On Mon, Jan 18, 2010 at 06:33:25PM -0800, Linus Torvalds wrote:
> > It means 1056 bytes for your disk image! Looks like a nice
> > improvement ;-) The change is pushed to the repository, so "git pull"
> > is your friend.
>
> Ok, so the above is fine, but since we do end up doing all reads as whole
> blocks anyway (and we always round up to a page size in order to do
> caching well), it doesn't really matter _that_ much once you get down to a
> single page (4kB) reads.
>
> Of course, depending on alignment etc, smaller reads can still be better
> (ie if that avoids crossing a page boundary), but realistically there are
> very much diminishing returns. A 4kB-aligned 4kB block is going to do as
> much IO as a 32-byte block does (assuming the 32-byte one doesn't cross a
> page).
>
> Also, I suspect that it's not a great idea to optimize for a totally empty
> medium. Sure, it's not a totally uncommon case, but I suspect that it's
The FS label (name) is usually the first entry in the root directory.
It means that read 32 bytes should be enough for non-empty disks too.
The problem is when there is not a FS label, then it reads all
entries.
> less common than having a couple of root directory entries (camera devices
> will almost always have exactly one subdirectory entry etc).
>
> But what you quote above will certainly work well for my case.
>
> Realistically, I'd suggest going for a 4kB blocking factor, though. That's
> likely a good mid-way point between "read all of it whether it's useful or
> not" and "assume it's almost empty".
Yes, I thought about it. Maybe I'll try it.
> > > However, the real problem is that you didn't fix the whole-disk case:
> >
> > Yes, ignore this problem for now. The ideal solution is not to call
> > blkid for the whole disk. We have to resolve this problem.
>
> So in the meantime, what's a good rule to disable it on a per-machine
> basis?
I hope I will disable UDF and UFS probing for small devices and then
blkid for your whole-disk will read 2048 bytes only.
If you want to disable it at all you need to modify udev rules :-(
Ask Kay for more details...
Karel
--
Karel Zak <kzak@xxxxxxxxxx>
--
To unsubscribe from this list: send the line "unsubscribe util-linux-ng" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
