Re: [libvirt] [PATCH 0/9] Slightly rework our streams
On 04/15/2016 09:51 AM, Michal Privoznik wrote:
This is not the big patch set that enables sparse streams. Not
just yet. I'm merely sending first few patches that prepare the
environment for that. These can, however, go in independent of
sparse streams.
Michal Privoznik (9):
daemonClientStream: Use unsigned int to store stream @serial
daemon stream: Prefer bool over unsigned int var:1
daemon stream: Convert @tx in daemonClientStream to bool
daemon stream: Remove useless empty lines from header file
virNetClientCallDispatchStream: Update comment
daemonStreamHandleRead: Rework to follow our coding pattern
Revert "rpc: Fix slow volume download (virsh vol-download)"
virnetclientstream: Process stream messages later
virStream{Recv,Send}All: Increase client buffer
daemon/stream.c | 116 +++++++++++++++----------------
daemon/stream.h | 2 -
src/libvirt-stream.c | 5 +-
src/rpc/virnetclient.c | 6 +-
src/rpc/virnetclientstream.c | 158 +++++++++++++++---------------------------
src/rpc/virnetserverprogram.c | 12 ++--
src/rpc/virnetserverprogram.h | 4 +-
7 files changed, 129 insertions(+), 174 deletions(-)
I'm assuming for patch 2-3, the memory for the int -> bool conversions
is all "local" to the client side...
I've made comments on a few patches... I'm not an expert in this space
by any stretch - hopefully Martin can look at patch 8 (and since it goes
w/ 7 - they're a matched set.
ACK for at at least 1-6... I'm OK with 7&8 (although I had a double
check type question in 8 regarding the while loop and usage of 'want' as
a loop control along with the st->rx that could have been Queue'd.
I'm OK with patch 9 as well - I think it's fine - a bit more memory for
a bit less back and forth. Standard tradeoff to me ;-). Perhaps someone
else may have agita over it though, so lets just be sure no one does
before pushing that one.
John