[...]
> Standard BSD writes metadata synchronously. This means it writes the size
> and block lists before the data. On a crash and fsck you get files that are
> correct to fsck (the block info is right) but whose data was never written.
> On Linux you may also not get the file itself if it was just created.
> fsck has to be a little smarter but there is no difference on a data
> integrity issue.
Isn't it that Linux may write data before writing metadata, so it could
write the data, but it doesn't show up in the (on-disk) file? As a result,
FFS might give you a file with garbage (allocated blocks that were never
written to), Linux might give you a short file (or loose it completely). Is
that correct?
Next question, how long are the respective windows of vulnerability to a
crash? To the next sync(2), i.e., on average 15s?
If the above is true, it looks like both are roughly equally vulnerable,
but I'd prefer the Linux failure mode.
-- Horst von Brand vonbrand@sleipnir.valparaiso.cl Casilla 9G, Viņa del Mar, Chile +56 32 672616
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/