[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [PATCH RFC 09/10] migration/dirtyrate: Expand dirty_bitmap to be tra
From: |
Markus Armbruster |
Subject: |
Re: [PATCH RFC 09/10] migration/dirtyrate: Expand dirty_bitmap to be tracked separately for devices |
Date: |
Mon, 02 May 2022 14:54:55 +0200 |
User-agent: |
Gnus/5.13 (Gnus v5.13) Emacs/27.2 (gnu/linux) |
Joao Martins <joao.m.martins@oracle.com> writes:
> Expand dirtyrate measurer that is accessible via HMP calc_dirty_rate
> or QMP 'calc-dirty-rate' to receive a @scope argument. The scope
> then restricts the dirty tracking to be done at devices only,
> while neither enabling or using the KVM (CPU) dirty tracker.
> The default stays as is i.e. dirty-ring / dirty-bitmap from KVM.
>
> This is useful to test, exercise the IOMMU dirty tracker and observe
> how much a given device is dirtying memory.
>
> Signed-off-by: Joao Martins <joao.m.martins@oracle.com>
[...]
> diff --git a/qapi/migration.json b/qapi/migration.json
> index 27d7b281581d..082830c6e771 100644
> --- a/qapi/migration.json
> +++ b/qapi/migration.json
> @@ -1793,6 +1793,19 @@
> { 'enum': 'DirtyRateMeasureMode',
> 'data': ['page-sampling', 'dirty-ring', 'dirty-bitmap'] }
>
> +##
> +# @DirtyRateScope:
> +#
> +# An enumeration of scope of measuring dirtyrate.
"dirtyrate" is not a word.
> +#
> +# @dirty-devices: calculate dirtyrate by devices only.
Please document @all, too.
> +#
> +# Since: 6.2
> +#
> +##
> +{ 'enum': 'DirtyRateScope',
> + 'data': ['all', 'dirty-devices'] }
> +
> ##
> # @DirtyRateInfo:
> #
> @@ -1827,6 +1840,7 @@
> 'calc-time': 'int64',
> 'sample-pages': 'uint64',
> 'mode': 'DirtyRateMeasureMode',
> + 'scope': 'DirtyRateScope',
Please document new member @scope.
> '*vcpu-dirty-rate': [ 'DirtyRateVcpu' ] } }
>
> ##
> @@ -1851,6 +1865,7 @@
> ##
> { 'command': 'calc-dirty-rate', 'data': {'calc-time': 'int64',
> '*sample-pages': 'int',
> + '*scope': 'DirtyRateScope',
> '*mode': 'DirtyRateMeasureMode'} }
>
> ##
[...]
- Re: [PATCH RFC 09/10] migration/dirtyrate: Expand dirty_bitmap to be tracked separately for devices,
Markus Armbruster <=