public inbox for [email protected]
help / color / mirror / Atom feedFrom: Melanie Plageman <[email protected]>
To: Andres Freund <[email protected]>
Cc: Andrey Borodin <[email protected]>
Cc: Kirill Reshke <[email protected]>
Cc: Chao Li <[email protected]>
Cc: Xuneng Zhou <[email protected]>
Cc: Robert Haas <[email protected]>
Cc: PostgreSQL Hackers <[email protected]>
Cc: Heikki Linnakangas <[email protected]>
Subject: Re: eliminate xl_heap_visible to reduce WAL (and eventually set VM on-access)
Date: Wed, 28 Jan 2026 18:16:10 -0500
Message-ID: <CAAKRu_bs+gZ83QDacmBxunPvCGnXJ05hxP2BDPJ3BGwdbGRXzg@mail.gmail.com> (raw)
In-Reply-To: <7ib3sa55sapwjlaz4sijbiq7iezna27kjvvvar4dpgkmadml6t@gfpkkwmdnepx>
References: <CABPTF7X1X810eT7hjRz9M69sjMv1Vui5gXQEGhqYXmkmywhHCQ@mail.gmail.com>
<CAAKRu_ZcK+ez71W7j+QzMo0gSafTsbGA3TU-fVSptw7yh41BEQ@mail.gmail.com>
<[email protected]>
<CAAKRu_ZCjHoRPfQ8AbMrFY8TOMCPAvZ0_m9SX7yg0edfTk45-g@mail.gmail.com>
<[email protected]>
<CAAKRu_a04jbDACwzRYwzDND31aPyf7Yvz9TAZrTr=+F5bK1aVA@mail.gmail.com>
<CALdSSPjcv25jmXm29X-MRWZBae6+HwcWfVH1PE8NfD=EMTnkAg@mail.gmail.com>
<CAAKRu_bwtBEzDwemyim1r6yYonw7FTyFr1HXG8vywCe-MdbPBQ@mail.gmail.com>
<[email protected]>
<CAAKRu_YQd=2KvomM+RHcpeDKj0bq+peJ=3W-fip+pkvzA-Jq9w@mail.gmail.com>
<7ib3sa55sapwjlaz4sijbiq7iezna27kjvvvar4dpgkmadml6t@gfpkkwmdnepx>
Thanks for the review!
I pushed v33 0001-0003 after incorporating your feedback.
On Fri, Jan 23, 2026 at 7:28 PM Andres Freund <[email protected]> wrote:
>
> On 2026-01-06 12:31:57 -0500, Melanie Plageman wrote:
>
> > + */
> > + if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
> > + {
> > + vacrel->vm_new_visible_pages++;
> > + if (presult.all_frozen)
> > + {
> > + vacrel->vm_new_visible_frozen_pages++;
> > + *vm_page_frozen = true;
>
> Not this patches fault, but I find "vm_new_visible_pages" and
> "vm_new_visible_frozen_pages" pretty odd names. The concept is all-visible and
> frozen. The page itself isn't visible or invisible...
I thought having the extra word "all" in there made it too long. And
since "vm" is there, that isn't set unless the page is
_all_-visible/all-frozen. But if you think it gives people the wrong
idea, I am willing to change it. I can omit vm and make it:
new_all_visible_all_frozen_pages
new_all_visible_pages
new_all_frozen_pages
Is that clearer?
> > From 5c65e73246b4968ddfa9d3739f53d0d8734b8727 Mon Sep 17 00:00:00 2001
> > From: Melanie Plageman <[email protected]>
> > Date: Tue, 2 Dec 2025 15:07:42 -0500
> > Subject: [PATCH v33 04/16] Set the VM in heap_page_prune_and_freeze()
> >
> > This has no independent benefit. It is meant for ease of review. As of
> > this commit, there is still a separate WAL record emitted for setting
> > the VM after pruning and freezing. But it is easier to review if moving
> > the logic into pruneheap.c is separate from setting the VM in the same
> > WAL record.
>
> It seems a bit noisy to refactor the related code in some of the preceding
> commits and then refactor it into a slightly different shape as part of this
> commit (c.f. heap_page_will_set_vm()).
I understand what you are saying. I don't see a good way to keep it
reviewable otherwise, though.
> It's also a bit odd that a function that sounds rather read-only does stuff
> like clearing VM/all-visible.
I thought about this a lot. Ultimately, I ended up keeping it the way it is.
I think the other option is changing from this:
do_set_vm = heap_page_will_set_vm(&prstate,
params->relation,
blockno, buffer, page,
vmbuffer,
params->reason,
do_prune, do_freeze,
prstate.lpdead_items,
&old_vmbits, &new_vmbits);
to this:
heap_page_prepare_vm_set(&prstate,
params->relation,
blockno, buffer, page,
vmbuffer,
params->reason,
do_prune, do_freeze,
prstate.lpdead_items,
&old_vmbits, &new_vmbits);
do_set_vm = (new_vmbits & VISIBILITYMAP_VALID_BITS) != 0;
or heap_page_plan_vm_set()
heap_page_will_set_vm() has symmetry with heap_page_will_freeze(), the
helper that decides whether or not we will freeze tuples. I like that
symmetry since heap_page_will_set_vm() decides whether or not to set
the VM.
Now, heap_page_plan/prepare_vm_set() does indirectly hint that
something like clearing VM/all-visible could happen -- if you
understand that preparing the VM to have bits set also includes
clearing any existing corruption. And "prepare" or "plan" has more
symmetry with prune_freeze_plan() -- though that function does not
make changes on the page.
