[PATCH] Kernel Module Loader vs. v2.5.43

[Rusty Russell <rusty@rustcorp.com.au>]

[-- Attachment #1: Type: text/plain, Size: 1501 bytes --]

Changes this version:
  o Updated to 2.5.43
  o Removes the "spurious failure" problem, which could
    theoretically occur during a failed unload attempt.  To do this, it
    uses a very lightweight refcount machanism in try_module_get(), and a
    complex machine-stopping maneuver on attempted unload.  This solution
    is inspired by discussions with Daniel Phillips.
  o You can use "no_module_init;" instead of declaring an init function.

It passes stress testing on my dual x86 box.

The full set of patches are available on my web page:
  http://www.kernel.org/pub/linux/kernel/people/rusty/patches

And as one convienent megapatch:
  http://www.kernel.org/pub/linux/kernel/people/rusty/patches/module-x86-18-10-2002.2.5.43.diff.gz

Cheers!
Rusty.
--
  Anyone who quotes me in their sig is an idiot. -- Rusty Russell.

D: This is an implementation of the in-kernel module loader extending
D: the try_inc_mod_count() primitive and making its use compulsory.
D: This has the benifit of simplicity, and similarity to the existing
D: scheme.  To reduce the cost of the constant increments and
D: decrements, reference counters are lockless and per-cpu.
D:
D: Modversions, Module parameters and kallsyms are not supported in
D: this patch, they come in future patches.  inter_module_get/put have been
D: eliminated: use symbol_get() and symbol_put().
D:
D: You will need the trivial replacement module utilities from:
D: http://www.kernel.org/pub/linux/people/rusty/module-init-tools-0.5.tar.gz


[-- Attachment #2: New Module Loader for 2.5.43 --]
[-- Type: application/octet-stream, Size: 36857 bytes --]