Il 19/03/2013 12:44, Alex Bligh ha scritto: > If this isn't true, we have a problem anyway with (e.g.) emulated > devices which don't use O_DIRECT anyway.
Yes, though that would be a libxl bug, not a QEMU bug. > And I had thought (I may be > wrong) using O_DIRECT does not guarantee no read caching with NFS; > O_DIRECT merely guarantees the page cache is not used under Linux > and isn't defined under POSIX: > http://www.spinics.net/lists/linux-nfs/msg17472.html Read caching on the server is fine, because it is the same server that was used for the writes. O_DIRECT bypasses the client's page cache, and that's enough for our purposes. > If it were just a write caching issue, we could use O_DSYNC instead of > O_DIRECT, which would at least ensure the copy from userspace. O_DSYNC is not necessary. We do issue the appropriate fsync/fdatasync. What O_DSYNC does is add an implicit fdatasync after every write, basically. Paolo