Ultimately, clearing the VM/page of corruption is pretty anomalous
from the rest of the code in heap_page_prune_and_freeze(). All other
changes to the page are done in a single critical section at the
bottom of the function.
I could see an argument for moving identify_and_fix_vm_corruption()
out of the helper and into heap_page_prune_and_freeze() but then we'd
have to move visibilitymap_get_status() out too. And that takes away a
lot of the benefit of encapsulating all that logic.
Ultimately, I don't like any of those alternative structures. But if
you prefer the names and return value change I have above
(heap_page_prepare/plan_vm_set()), I'm fine with going with that.
> Why are we not doing fixing up of the page *before* we prune it? It's a bit
> insane that we do the WAL logging for pruning, which in turn will often
> include an FPI, before we do the fixups. The fixes aren't WAL logged, so this
> actually leads to the standby getting further out of sync.
>
> I realize this isn't your mess, but brrr.
Well, after this patch set, clearing the VM does happen before we emit
WAL for pruning. It wouldn't be hard to move the corruption fixups to
the beginning of heap_page_prune_and_freeze() in the new code
structure. But it would split visibility map-related logic into two
parts of heap_page_prune_and_freeze(). Would it be worth it? What
benefit would we get? Do you just feel that it should logically come
first?
> Do we actually forsee a case where only one of HEAP_PAGE_PRUNE_FREEZE |
> HEAP_PAGE_PRUNE_UPDATE_VM would be set?
Yes, when setting the VM on-access, it is too expensive to call
heap_prepare_freeze_tuple() on each tuple. I could work on trying to
optimize it, but it isn't currently viable.
> > - if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
> > + if ((presult.old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0 &&
> > + (presult.new_vmbits & VISIBILITYMAP_ALL_VISIBLE) != 0)
> > {
> > vacrel->vm_new_visible_pages++;
> > - if (presult.all_frozen)
> > + if ((presult.new_vmbits & VISIBILITYMAP_ALL_FROZEN) != 0)
> > {
> > vacrel->vm_new_visible_frozen_pages++;
> > *vm_page_frozen = true;
> > }
> > }
> > - else if ((old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0 &&
> > - presult.all_frozen)
> > + else if ((presult.old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0 &&
> > + (presult.new_vmbits & VISIBILITYMAP_ALL_FROZEN) != 0)
> > {
> > + Assert((presult.new_vmbits & VISIBILITYMAP_ALL_VISIBLE) != 0);
> > vacrel->vm_new_frozen_pages++;
> > *vm_page_frozen = true;
> > }
>
> It's a bit odd that we figure out all of this by inspecting old/new vmbits and
> have that logic in multiple places.
I changed PruneFreezeResult to have just the counters that have to be
reflected in vacrel for logging (vm_new_frozen_pages, etc) instead of
passing the bits back.
> > Subject: [PATCH v33 05/16] Move VM assert into prune/freeze code
>
> Feels like a somewhat too narrow description, given that it changes the API
> for heap_page_prune_and_freeze() by removing variables from PruneFreezeResult.
I've tried to fix this.
> > +/*
> > + * Wrapper for heap_page_would_be_all_visible() which can be used for callers
> > + * that expect no LP_DEAD on the page. Currently assert-only, but there is no
> > + * reason not to use it outside of asserts.
> > + */
>
> If so, why would we want it in pruneheap.c? Seems a bit odd to have
> heap_page_would_be_all_visible() defined in vacuumlazy.c but defined
> heap_page_is_all_visible() in pruneheap.c.
You're right. I've kept them both in vacuumlazy.c
> > From cdf5776fadeae3430c692999b37f8a7ec944bda1 Mon Sep 17 00:00:00 2001
> > From: Melanie Plageman <[email protected]>
> > Date: Tue, 2 Dec 2025 16:16:22 -0500
> > +static TransactionId
> > +get_conflict_xid(bool do_prune, bool do_freeze, bool do_set_vm,
> > + uint8 old_vmbits, uint8 new_vmbits,
> > + TransactionId latest_xid_removed, TransactionId frz_conflict_horizon,
> > + TransactionId visibility_cutoff_xid)
> > +{
>
> The logic for horizons is now split between this and "Calculate what the
> snapshot conflict horizon should be for a record" in heap_page_will_freeze().
That is true in master too. We determine frz_conflict_horizon in
heap_page_will_freeze() and later before emitting the WAL record
decide which of the latest_xid_removed and frz_conflict_horizon that
we should use as the snapshot conflict horizon for the combined
record.
All I've done is expand that part (the part before emitting the WAL
record) a bit because now we have to consider what the horizon would
be if we set the VM.
If I really wanted to calculate it only in a single place, I could
maintain a new variable, all_frozen_except_dead, and remove the
frz_conflict_horizon from heap_page_will_freeze(). Then, in
get_conflict_xid(), I could have the following logic:
if (do_set_vm)
conflict_xid = visibility_cutoff_xid;
else if (do_freeze)
{
if (all_frozen_except_dead)
conflict_xid = visibility_cutoff_xid;
else
{
conflict_xid = OldestXmin;
TransactionIdRetreat(conflict_xid);
}
}
else
conflict_xid = InvalidTransactionId;
I think using all_frozen_except_dead while maintaining
visibility_cutoff_xid (in heap_prune_record_unchanged_lp_normal()) has
the potential to be confusing, though. We'd need to keep updating
visibility_cutoff_xid when all_visible is false but
all_frozen_except_dead is true as well as when all_visible is true.
