在 2022/1/18 9:00, Peter Xu 写道:
On Mon, Jan 17, 2022 at 10:00:57PM +0800, Hyman Huang wrote:
This algorithm seems works even worse than the previous version, could you have
a look on what's wrong?
What number the dirty-ring-size of qemu be configured? is it the same as
previous version test?
It should be the same 4096.
The test environment can be slightly different, I used a larger guest this time
(20G, 40 cores), though. Previously it should be a few gig only with a few
cores.
Ok, i'll work this out.
Side note: would you also consider picking up this patch along with the series?
Of course yes, i think this patch can reduce time overhead of
memory_global_dirty_log_start/memory_global_dirty_log_stop but need some
migration tests, i'll do that and once all these be ready, i'll cherry
pick the series before dirtylimit patchset.
https://lore.kernel.org/qemu-devel/20211130080028.6474-1-pet...@redhat.com/
I wanted to make sure it lands before this series, e.g., when enabled both
dirty limit and migration, disabling dirty limit might trigger the bug already.
> Thanks,
--
Best regard
Hyman Huang(黄勇)
