Re: [PATCHES] Post-special page storage TDE support - Mailing list pgsql-hackers

Hi David,

> I have finished the reworking of this particular patch series, and have tried to
> organize this in such a way that it will be easily reviewable.  It is
> constructed progressively to be able to follow what is happening here.  As such,
> each individual commit is not guaranteed to compile on its own, so the whole
> series would need to be applied before it works. (It does pass CI tests.)
>
> Here is a brief roadmap of the patches; some of them have additional details in
> the commit message describing a little more about them.
>
> These two patches do some refactoring of existing code to make a common place to
> modify the definitions:
>
> v3-0001-refactor-Create-PageUsableSpace-to-represent-spac.patch
> v3-0002-refactor-Make-PageGetUsablePageSize-routine.patch
>
> These two patches add the ReservedPageSize variable and teach PageInit to use to
> adjust sizing accordingly:
>
> v3-0003-feature-Add-ReservedPageSize-variable.patch
> v3-0004-feature-Adjust-page-sizes-at-PageInit.patch
>
> This patch modifies the definitions of 4 symbols to be computed based on
> PageUsableSpace:
>
> v3-0005-feature-Create-Calc-Limit-and-Dynamic-forms-for-f.patch
>
> These following 4 patches are mechanical replacements of all existing uses of
> these symbols; this provides both visibility into where the existing symbol is
> used as well as distinguishing between parts that care about static allocation
> vs dynamic usage.  The only non-mechanical change is to remove the definition of
> the old symbol so we can be guaranteed that all uses have been considered:
>
> v3-0006-chore-Split-MaxHeapTuplesPerPage-into-Limit-and-D.patch
> v3-0007-chore-Split-MaxIndexTuplesPerPage-into-Limit-and-.patch
> v3-0008-chore-Split-MaxHeapTupleSize-into-Limit-and-Dynam.patch
> v3-0009-chore-Split-MaxTIDsPerBTreePage-into-Limit-and-Dy.patch
>
> The following patches are related to required changes to support dynamic toast
> limits:
>
> v3-0010-feature-Add-hook-for-setting-reloptions-defaults-.patch
> v3-0011-feature-Dynamically-calculate-toast_tuple_target.patch
> v3-0012-feature-Add-Calc-options-for-toast-related-pieces.patch
> v3-0013-chore-Replace-TOAST_MAX_CHUNK_SIZE-with-ClusterTo.patch
> v3-0014-chore-Translation-updates-for-TOAST_MAX_CHUNK_SIZ.patch
>
> In order to calculate some of the sizes, we need to include nbtree.h internals,
> but we can't use in front-end apps, so we separate out the pieces we care about
> into a separate include and use that:
>
> v3-0015-chore-Split-nbtree.h-structure-defs-into-an-inter.patch
>
> This is the meat of the patch; provide a common location for these
> block-size-related constants to be computed using the infra that has been set up
> so far.  Also ensure that we are properly initializing this in front end and
> back end code.  A tricky piece here is we have two separate include files for
> blocksize.h; one which exposes externs as consts for optimizations, and one that
> blocksize.c itself uses without consts, which it uses to create/initialized the
> vars:
>
> v3-0016-feature-Calculate-all-blocksize-constants-in-a-co.patch
>
> Add ControlFile and GUC support for reserved_page_size:
>
> v3-0017-feature-ControlFile-GUC-support-for-reserved_page.patch
>
> Add initdb support for reserving page space:
>
> v3-0018-feature-Add-reserved_page_size-to-initdb-bootstra.patch
>
> Fixes for pg_resetwal:
>
> v3-0019-feature-Updates-for-pg_resetwal.patch
>
> The following 4 patches mechanically replace the Dynamic form to use the new
> Cluster variables:
>
> v3-0020-chore-Rename-MaxHeapTupleSizeDynamic-to-ClusterMa.patch
> v3-0021-chore-Rename-MaxHeapTuplesPerPageDynamic-to-Clust.patch
> v3-0022-chore-Rename-MaxIndexTuplesPerPageDynamic-to-Clus.patch
> v3-0023-chore-Rename-MaxTIDsPerBTreePageDynamic-to-Cluste.patch
>
> Two pieces of optimization required for visibility map:
>
> v3-0024-optimization-Add-support-for-fast-non-division-ba.patch
> v3-0025-optimization-Use-fastdiv-code-in-visibility-map.patch
>
> Update bufpage.h comments:
>
> v3-0026-doc-update-bufpage-docs-w-reserved-space-data.patch
>
> Fixes for bloom to use runtime size:
>
> v3-0027-feature-Teach-bloom-about-PageUsableSpace.patch
>
> Fixes for FSM to use runtime size:
>
> v3-0028-feature-teach-FSM-about-reserved-page-space.patch
>
> I hope this makes sense for reviewing, I know it's a big job, so breaking things up a little more and organizing will
hopefullyhelp.
 

Just wanted to let you know that the patchset seems to need a rebase,
according to cfbot.

Best regards,
Aleksander Alekseev (wearing a co-CFM hat)