On 15.04.2016 16:59, Peter Krempa wrote:
On Fri, Apr 15, 2016 at 16:45:07 +0200, Michal Privoznik wrote:
On 15.04.2016 13:53, Richard W.M. Jones wrote:
In a few places in libvirt we busy-wait for events, for example qemu creating a monitor socket. This is problematic because:
- We need to choose a sufficiently small polling period so that libvirt doesn't add unnecessary delays.
- We need to choose a sufficiently large polling period so that the effect of busy-waiting doesn't affect the system.
The solution to this conflict is to use an exponential backoff.
This patch adds two functions to hide the details, and modifies a few places where we currently busy-wait. --- src/fdstream.c | 10 +++--- src/libvirt_private.syms | 2 ++ src/qemu/qemu_agent.c | 9 +++--- src/qemu/qemu_monitor.c | 10 +++--- src/util/virtime.c | 81 ++++++++++++++++++++++++++++++++++++++++++++++++ src/util/virtime.h | 11 +++++++ 6 files changed, 111 insertions(+), 12 deletions(-)
ACK. Sorry for making you send v5 to such simple patch.
GCC 5.3.0 doesn't like this very much:
qemu/qemu_agent.c:238:8: error: 'ret' may be used uninitialized in this function [-Werror=maybe-uninitialized] if (ret != 0) {
qemu/qemu_monitor.c:369:8: error: 'ret' may be used uninitialized in this function [-Werror=maybe-uninitialized] if (ret != 0) {
The static analyzer considers the possibility that the while loop will never execute, which would be possible in a very strange lockup of the host.
Please initialize ret to -1 before pushing in those two functions.
Peter
I wonder why my compiler (gcc-4.9.3) hadn't caught those ... Well, thanks Peter for pointing it out. Michal