On Mon, 29 Sep 2008, Andrew Doran wrote:
On Mon, Sep 29, 2008 at 02:04:40PM +0100, Stephen Borrill wrote:ftp://ftp.netbsd.org/pub/NetBSD/misc/sborrill/aac64bit.diff+ aprint_normal_dev(&sc->sc_dv, "Enable 64-bit array support\n"); Don't see why the user cares once it is working.
I think it's a useful diagnostic about the features of the card (that can't easily be deduced by looking at the "Controller supports" line. I think it's more useful than the existing "New comm. interface enabled" message.
+ size = htole16(sizeof(fib->Header) + size); + fib->Header.Size = htole16(size); One too many calls to htole16(), otherwise it looks good to me.
This is present in the original parts of ld_aac.c too (see around lines 221). I couldn't work out why either, but seeing it was consistent I didn't want to touch it! I guessed bad things would happen on big-endian platforms.
Note that the ld driver needs changing to handle 64-bit dump offsets.
Noted, though in mitigation this is only going to affect people who've got ld0b starting > 2TB.
-- Stephen