And because we don't care about all_visible_except_dead, it gets even
more confusing to make sure we are maintaining the right variables in
the right situations.
Alternatively, we could keep maintenance of visibility_cutoff_xid the
same and only use all_frozen_except_dead to avoid having the conflict
xid calculation in two places. We would just set it after
prune_freeze_plan() and use it the way it is in the snippet above. But
I don't know if this is better than just having a separate freeze
conflict horizon calculated in the will_freeze code. It is just as
confusing to understand and just as many variables but in a different
place.
For now, I've kept it as is.
> Although I guess I don't understand that code:
>
> /*
> * Calculate what the snapshot conflict horizon should be for a record
> * freezing tuples. We can use the visibility_cutoff_xid as our cutoff
> * for conflicts when the whole page is eligible to become all-frozen
> * in the VM once we're done with it. Otherwise, we generate a
> * conservative cutoff by stepping back from OldestXmin.
> */
> if (prstate->all_frozen)
> prstate->frz_conflict_horizon = prstate->visibility_cutoff_xid;
> else
> {
> /* Avoids false conflicts when hot_standby_feedback in use */
> prstate->frz_conflict_horizon = prstate->cutoffs->OldestXmin;
> TransactionIdRetreat(prstate->frz_conflict_horizon);
> }
>
> Why does it make sense to use OldestXmin? Consider e.g. the case where there
> is one very old tuple that needs to be frozen and one new live tuple on a
> page. Because of the new tuple we can't mark the page all-frozen. But there's
> also no reason to not use much less aggressive horizon than OldestXmin, namely
> the newer of xmin,xmax of the old frozen tuple?
We don't track the newest frozen xmin right now. Doing so wouldn't be
free (i.e. more comparisons which may matter in a query without much
other overhead). And we can't get rid of any of the other cutoffs we
track. We'd still need to maintain the visibility_cutoff_xid and
latest_removed_xid. Which also means it doesn't simplify the code.
The only purpose it would serve is to make the snapshot conflict
horizon more accurate/more aggressive when we freeze tuples, which
would lead to canceling less queries than master -- which is outside
the purview of this patch. There's also a set of complications around
maintaining this number accurately mentioned by Peter in [1].
I've added a new patch in the series 0001 that expands the comment
about this in heap_page_will_freeze() and describes that we are using
a coarse cutoff because we don't track anything else.
But I don't think changing this behavior is a blocker for this feature.
> I also don't understand what the "false conflicts" thing is referencing.
Yea, neither do I. I ported that comment over (it's from before I
started modifying this code) and have never really understood what it
meant -- wouldn't we have more false conflicts if we use a newer
cutoff? (because OldestXmin will be newer than visibility_cutoff_xid).
If the page isn't all-visible, we won't maintain
visibility_cutoff_xid. But if we did actually track the newest live
tuple xmin on the page, that could very well be newer than OldestXmin
and thus using OldestXmin would cancel less and avoid more false
conflicts. But that feels like a _very_ big stretch (in a hypothetical
world that doesn't exist). Anyway, I've deleted the comment (in 0001)
since it clearly is not adding value.
> > + TransactionId conflict_xid;
> > +
> > + /*
> > + * We can omit the snapshot conflict horizon if we are not pruning or
> > + * freezing any tuples and are setting an already all-visible page
> > + * all-frozen in the VM. In this case, all of the tuples on the page must
> > + * already be visible to all MVCC snapshots on the standby.
> > + */
>
> The last sentence here is a bit confusing, because they don't just need to
> already be visible to everyone, they already need to be frozen. Right?
Right. Well that also means that all tuples are visible to all MVCC
snapshots on the standby too -- I picked that language up from
somewhere else in the code and thought it sounded good. But I've
edited it to say frozen, which is more accurate.
> > + if (!do_prune &&
> > + !do_freeze &&
> > + do_set_vm &&
>
> I'm confused by the do_set_vm check here. Doesn't it mean that we will *not*
> return InvalidTransactionId if !prstate->attempt_update_vm? I don't undestand
> why that would make sense.
>
> I guess we'll compute a bogus cutoff in that cse, but never use it, since
> we'll also not emit WAL? Or maybe we'll just unnecessarily go through the
> code below, because the code turns out to do ok regardles? It's confusing
> either way.
Ah no, that was a relic from when I didn't clear the new vmbits in the
event that they were the same as old_vmbits. Now that I do that, I
don't need to qualify it with do_set_vm anymore. I've fixed that.
> > + (old_vmbits & VISIBILITYMAP_ALL_VISIBLE) != 0 &&
> > + (new_vmbits & VISIBILITYMAP_ALL_FROZEN) != 0)
>
> I wonder if some of this code would end up cleaner if we tracked the bits we
> intend to add, rather than the target set of bits is.
I played around with this idea. Unfortunately, this specific check
wouldn't be much simpler, as I'd have to check that _only_ the
all-frozen bit is set. And I think that it is probably more clear with
explicit old and new vmbits. Those make it clear that we are setting a
formerly all-visible page all-frozen.
