Subject: re: fixed sized disk structures
To: Reinoud Zandijk <firstname.lastname@example.org>
From: Bill Studenmund <email@example.com>
Date: 02/23/2001 08:19:00
On Fri, 23 Feb 2001, Reinoud Zandijk wrote:
> On Fri, 23 Feb 2001, matthew green wrote:
> > sparc64 is not currently broken. it was *way* back, but you and jason
> > made me change time_t to be 32 bits. eeh is dealing with the fact that
> > on 64 bit solaris, time_t is 64 bits.
> Hmm.... well if it can solve future problems, please make it uint_64 again
> :) ... ppl might have some initial problems swiching over but rather have
> a ``smart'' fsck solve it once than giving trouble later :(
Note: time_t will be a signed quantity so that we can express time
differences with it.
Using 64-bit times in ufs won't work for a number of reasons. One is that
we can't fit that into the inode (we don't have three 32-bit ints to
spare). Second, we need to keep the inode footprint as it is so that the
boot loaders on a number of ports (where the machine's ROMs know how to
read ffs) will keep working.
The tentative plan is to combine the time fields (Xtime and Xtimensec) in
the inode, and split them say 20 bits sub-second and 44 bits of
seconds. That would give us microsecond accuracy and keep us going for
many many milenia.