"fast data access mmu miss" on kernels w/o "makeoptions DEBUG=-g"

[Marius Strobl]

On Sat, Aug 27, 2005 at 10:20:08PM +0900, R. Imura wrote:
> On Sat, Aug 27, 2005 at 02:55:03PM +0200, Marius Strobl wrote:
> > On Sat, Aug 27, 2005 at 02:58:41PM +0900, R. Imura wrote:
> > > Hi,
> > > 
> > > On Wed, Aug 24, 2005 at 07:14:37PM +0200, Marius Strobl wrote:
> > > > > > smbfs does not work on sparc64 as smbfs assumes it runs on
> > > > > > architectures that allow on-aligned memory access. It also needs
> > > > > > big-endian clean up too.(See sys/mchain.h)
> > > > > 
> > > > > To begin with, you might want this patch. :)
> > > > > 
> > > > 
> > > > You probably should go the NetBSD route and change the consumers
> > > > of these macros to use the byteorder(9) functions like htole16(9)
> > > > directly. Btw., NetBSD seems to also have fixed the alignment
> > > > issues.
> > > 
> > > I was too rushed that these macros in my patch are no longer used
> > > in our system, so we can safe to remove them.
> > > 
> > > Fixes are done in 3 years ago.
> > > http://docs.freebsd.org/cgi/mid.cgi?200212161620.gBGGK6uX080655
> > > 
> > 
> > That commit only dealt with the kernel part, these macros are still
> > used in the userland part:
> > > grep -r htoles /usr/src/contrib/smbfs/*
> > /usr/src/contrib/smbfs/lib/smb/nb_name.c:#define        NBENCODE(c)     (htoles((u_short)(((u_char)(c) >> 4) | \
> > 
> > Marius
> 
> I see. :)
> Then, could you test this patch?
> 

While it should fix compilation on sparc64 it still doesn't deal
with the alignment issues. Please have a look at the changes that
went into the respective kernel and userland parts in NetBSD if
you want to get it working on FreeBSD/sparc64, too.

Marius