In some of the other places where I use old/new vmbits, I tried only
using the delta/bits that need to be newly set. But, for example, in
visibilitymap_set(), I need both the all-visible and all-frozen bits
to be passed (even if all-visible is already set), because it asserts
that if all-frozen is passed both all-visible and all-frozen are
passed. If I only keep track of the new bits that need to be set, I
would require other logic before visibilitymap_set(). And it makes
sense that visibilitymap_set() requires both because you don't ever
want people setting just all-frozen -- and you don't want the API to
make it easy to do that.
> > + /*
> > + * If we are updating the VM, the conflict horizon is almost always the
> > + * visibility cutoff XID.
> > + *
> > + * Separately, if we are freezing any tuples, as an optimization, we can
> > + * use the visibility_cutoff_xid as the conflict horizon if the page will
> > + * be all-frozen.
>
> What does "as an optimization" mean here?
What I meant was that visibility_cutoff_xid is going to be older than
OldestXmin for an all-frozen page, so it will lead to canceling fewer
queries. So, using it is kind of an "optimization". But I re-read that
comment and it was way too confusing. I actually cut that whole
paragraph because it should be discussed in heap_page_will_freeze()
where we are actually handling it.
> Note that the code actually uses visibility_cutoff_xid even if the page is
> just marked all-visible, but not all-frozen (due to the do_set_vm check being
> earlier)
And that is correct. But the comment (which is now removed) was
misleading, you are right.
> > This is true even if there are LP_DEAD line pointers
> > + * because we ignored those when maintaining the visibility_cutoff_xid.
>
> I must just be missing something because I can't follow this at all. I guess
> it could be correct because we later then add in knowledge of removed xids in
> via the TransactionIdFollows check below? But if that's it, this is extremely
> confusingly worded.
Yes, we add in latest_xid_removed which is what makes it correct. But
I agree that the wording was terrible. I've cut those details from
get_conflict_xid() and kept them where they are relevant in
heap_page_will_freeze().
- Melanie
[1] https://www.postgresql.org/message-id/CAH2-WzkB-Pt3zPeTXvMik6jcJn%2BdcpUqO-tt_hc13bD6sGRLPg%40mail.g...
Attachments:
[text/x-patch] v34-0001-Clarify-some-heap-tuple-freezing-related-comment.patch (3.3K, 2-v34-0001-Clarify-some-heap-tuple-freezing-related-comment.patch)
download | inline diff:
From e76a68f196d96f269d5719b46d8ba1c9b8950870 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Wed, 28 Jan 2026 12:51:26 -0500
Subject: [PATCH v34 01/14] Clarify some heap tuple freezing related comments
Some of the comments about calculating the snapshot conflict horizon for
a record freezing tuples needed further elaboration.
---
src/backend/access/heap/pruneheap.c | 30 +++++++++++++++++++----------
1 file changed, 20 insertions(+), 10 deletions(-)
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index 632c2427952..b9d2b48104c 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -736,17 +736,26 @@ heap_page_will_freeze(Relation relation, Buffer buffer,
heap_pre_freeze_checks(buffer, prstate->frozen, prstate->nfrozen);
/*
- * Calculate what the snapshot conflict horizon should be for a record
- * freezing tuples. We can use the visibility_cutoff_xid as our cutoff
- * for conflicts when the whole page is eligible to become all-frozen
- * in the VM once we're done with it. Otherwise, we generate a
+ * Determine the snapshot conflict horizon for freezing tuples on the
+ * page.
+ *
+ * We don't track the newest xmin that will become frozen, so we must
+ * use a coarser (more conservative) cutoff as the conflict horizon.
+ *
+ * We can use the visibility_cutoff_xid as our cutoff for conflicts
+ * when the whole page is eligible to become all-frozen in the VM once
+ * we're done with it. Otherwise, we generate an even more
* conservative cutoff by stepping back from OldestXmin.
+ *
+ * Ignoring dead items when all other tuples will be frozen allows us
+ * to pick an older horizon (visibility_cutoff_xid will be older than
+ * OldestXmin). We will later adjust this horizon to account for dead
+ * items, moving it forward if the newest removed xid is newer.
*/
if (prstate->all_frozen)
prstate->frz_conflict_horizon = prstate->visibility_cutoff_xid;
else
{
- /* Avoids false conflicts when hot_standby_feedback in use */
prstate->frz_conflict_horizon = prstate->cutoffs->OldestXmin;
TransactionIdRetreat(prstate->frz_conflict_horizon);
}
@@ -877,11 +886,12 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
/*
* While scanning the line pointers, we did not clear
* all_visible/all_frozen when encountering LP_DEAD items because we
- * wanted the decision whether or not to freeze the page to be unaffected
- * by the short-term presence of LP_DEAD items. These LP_DEAD items are
- * effectively assumed to be LP_UNUSED items in the making. It doesn't
- * matter which vacuum heap pass (initial pass or final pass) ends up
- * setting the page all-frozen, as long as the ongoing VACUUM does it.
+ * wanted the decision whether or not to opportunistically freeze the page
+ * to be unaffected by the short-term presence of LP_DEAD items. These
+ * LP_DEAD items are effectively assumed to be LP_UNUSED items in the
+ * making. It doesn't matter which vacuum heap pass (initial pass or final
+ * pass) ends up setting the page all-frozen, as long as the ongoing
+ * VACUUM does it.
