[PATCH 0/2] More clarification on maximum limits

[Eric Blake <eblake@redhat.com>]

Edgar recently proposed a patch to qemu that would allow using
larger WRITE_ZEROES requests than the maximum block size advertised
during NBD_OPT_GO, but I had questions whether that would be
portable across other servers/clients.  Here's my attempt at a
documentation patch update that would officially sanction the
different limits between NBD_CMD_WRITE (often capped at 32M due
to denial-of-service concerns) and NBD_CMD_WRITE_ZEROES (often
a much larger limit, and even if it is overlarge, servers will
still give a graceful EINVAL rather than a hard disconnect).

The first patch is less controversial than the second, but I'm
hoping that we can accept both.  Since NBD_OPT_GO is already
mainline, I'm proposing these patches directly on mainline,
rather than in an extension-* branch; if needed, I can quickly
post qemu patches for both client and server which would implement
the second patch if we want an existing implementation before
accepting the patch.

Eric Blake (2):
  doc: Clarify maximum size limits
  doc: Add alternate trim/zero limits

 doc/proto.md | 108 +++++++++++++++++++++++++++++++++++++++++++++--------------
 1 file changed, 82 insertions(+), 26 deletions(-)

-- 
2.14.3