On Wednesday 17 October 2007 09:48, Eric W. Biederman wrote:
quoted text
> Nick Piggin <nickpiggin@yahoo.com.au> writes:
> > On Wednesday 17 October 2007 07:28, Theodore Tso wrote:
> >> On Tue, Oct 16, 2007 at 05:47:12PM +1000, Nick Piggin wrote:
> >> > +	/*
> >> > +	 * ram device BLKFLSBUF has special semantics, we want to actually
> >> > +	 * release and destroy the ramdisk data.
> >> > +	 */
> >>
> >> We won't be able to fix completely this for a while time, but the fact
> >> that BLKFLSBUF has special semantics has always been a major wart.
> >> Could we perhaps create a new ioctl, say RAMDISKDESTORY, and add a
> >> deperecation printk for BLKFLSBUF when passed to the ramdisk?  I doubt
> >> there are many tools that actually take advantage of this wierd aspect
> >> of ramdisks, so hopefully it's something we could remove in a 18
> >> months or so...
> >
> > It would be nice to be able to do that, I agree. The new ramdisk
> > code will be able to flush the buffer cache and destroy its data
> > separately, so it can actually be implemented.
>
> So the practical problem are peoples legacy boot setups but those
> are quickly going away.

After that, is the ramdisk useful for anything aside from testing?


quoted text
> The sane thing is probably something that can be taken as a low
> level format command for the block device.
>
> Say: dd if=/dev/zero of=/dev/ramX

We have 2 problems. First is that, for testing/consistency, we
don't want BLKFLSBUF to throw out the data. Maybe hardly anything
uses BLKFLSBUF now, so it could be just a minor problem, but still
one to fix.

Second is actually throwing out the ramdisk data. dd from /dev/null
isn't trivial because it isn't a "command" from the kernel's POV.
rd could examine the writes to see if they are zero and page aligned,
I suppose... but if you're transitioning everyone over to a new
method anyway, might as well make it a nice one ;)


quoted text
> I know rewriting the drive with all zeroes can cause a modern
> disk to redo it's low level format.  And that is something
> we can definitely implement without any backwards compatibility
> problems.
>
> Hmm. Do we have anything special for punching holes in files?
> That would be another sane route to take to remove the special
> case for clearing the memory.

truncate_range, I suppose. A file descriptor syscall based
alternative for madvise would be nice though (like fallocate).

We could always put something in /sys/block/ram*/xxx
-
unsubscribe notice
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/