(2010/09/04 18:24), Alexander Graf wrote:
On 03.09.2010, at 10:34, Takuya Yoshikawa wrote:
This is the 2nd version of get_dirty_log cleanup.
Changelog:
In version 1, I changed the timing of copy_to_user() in the
powerpc's get_dirty_log by mistake. This time, I've kept the
timing and tests on ppc box now look OK to me!
I seem to get a pretty big chunk of false positives with this set applied. Qemu's vnc server tries to be clever and searches for real changes inside the pages it finds dirty, so the only perceptible thing there is an increased amount of CPU being wasted.
In MOL instead, the VNC server just directly sends pages that are marked dirty. And I get a full new screen update on every update cycle.
Please check with some debugging code if the dirty region really from the dirty bitmap is really only the one that was updated :).
Interesting behavior! This functionality might be more sensitive than I imagined.
I'll investigate with some debugging code as you suggest!
Thank you,
Takuya
--
To unsubscribe from this list: send the line "unsubscribe kvm-ia64" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
