Nathan Coulson wrote: > Just had a few thoughts that could enhance the page (or possibly make > it worse if I misunderstood linux aio. I imagine there's a reason why > it was as hard to find as it was). > > > > linux Asynchronous I/O > (http://ftp.debian.org/debian/pool/main/liba/libaio/libaio_0.3.109.orig.tar.gz) > > I have not found a modern homepage for it (best I found was > http://lse.sourceforge.net/io/aio.html), but it sounds like it can > give a io boost to storage devices under qemu, when your storage > devices is a partition. Also sounds like it can cause corruption when > your storage device is a loopback image/file. > > Installed via make prefix=/usr install. > > Automatically detected by qemu, or manually enabled by > --enable-linux-aio, and used by the client when you call -drive > file=...,aio=native > > Note about possible corruption using aio=native on file disk images: > https://access.redhat.com/site/articles/41313 > > (possibly more info then we want to cover on the page, but *shrug* > can't hurt in this email. My actual recommendation is to provide a > link to libaio as an optional dependency. If we add a note saying it > can be activated with aio=native, I would also recommend mentioning > that the user use it solely on actual partitions [which will be fine > for a lvm based system, probably a headache otherwise]).
There are so many optional dependencies for qemu that are not in the book, I didn't list them. I note that libaio is listed as an optional dependency for mysql. > For qemu-ifup and qemu-ifdown, we use >> (append) to create them. We > could use > to create a new file instead (A user will get the same > results if they run it multiple times, there are no benefits to > appending to those files) Yes, that looks like an oversight. I've fixed it in my sandbox and it will be updated at my next commit. > Also there is the spice project, (which has given me more problems > then it claims to solve so far [probably my fault]), which may be > worth linking to. http://spice-space.org/. I think it can allow you > to share clipboards, and provides a option for paravirtualized > graphics [which is the part I was experimenting unsuccessfully with on > a windows 64bit client] Something to consider, but I think that would end up as a page by itself in the virtualization chapter. -- Bruce -- http://linuxfromscratch.org/mailman/listinfo/blfs-dev FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
