Re: [PATCH /3] sh64: remove cli()/sti() from arch/sh64/*

[Jim Nelson <james4765@cwazy.co.uk>]

Al Viro wrote:

>On Tue, Jan 04, 2005 at 08:22:47PM -0600, James Nelson wrote:
>  
>
>>This series of patches is to remove the last cli()/sti() function calls in arch/sh64.
>>    
>>
>
>Wait a minute.  Is that just a blanket search-and-replace job?  There is
>a reason why cli/sti is marked obsolete instead of being silently #define'd
>that way.  Namely, in a lot of cases users of cli/sti are actually racy.
>
>For such instances replacing these with local_... would not improve anything
>(obviously) *and* would hide a trouble spot by silencing a warning.
>
>I'm not familiar with the architectures in question, so it might very well
>be that all replacements so far had been correct.  However, I would really
>like to see rationale for each of those warning removals to go along with
>the patches.
>
>Note that basically you are doing "remove the warning in foo.c:42 and
>keep the current behaviour".  The missing part is "current behaviour is,
>in fact, correct in that place and does not deserve a warning because
><list of reasons>".
>
>  
>
Everything I've looked at so far has been for single-processor systems 
AFAICT - embedded processors, evaluation boards, etc.  I do not pretend 
to have intimate familiarity with the hardware in question, and I will 
be much more careful when I reach anything that can be plugged into an 
SMP box, but I was grabbing the low-hanging fruit first.  The nasty 
stuff (drivers/char, for example) will come later.

That's why I cc'd the arch maintainers - figured they'd whack me with a 
cluebat if I'd overlooked something.