On Mon, Mar 19, 2007 at 07:31:11AM +0100, Rudolf Marek wrote: > Maybe OpenVZ needs both safe and unsafe variants in msr-on-cpu? Frankly I don't understand when one should use rdmsr_safe() instead of rdmsr(). p4-clockmod driver worked fine with plain rdmsr/wrmsr. For general education, why this driver uses _safe functions? What's wrong with http://ssh.cz/~ruik/patches/add-msr-io-safe.patch is: a) naming: _on_cpu is suffix and thus should be last FOO(...) => FOO_on_cpu(unsigned int cpu, ...) b) static qualifier should be first. c) coding style in wrmsr_on_cpu/rdmsr_on_cpu/... static int foo(...) { ... } d) +#include <linux/errno.h> what for? those dummy inlines don't include -E* > >> On URL bellow is current version of coretemp driver. > >> > >> http://ssh.cz/~ruik/patches/ > >> > >> What was fixed? I fixed the issues Jean pointed out, I created separate patch > >> for ...msr_safe stuff. > >> > > In arch/i386/lib/msr-on-cpu.c, I'm not sure if you really need to create > > 2 versions of the functions (safe and unsafe). Anyway, no one uses these > > functions in the kernel yet, so I think you can simply convert them to > > safe versions. > > > -snip- > > I will answer the rest in the evening.