*
* Now that we finished determining whether or not to freeze the page,
* update all_visible and all_frozen so that they reflect the true state
--
2.43.0
[text/x-patch] v34-0002-Set-the-VM-in-heap_page_prune_and_freeze.patch (27.4K, 3-v34-0002-Set-the-VM-in-heap_page_prune_and_freeze.patch)
download | inline diff:
From 9a6cc0cb97af38cf6eedf1a39eede8f3f9926cb8 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Tue, 2 Dec 2025 15:07:42 -0500
Subject: [PATCH v34 02/14] Set the VM in heap_page_prune_and_freeze()
This has no independent benefit. It is meant for ease of review. As of
this commit, there is still a separate WAL record emitted for setting
the VM after pruning and freezing. But it is easier to review if moving
the logic into pruneheap.c is separate from setting the VM in the same
WAL record.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/7ib3sa55sapwjlaz4sijbiq7iezna27kjvvvar4dpgkmadml6t%40gfpkkwmdnepx
---
src/backend/access/heap/pruneheap.c | 318 +++++++++++++++++++++++----
src/backend/access/heap/vacuumlazy.c | 163 +-------------
src/include/access/heapam.h | 20 ++
3 files changed, 303 insertions(+), 198 deletions(-)
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index b9d2b48104c..014c3c92d6c 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -19,7 +19,7 @@
#include "access/htup_details.h"
#include "access/multixact.h"
#include "access/transam.h"
-#include "access/visibilitymapdefs.h"
+#include "access/visibilitymap.h"
#include "access/xlog.h"
#include "access/xloginsert.h"
#include "commands/vacuum.h"
@@ -44,6 +44,8 @@ typedef struct
bool mark_unused_now;
/* whether to attempt freezing tuples */
bool attempt_freeze;
+ /* whether or not to attempt updating the VM */
+ bool attempt_update_vm;
struct VacuumCutoffs *cutoffs;
/*-------------------------------------------------------
@@ -140,16 +142,17 @@ typedef struct
* all_visible and all_frozen indicate if the all-visible and all-frozen
* bits in the visibility map can be set for this page after pruning.
*
- * visibility_cutoff_xid is the newest xmin of live tuples on the page.
- * The caller can use it as the conflict horizon, when setting the VM
- * bits. It is only valid if we froze some tuples, and all_frozen is
- * true.
+ * visibility_cutoff_xid is the newest xmin of live tuples on the page. It
+ * can be used as the conflict horizon when setting the VM or when
+ * freezing all the tuples on the page. It is only valid when all the live
+ * tuples on the page are all-visible.
*
* NOTE: all_visible and all_frozen initially don't include LP_DEAD items.
* That's convenient for heap_page_prune_and_freeze() to use them to
- * decide whether to freeze the page or not. The all_visible and
- * all_frozen values returned to the caller are adjusted to include
- * LP_DEAD items after we determine whether to opportunistically freeze.
+ * decide whether to opportunistically freeze the page or not. The
+ * all_visible and all_frozen values ultimately used to set the VM are
+ * adjusted to include LP_DEAD items after we determine whether or not to
+ * opportunistically freeze.
*/
bool all_visible;
bool all_frozen;
@@ -191,6 +194,17 @@ static void page_verify_redirects(Page page);
static bool heap_page_will_freeze(Relation relation, Buffer buffer,
bool did_tuple_hint_fpi, bool do_prune, bool do_hint_prune,
PruneState *prstate);
+static void identify_and_fix_vm_corruption(Relation rel, Buffer heap_buffer,
+ BlockNumber heap_blk, Page heap_page, int nlpdead_items,
+ Buffer vmbuffer,
+ uint8 *vmbits);
+static bool heap_page_will_set_vm(PruneState *prstate,
+ Relation relation,
+ BlockNumber heap_blk, Buffer heap_buffer, Page heap_page,
+ Buffer vmbuffer,
+ int nlpdead_items,
+ uint8 *old_vmbits,
+ uint8 *new_vmbits);
/*
@@ -280,6 +294,7 @@ heap_page_prune_opt(Relation relation, Buffer buffer)
PruneFreezeParams params = {
.relation = relation,
.buffer = buffer,
+ .vmbuffer = InvalidBuffer,
.reason = PRUNE_ON_ACCESS,
.options = 0,
.vistest = vistest,
@@ -341,6 +356,8 @@ prune_freeze_setup(PruneFreezeParams *params,
/* cutoffs must be provided if we will attempt freezing */
Assert(!(params->options & HEAP_PAGE_PRUNE_FREEZE) || params->cutoffs);
prstate->attempt_freeze = (params->options & HEAP_PAGE_PRUNE_FREEZE) != 0;
+ prstate->attempt_update_vm =
+ (params->options & HEAP_PAGE_PRUNE_UPDATE_VM) != 0;
prstate->cutoffs = params->cutoffs;
/*
@@ -396,51 +413,54 @@ prune_freeze_setup(PruneFreezeParams *params,
prstate->frz_conflict_horizon = InvalidTransactionId;
/*
- * Vacuum may update the VM after we're done. We can keep track of
- * whether the page will be all-visible and all-frozen after pruning and
- * freezing to help the caller to do that.
+ * Track whether the page could be marked all-visible and/or all-frozen.
+ * This information is used for opportunistic freezing and for updating
+ * the visibility map (VM) if requested by the caller.
*
- * Currently, only VACUUM sets the VM bits. To save the effort, only do
- * the bookkeeping if the caller needs it. Currently, that's tied to
- * HEAP_PAGE_PRUNE_FREEZE, but it could be a separate flag if you wanted
- * to update the VM bits without also freezing or freeze without also
- * setting the VM bits.
