Linux module hashing mechanism, which I've never fully understood. My experience with it is that it can prevent you from using a binary-only module (AFS is the one I'm thinking of) even if the interfaces haven't really changed. It seems like a poor substitute for proper management and versioning of kernel interfaces. I suppose much of this becomes a non-issue outside of -current, and maybe that's why no one has seemed worried about it. Even in -stable, though, I suppose you could have incompatible kernel interfaces introduced that might break modules. It just seems to me that one should consider /kernel plus /modules/* to be more of unit. Thanks for any insights. Tom Pavel Stanford Linear Accelerator Center pavel@slac.stanford.edu http://www.slac.stanford.edu/~pavel/ To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?0FEF008IZ5QDMA>