On Mon, Jun 26, 2023 at 5:30 PM David Howells [off-list ref] wrote:

Ilya Dryomov [off-list ref] wrote:

quoted

write_partial_message_data() is net/ceph/messenger_v1.c specific, so it
doesn't apply here.  I would suggest squashing the two net/ceph patches
into one since even the titles are the same.

I would, but they're now applied to net-next, so we need to patch that.

I don't see a problem with that given that the patches themselves have
major issues (i.e. it's not just a commit message/title nit).

quoted

quoted

  * Write as much as possible.  The socket is expected to be corked,
- * so we don't bother with MSG_MORE/MSG_SENDPAGE_NOTLAST here.
+ * so we don't bother with MSG_MORE here.
  *
  * Return:
- *   1 - done, nothing (else) to write
+ *  >0 - done, nothing (else) to write

It would be nice to avoid making tweaks like this to the outer
interface as part of switching to a new internal API.

Ok.  I'll change that and wrap the sendmsg in a loop.  Though, as I asked in
an earlier reply, why is MSG_DONTWAIT used here?

See my reply there.

quoted

quoted

+       if (WARN_ON(!iov_iter_is_bvec(&con->v2.out_iter)))
+               return -EINVAL;

Previously, this WARN_ON + error applied only to the "try sendpage"
path.  There is a ton of kvec usage in net/ceph/messenger_v2.c, so I'm
pretty sure that placing it here breaks everything.

This should have been removed as MSG_SPLICE_PAGES now accepts KVEC and XARRAY
iterators also.

Btw, is it feasible to use con->v2.out_iter_sendpage to apply MSG_SPLICE_PAGES
to the iterator to be transmitted as a whole?  It seems to be set depending on
iterator type.

I'm not sure I understand what you mean by "transmitted as a whole".
con->v2.out_iter_sendpage is set only when zerocopy is desired.  If the
underlying data is not guaranteed to remain stable, zerocopy behavior
is not safe.

Thanks,

                Ilya