+ * Currently, only VACUUM performs freezing, but other callers may in the
+ * future. Visibility bookkeeping is required not just for setting the VM
+ * bits, but also for opportunistic freezing: we only consider freezing if
+ * the page would become all-frozen, or if it would be all-frozen except
+ * for dead tuples that VACUUM will remove. If attempt_update_vm is false,
+ * we will not set the VM bit even if the page is found to be all-visible.
*
- * In addition to telling the caller whether it can set the VM bit, we
- * also use 'all_visible' and 'all_frozen' for our own decision-making. If
- * the whole page would become frozen, we consider opportunistically
- * freezing tuples. We will not be able to freeze the whole page if there
- * are tuples present that are not visible to everyone or if there are
- * dead tuples which are not yet removable. However, dead tuples which
- * will be removed by the end of vacuuming should not preclude us from
- * opportunistically freezing. Because of that, we do not immediately
- * clear all_visible and all_frozen when we see LP_DEAD items. We fix
- * that after scanning the line pointers. We must correct all_visible and
- * all_frozen before we return them to the caller, so that the caller
- * doesn't set the VM bits incorrectly.
+ * If HEAP_PAGE_PRUNE_UPDATE_VM is passed without HEAP_PAGE_PRUNE_FREEZE,
+ * prstate.all_frozen must be initialized to false, since we will not call
+ * heap_prepare_freeze_tuple() for each tuple.
+ *
+ * Dead tuples that will be removed by the end of vacuum should not
+ * prevent opportunistic freezing. Therefore, we do not clear all_visible
+ * and all_frozen when we encounter LP_DEAD items. Instead, we correct
+ * them after deciding whether to freeze, but before updating the VM, to
+ * avoid setting the VM bits incorrectly.
+ *
+ * If neither freezing nor VM updates are requested, we skip the extra
+ * bookkeeping. In this case, initializing all_visible to false allows
+ * heap_prune_record_unchanged_lp_normal() to bypass unnecessary work.
*/
if (prstate->attempt_freeze)
{
prstate->all_visible = true;
prstate->all_frozen = true;
}
+ else if (prstate->attempt_update_vm)
+ {
+ prstate->all_visible = true;
+ prstate->all_frozen = false;
+ }
else
{
- /*
- * Initializing to false allows skipping the work to update them in
- * heap_prune_record_unchanged_lp_normal().
- */
prstate->all_visible = false;
prstate->all_frozen = false;
}
/*
- * The visibility cutoff xid is the newest xmin of live tuples on the
- * page. In the common case, this will be set as the conflict horizon the
- * caller can use for updating the VM. If, at the end of freezing and
- * pruning, the page is all-frozen, there is no possibility that any
- * running transaction on the standby does not see tuples on the page as
- * all-visible, so the conflict horizon remains InvalidTransactionId.
+ * The visibility cutoff xid is the newest xmin of live, committed tuples
+ * older than OldestXmin on the page. This field is only kept up-to-date
+ * if the page is all-visible. As soon as a tuple is encountered that is
+ * not visible to all, this field is unmaintained. As long as it is
+ * maintained, it can be used to calculate the snapshot conflict horizon
+ * when updating the VM and/or freezing all the tuples on the page.
*/
prstate->visibility_cutoff_xid = InvalidTransactionId;
}
@@ -784,10 +804,145 @@ heap_page_will_freeze(Relation relation, Buffer buffer,
return do_freeze;
}
+/*
+ * Helper to correct any corruption detected on a heap page and its
+ * corresponding visibility map page after pruning but before setting the
+ * visibility map. It examines the heap page, the associated VM page, and the
+ * number of dead items previously identified.
+ *
+ * This function must be called while holding an exclusive lock on the heap
+ * buffer, and the dead items must have been discovered under that same lock.
+ *
+ * The provided vmbits must reflect the current state of the VM block
+ * referenced by vmbuffer. Although we do not hold a lock on the VM buffer, it
+ * is pinned, and the heap buffer is exclusively locked, ensuring that no
+ * other backend can update the VM bits corresponding to this heap page.
+ *
+ * If it clears corruption, it will zero out vmbits.
+ */
+static void
+identify_and_fix_vm_corruption(Relation rel, Buffer heap_buffer,
+ BlockNumber heap_blk, Page heap_page,
+ int nlpdead_items,
+ Buffer vmbuffer,
+ uint8 *vmbits)
+{
+ Assert(visibilitymap_get_status(rel, heap_blk, &vmbuffer) == *vmbits);
+
+ Assert(BufferIsLockedByMeInMode(heap_buffer, BUFFER_LOCK_EXCLUSIVE));
+
+ /*
+ * As of PostgreSQL 9.2, the visibility map bit should never be set if the
+ * page-level bit is clear. However, it's possible that the bit got
+ * cleared after heap_vac_scan_next_block() was called, so we must recheck
+ * with buffer lock before concluding that the VM is corrupt.
