This is a followup to Daniel's review in thread
https://www.redhat.com/archives/libvir-list/2019-December/msg00978.html.
Hi Daniel,
In this series, patch 1/2 is trying to fix a bug that using the 64bit,
instead of the current 32bit, interface, to hold cache monitor value
reported by underlying 64bit counter.
In last series, you have suggested
of adding a new field such as "cpu.cache.monitor.%zu.bank.%zu.bytes64"
for new 64bit counter result, and keep the old but already exported 32bit
field, to avoid reporting a truncated cache monitor result to user.
But considering the fact that, in reality, CPU cache size will never
exceed 4GB in size, it is safe to truccate a 64bit counter to 32 bit
since the counter is counting for the cache size that vcpus are using.
Change List:
v2:
* Using old "cpu.cache.monitor.%zu.bank.%zu.bytes" field for cache
monitor result, and keeping the interface consistent.
* Add 'virsh domstats --memory' document in file 'virsh.rst'
Wang Huaqiang (2):
util, resctrl: using 64bit interface instead of 32bit for counters
Introduce command 'virsh domstats --memory' for reporting memory BW
docs/manpages/virsh.rst | 22 +++++-
include/libvirt/libvirt-domain.h | 1 +
src/libvirt-domain.c | 21 ++++++
src/qemu/qemu_driver.c | 118 ++++++++++++++++++++++++++++++-
src/util/virfile.c | 40 +++++++++++
src/util/virfile.h | 2 +
src/util/virresctrl.c | 6 +-
src/util/virresctrl.h | 2 +-
tools/virsh-domain-monitor.c | 7 ++
9 files changed, 211 insertions(+), 8 deletions(-)
--
2.23.0