Hi,
On Wed, 18 Aug 1999 17:39:04 -0600, "Peter J. Braam" <[EMAIL PROTECTED]>
said:
> Can you explain to me how journaling really helps here? It was my
> understanding that ultimately Stephen will only do journaling of
> meta-data and then it is not clear to me how to fix these problems.
The journaling code has to impose the extra rule that any newly
allocated file data is always flushed to disk before we commit the
transaction which allocated that data. You will never see stale data.
> I also wonder if POSIX or something like that doesn't actually mandate
> that data blocks can _never_ migrate from one file to another like
> this.
No, POSIX doesn't talk about filesystem recoverability, and doesn't talk
about media synchrosation at all except in the context of O_*SYNC and
f*sync.
--Stephen