+ */
+ if (!PageIsAllVisible(heap_page) &&
+ ((*vmbits & VISIBILITYMAP_VALID_BITS) != 0))
+ {
+ ereport(WARNING,
+ (errcode(ERRCODE_DATA_CORRUPTED),
+ errmsg("page is not marked all-visible but visibility map bit is set in relation \"%s\" page %u",
+ RelationGetRelationName(rel), heap_blk)));
+
+ visibilitymap_clear(rel, heap_blk, vmbuffer,
+ VISIBILITYMAP_VALID_BITS);
+ *vmbits = 0;
+ }
+
+ /*
+ * It's possible for the value returned by
+ * GetOldestNonRemovableTransactionId() to move backwards, so it's not
+ * wrong for us to see tuples that appear to not be visible to everyone
+ * yet, while PD_ALL_VISIBLE is already set. The real safe xmin value
+ * never moves backwards, but GetOldestNonRemovableTransactionId() is
+ * conservative and sometimes returns a value that's unnecessarily small,
+ * so if we see that contradiction it just means that the tuples that we
+ * think are not visible to everyone yet actually are, and the
+ * PD_ALL_VISIBLE flag is correct.
+ *
+ * There should never be LP_DEAD items on a page with PD_ALL_VISIBLE set,
+ * however.
+ */
+ else if (PageIsAllVisible(heap_page) && nlpdead_items > 0)
+ {
+ ereport(WARNING,
+ (errcode(ERRCODE_DATA_CORRUPTED),
+ errmsg("page containing LP_DEAD items is marked as all-visible in relation \"%s\" page %u",
+ RelationGetRelationName(rel), heap_blk)));
+
+ PageClearAllVisible(heap_page);
+ MarkBufferDirty(heap_buffer);
+ visibilitymap_clear(rel, heap_blk, vmbuffer,
+ VISIBILITYMAP_VALID_BITS);
+ *vmbits = 0;
+ }
+}
+
+/*
+ * Decide whether to set the visibility map bits (all-visible and all-frozen)
+ * for heap_blk using information from the PruneState and VM.
+ *
+ * This function does not actually set the VM bits or page-level visibility
+ * hint, PD_ALL_VISIBLE.
+ *
+ * However, if it finds that the page-level visibility hint or VM is
+ * corrupted, it will fix them by clearing the VM bits and visibility hint.
+ * This does not need to be done in a critical section.
+ *
+ * Returns true if one or both VM bits should be set, along with returning the
+ * current value of the VM bits in *old_vmbits and the desired new value of
+ * the VM bits in *new_vmbits.
+ *
+ * If the VM should not be set, it returns false. If we won't consider
+ * updating the VM, *old_vmbits will be 0, regardless of the current value of
+ * the VM bits.
+ */
+static bool
+heap_page_will_set_vm(PruneState *prstate,
+ Relation relation,
+ BlockNumber heap_blk, Buffer heap_buffer, Page heap_page,
+ Buffer vmbuffer,
+ int nlpdead_items,
+ uint8 *old_vmbits,
+ uint8 *new_vmbits)
+{
+ *old_vmbits = 0;
+ *new_vmbits = 0;
+
+ if (!prstate->attempt_update_vm)
+ return false;
+
+ *old_vmbits = visibilitymap_get_status(relation, heap_blk,
+ &vmbuffer);
+
+ /* We do this even if not all-visible */
+ identify_and_fix_vm_corruption(relation, heap_buffer, heap_blk, heap_page,
+ nlpdead_items, vmbuffer,
+ old_vmbits);
+
+ if (!prstate->all_visible)
+ return false;
+
+ *new_vmbits = VISIBILITYMAP_ALL_VISIBLE;
+
+ if (prstate->all_frozen)
+ *new_vmbits |= VISIBILITYMAP_ALL_FROZEN;
+
+ if (*new_vmbits == *old_vmbits)
+ {
+ *new_vmbits = 0;
+ return false;
+ }
+
+ return true;
+}
+
/*
* Prune and repair fragmentation and potentially freeze tuples on the
- * specified page.
+ * specified page. If the page's visibility status has changed, update it in
+ * the VM.
*
* Caller must have pin and buffer cleanup lock on the page. Note that we
* don't update the FSM information for page on caller's behalf. Caller might
@@ -802,12 +957,13 @@ heap_page_will_freeze(Relation relation, Buffer buffer,
* tuples if it's required in order to advance relfrozenxid / relminmxid, or
* if it's considered advantageous for overall system performance to do so
* now. The 'params.cutoffs', 'presult', 'new_relfrozen_xid' and
- * 'new_relmin_mxid' arguments are required when freezing. When
- * HEAP_PAGE_PRUNE_FREEZE option is passed, we also set presult->all_visible
- * and presult->all_frozen after determining whether or not to
- * opportunistically freeze, to indicate if the VM bits can be set. They are
- * always set to false when the HEAP_PAGE_PRUNE_FREEZE option is not passed,
- * because at the moment only callers that also freeze need that information.
+ * 'new_relmin_mxid' arguments are required when freezing.
+ *
+ * If HEAP_PAGE_PRUNE_UPDATE_VM is set in params and the visibility status of
+ * the page has changed, we will update the VM at the same time as pruning and
+ * freezing the heap page. We will also update presult->old_vmbits and
+ * presult->new_vmbits with the state of the VM before and after updating it
+ * for the caller to use in bookkeeping.
*
* presult contains output parameters needed by callers, such as the number of
* tuples removed and the offsets of dead items on the page after pruning.
