On Thu, Dec 08, 2005 at 10:57:35AM -0600, Rob Landley wrote:
[...]
> What we really want are write barriers (which I don't believe Linux exposes
> to
> userspace but I'm not sure), and for UBD to propogate through actual fsync
> requests to the underlying OS.
I did some work on this six weeks or so ago, on the
(non-AIO) ubd implementation. You need to switch on write
barriers explicitly for ext3 (for some reason marking
block dev requests as barriers is off by default, even on
block devices which advertise support for barriers),
change ubd to advertise same, and implement barrier
requests as fdatasync(2). I *think* this is correct,
assuming that the host (and its storage hardware) honours
fdatasync properly. I published a patch, I think -- search
the uml-devel list archives.
You certainly don't want to run ubd in non-sync mode for
important data without this, unless you don't care much
about your data.
--
Don't argue with idiots-- they drag you down to their level,
and then they win because they're playing on home ground.
-------------------------------------------------------
This SF.net email is sponsored by: Splunk Inc. Do you grep through log files
for problems? Stop! Download the new AJAX search engine that makes
searching your log files as easy as surfing the web. DOWNLOAD SPLUNK!
http://ads.osdn.com/?ad_id=7637&alloc_id=16865&op=click
_______________________________________________
User-mode-linux-user mailing list
User-mode-linux-user@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user
