Re: [PATCH] drivers/net/tile/: on-chip network drivers for the tile architecture

[Eric Dumazet <eric.dumazet@gmail.com>]

Le mercredi 03 novembre 2010 à 13:37 -0400, Chris Metcalf a écrit :
> Stephen, thanks for your feedback!
> 
> On 11/3/2010 12:59 PM, Stephen Hemminger wrote:
> > 1. MUST not use volatile, see volatile-considered-harmful.txt
> 
> The "harmful" use of volatile is in trying to fake out SMP.  Believe me,
> with a 64-core architecture, we know our SMP guidelines. :-)  Our use here
> is simply to force the compiler to issue a load, for the side-effect of
> populating the TLB, for example.
> 
> However, your response does suggest that simply the syntactic use of
> "volatile" will cause a red flag for readers.  I'll move this to an inline
> function in a header with a comment explaining what it's for, and use the
> function instead.

Please read Documentation/volatile-considered-harmful.txt

Then if there is a problem, we can make change to the documentation, but
volatile use in new code is _strictly_ forbidden.

ACCESS_ONCE() is your friend, we might document it in
Documentation/volatile-considered-harmful.txt