On 03/12/2018 07:49 AM, Daniel P. Berrangé wrote:
An update of:
v1: https://lists.gnu.org/archive/html/qemu-devel/2017-12/msg04618.html
v2: https://lists.gnu.org/archive/html/qemu-devel/2017-12/msg04706.html
v3: https://lists.gnu.org/archive/html/qemu-devel/2017-12/msg04892.html
v4: https://lists.gnu.org/archive/html/qemu-devel/2018-02/msg00950.html
This enables fixing a long standing problem that libvirt has with
starting up QEMU. It has to busy-wait retrying connect() on the QMP
monitor socket until QEMU finally creates & listens on it, but at same
time must be careful to not wait forever if QEMU exits.
With this patch series, libvirt can simply pass in a pre-opened UNIX domain
socket file descriptor, which it can immediately connect to with no busy-wait.
NB, this will generate one expected failure with patchew / checkpatch.pl
Changed in v5:
- Docs typo (Eric)
- Fix errno setting in int parsing code (Eric)
Daniel P. Berrangé (9):
char: don't silently skip tn3270 protocol init when TLS is enabled
cutils: add qemu_strtoi & qemu_strtoui parsers for int/unsigned int
types
sockets: pull code for testing IP availability out of specific test
sockets: strengthen test suite IP protocol availability checks
sockets: move fd_is_socket() into common sockets code
sockets: check that the named file descriptor is a socket
sockets: allow SocketAddress 'fd' to reference numeric file
descriptors
char: refactor parsing of socket address information
char: allow passing pre-opened socket file descriptor at startup
Whose tree should this go through? It's got some QAPI impact, so I'm
okay including it in my QAPI pull request later today, if there is no
better tree...
--
Eric Blake, Principal Software Engineer
Red Hat, Inc. +1-919-301-3266
Virtualization: qemu.org | libvirt.org