While not directly related to NBD_CMD_WRITE_ZEROES, the qemu team discovered that it is useful if a server can advertise whether an export is in a known-all-zeroes state at the time the client connects.
Signed-off-by: Eric Blake <ebl...@redhat.com> --- doc/proto.md | 5 +++++ 1 file changed, 5 insertions(+) This replaces the following qemu patch attempt: https://lists.gnu.org/archive/html/qemu-devel/2016-12/msg00357.html which tried to add NBD_CMD_HAS_ZERO_INIT with poor semantics. The semantics in this proposal should be much better. Patch is to the merge of the master branch and the extension-write-zeroes branch. By the way, qemu 2.8 is due to be released "real soon now", and implements NBD_CMD_WRITE_ZEROES, so maybe it is time to consider promoting the extension-write-zeroes branch into master. diff --git a/doc/proto.md b/doc/proto.md index afe71fc..7e4ec7f 100644 --- a/doc/proto.md +++ b/doc/proto.md @@ -697,6 +697,11 @@ The field has the following format: the export. - bit 9, `NBD_FLAG_SEND_BLOCK_STATUS`: defined by the experimental `BLOCK_STATUS` [extension](https://github.com/NetworkBlockDevice/nbd/blob/extension-blockstatus/doc/proto.md). +- bit 10, `NBD_FLAG_INIT_ZEROES`: Indicates that the server guarantees + that at the time transmission phase begins, all offsets within the + export read as zero bytes. Clients MAY use this information to + avoid writing to sections of the export that should still read as + zero after the client is done writing. Clients SHOULD ignore unknown flags. -- 2.9.3