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

R. Imura imura at ryu16.org
Tue Sep 6 08:21:35 PDT 2005

On Mon, Aug 29, 2005 at 12:10:58PM +0900, Pyun YongHyeon wrote:
> Supporting smbfs on sparc64 is on my TODO list. But it's priority
> is low due to a lack of time and other pending works. Based on
> NetBSD's changes I managed to build smbfs. But it didn't work and it
> needs more investigation. While reading the code I noticed there
> is a bug in smbfs_hash() in sys/fs/smbfs/smbfs_node.c. NetBSD
> switched to use hash32_strn() defined in <sys/hash.h>. I think it
> also has influence on our implementation. Would you please check
> the function?
> (http://cvsweb.netbsd.org/bsdweb.cgi/src/sys/fs/smbfs/smbfs_node.c, rev. 1.12)

IMHO, we are not fully required to fix this problem at first
because the mission of smbfs_hash() is to create unique i-node number.
(however smbfs_hash can be replaced by any hash algorithm,
though. Our smbfs_hash is called FNV hash according to google.)

I hope you also look at Darwin's code, because I believe it may
bring us some benefit easily if we keep differences small between Darwin's
and ours, since Darwin has many improvement.
For example, NetBSD had introduced SMBRQ_PUTLE16 macro but Darwin hadn't.
I think this doesn't mean SMBRQ_PUTLE16 is an essential feature.
BSD licensed Darwin's smbfs can be found at
http://darwinsource.opendarwin.org/ .

BTW, I found we can obtain macros from NetBSD in

- R. Imura

More information about the freebsd-sparc64 mailing list