cvs commit: src/sys/net if_ethersubr.c

[M. Warner Losh]

In message: <20061212183825.GF781 at funkthat.com>
            John-Mark Gurney <gurney_j at resnet.uoregon.edu> writes:
: Luigi Rizzo wrote this message on Tue, Dec 12, 2006 at 06:24 -0800:
: > On Tue, Dec 12, 2006 at 03:17:59PM +0100, Bernd Walter wrote:
: > > On Tue, Dec 12, 2006 at 05:57:56AM -0800, Luigi Rizzo wrote:
: > ...
: > > > then i suppose the proper fix is to revert to m_copy() and
: > > > work on if_simloop() so that 1. it handles a readonly chain, and
: > > > 2. when doing so, it passes up a properly aligned packet...
: > > 
: > > Can't comment on this, as I don't have enough knowledge about network
: > > code.
: > > According to the xscale report it was likely never properly aligned,
: > > the alignment obviously just moved with your change.
: > > 
: > > > however note that there is already some code in net/if_loop.c::if_simloop(),
: > > > just that it uses this:
: > > > 
: > > > 	#if defined(__ia64__) || defined(__sparc64__)
: > ...
: > > > to detect whether the architecture is alignment-sensitive.
: > > > Is there any other identifier that we can use to check ?
: > > 
: > > I wonder how many of these are missing __arm__?
: > 
: > and this is why i wonder if there is a generic identifier
: > that we can use to provide alignment.
: 
: How about?
: #define __NO_STRICT_ALIGNMENT
: 
: It's already defined for amd64 and i386, so all you have to do it test
: that it isn't defined...

__NO_STRICT_ALIGNMENT is relatively new.  Its definition in _types.h,
while established practice in NetBSD, met with considerable
controversy when I tried to expand it in FreeBSD land.  However,
putting things in param.h met with no resistance whatsoever and was
actively suggested.  Maybe we should move this straggler from _types.h
to param.h and start using it more often.

Warner