On Sun, Dec 15, 2019 at 02:39:20AM +0000, Wang, Huaqiang wrote:
Hi Daniel,
Thanks for your review.
About patch 1/5, I understand we should be very cautious when we changes the determined
interface.
I'd like to reserved the old 32bit interface and propose a new 64bit interface just
as you
suggested if you still think so after you got my intention.
Here actually I want to fix a bug, maybe I should title this patch as *bug fixing*.
Reason is the
underlying hardware, the cache monitor and memory bandwidth counters, are 64bit width.
Using 32bit interface to access these counters are problematic. This bug is not found
because
this interface is only used for tracking the amount of cache that used before this patch,
normally
the occupied cache will not exceed 4GB range. (32bit counter can counter value up to
4GB).
But for memory, this counter records the data passing through the memory controller
issued
by this CPU in bytes and accumulatively, this value can easily exceed the 4GB bound, so
I
don’t want to reserve the old 32 bit interface and let user use it, because it will
report incorrect value.
We simply have to document the limitati onof the old interface. We can
*NOT* change it, because it WILL break API compatibility for apps that
deserailize the current data.
Regards,
Daniel
--
|:
https://berrange.com -o-
https://www.flickr.com/photos/dberrange :|
|:
https://libvirt.org -o-
https://fstop138.berrange.com :|
|:
https://entangle-photo.org -o-
https://www.instagram.com/dberrange :|