On Mon, Aug 19, 2013 at 12:13:14PM +0200, Geert Uytterhoeven wrote: > Hi Richard, > > On Mon, Aug 19, 2013 at 11:18 AM, Richard W.M. Jones <[email protected]> > wrote: > > On Sun, Aug 18, 2013 at 12:09:34AM +0200, Richard Weinberger wrote: > >> UML's block device driver does not support write barriers, > >> to support this this patch adds REQ_FLUSH suppport. > >> Every time the block layer sends a REQ_FLUSH we fsync() now > >> our backing file to guarantee data consistency. > > > > This fixes the sync problem I saw before. So: > > What's the impact on your performance figures?
It just moves the sync into the main process, thus making the "write-no-upload" test time *increase* to what we expected (the same as KVM). This doesn't address the main performance issue, which appears to lie in the serial port emulation. Rich. -- Richard Jones, Virtualization Group, Red Hat http://people.redhat.com/~rjones virt-df lists disk usage of guests without needing to install any software inside the virtual machine. Supports Linux and Windows. http://people.redhat.com/~rjones/virt-df/ -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
