lkml.org 
[lkml]   [2011]   [Jun]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Date
From
SubjectRe: [PATCH 4/8] x86, mce: rename bootparam parser
On Fri, Jun 17, 2011 at 03:25:51PM -0700, Luck, Tony wrote:
> > and leave this like this. "nomce" is the same as "mce=off" and frankly,
> > I'd like to remove this redundancy, thus no need to do the code
> > relocation. In addition, I don't think there are lots of systems running
> > with "nomce" so I really think we should drop it.
> >
> > So Ingo, hpa, what is the proper way to remove early setup params? Maybe
> > through Documentation/feature-removal-schedule.txt?
>
> Though it seems odd to me that anyone would want to turn mce off,
> the fact that we have two ways to do so would indicate that people
> do (or at least did) want to do this.

Yeah, the only usecase I could think of for this is people doing their
own userspace DRAM ECC evaluation like google. But they'd still need to
have an #MC handler... or they could parse syslog for the output from
the default unexpected_machine_check().. hm.

> It seems like we'd need a long "deprecated" period (till all the
> major OSVs turn out a new release) to remove this without surprises.
> If the "nomce" option just disappears, then people using it will
> not realize that their boot time argument was ignored (until they
> see a reported error).

Yeah, we could do a grace period with a warning:

/*
* Old style boot options parsing. Only for compatibility.
*/
static int __init mcheck_disable(char *str)
{
pr_err("\"nomce\" boot param is deprecated. Please use \"mce=off\"\n");
mce_disabled = 1;
return 1;
}
__setup("nomce", mcheck_disable);
and then kill it.

--
Regards/Gruss,
Boris.


\
 
 \ /
  Last update: 2011-06-18 10:41    [from the cache]
©2003-2011 Jasper Spaans