@@ -832,13 +988,18 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
MultiXactId *new_relmin_mxid)
{
Buffer buffer = params->buffer;
+ Buffer vmbuffer = params->vmbuffer;
Page page = BufferGetPage(buffer);
+ BlockNumber blockno = BufferGetBlockNumber(buffer);
PruneState prstate;
bool do_freeze;
bool do_prune;
bool do_hint_prune;
+ bool do_set_vm;
bool did_tuple_hint_fpi;
int64 fpi_before = pgWalUsage.wal_fpi;
+ uint8 new_vmbits;
+ uint8 old_vmbits;
/* Initialize prstate */
prune_freeze_setup(params,
@@ -1021,6 +1182,71 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
*new_relmin_mxid = prstate.pagefrz.NoFreezePageRelminMxid;
}
}
+
+ /* Now update the visibility map and PD_ALL_VISIBLE hint */
+ Assert(!prstate.all_visible || (prstate.lpdead_items == 0));
+
+ do_set_vm = heap_page_will_set_vm(&prstate,
+ params->relation,
+ blockno,
+ buffer,
+ page,
+ vmbuffer,
+ prstate.lpdead_items,
+ &old_vmbits,
+ &new_vmbits);
+
+ /*
+ * new_vmbits should be 0 regardless of whether or not the page is
+ * all-visible if we do not intend to set the VM.
+ */
+ Assert(do_set_vm || new_vmbits == 0);
+
+ /* Set the visibility map and page visibility hint, if relevant */
+ if (do_set_vm)
+ {
+ Assert(prstate.all_visible);
+
+ /*
+ * It should never be the case that the visibility map page is set
+ * while the page-level bit is clear (and if so, we cleared it above),
+ * but the reverse is allowed (if checksums are not enabled).
+ * Regardless, set both bits so that we get back in sync.
+ *
+ * The heap buffer must be marked dirty before adding it to the WAL
+ * chain when setting the VM. We don't worry about unnecessarily
+ * dirtying the heap buffer if PD_ALL_VISIBLE is already set, though.
+ * It is extremely rare to have a clean heap buffer with
+ * PD_ALL_VISIBLE already set and the VM bits clear, so there is no
+ * point in optimizing it.
+ */
+ PageSetAllVisible(page);
+ MarkBufferDirty(buffer);
+
+ /*
+ * If the page is being set all-frozen, we pass InvalidTransactionId
+ * as the cutoff_xid, since a snapshot conflict horizon sufficient to
+ * make everything safe for REDO was logged when the page's tuples
+ * were frozen.
+ */
+ Assert(!prstate.all_frozen ||
+ !TransactionIdIsValid(presult->vm_conflict_horizon));
+
+ visibilitymap_set(params->relation, blockno, buffer,
+ InvalidXLogRecPtr,
+ vmbuffer, presult->vm_conflict_horizon,
+ new_vmbits);
+
+ if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
+ {
+ presult->new_all_visible_pages = 1;
+ if (presult->all_frozen)
+ presult->new_all_visible_frozen_pages = 1;
+ }
+ else if ((old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0 &&
+ presult->all_frozen)
+ presult->new_all_frozen_pages = 1;
+ }
}
@@ -1495,6 +1721,8 @@ heap_prune_record_unchanged_lp_normal(Page page, PruneState *prstate, OffsetNumb
{
TransactionId xmin;
+ Assert(prstate->attempt_update_vm);
+
if (!HeapTupleHeaderXminCommitted(htup))
{
prstate->all_visible = false;
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 4be267ff657..323b8e3dde3 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -422,11 +422,6 @@ static void find_next_unskippable_block(LVRelState *vacrel, bool *skipsallvis);
static bool lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf,
BlockNumber blkno, Page page,
bool sharelock, Buffer vmbuffer);
-static void identify_and_fix_vm_corruption(Relation rel, Buffer heap_buffer,
- BlockNumber heap_blk, Page heap_page,
- int nlpdead_items,
- Buffer vmbuffer,
- uint8 *vmbits);
static int lazy_scan_prune(LVRelState *vacrel, Buffer buf,
BlockNumber blkno, Page page,
Buffer vmbuffer,
@@ -1960,81 +1955,6 @@ cmpOffsetNumbers(const void *a, const void *b)
return pg_cmp_u16(*(const OffsetNumber *) a, *(const OffsetNumber *) b);
}
-/*
- * Helper to correct any corruption detected on a heap page and its
- * corresponding visibility map page after pruning but before setting the
- * visibility map. It examines the heap page, the associated VM page, and the
- * number of dead items previously identified.
- *
- * This function must be called while holding an exclusive lock on the heap
- * buffer, and the dead items must have been discovered under that same lock.
-
- * The provided vmbits must reflect the current state of the VM block
- * referenced by vmbuffer. Although we do not hold a lock on the VM buffer, it
- * is pinned, and the heap buffer is exclusively locked, ensuring that no
- * other backend can update the VM bits corresponding to this heap page.
- *
- * If it clears corruption, it will zero out vmbits.
- */
-static void
-identify_and_fix_vm_corruption(Relation rel, Buffer heap_buffer,
- BlockNumber heap_blk, Page heap_page,
- int nlpdead_items,
- Buffer vmbuffer,
- uint8 *vmbits)
-{
- Assert(visibilitymap_get_status(rel, heap_blk, &vmbuffer) == *vmbits);
-
- Assert(BufferIsLockedByMeInMode(heap_buffer, BUFFER_LOCK_EXCLUSIVE));
-
- /*
- * As of PostgreSQL 9.2, the visibility map bit should never be set if the
- * page-level bit is clear. However, it's possible that the bit got
- * cleared after heap_vac_scan_next_block() was called, so we must recheck
- * with buffer lock before concluding that the VM is corrupt.
- */
- if (!PageIsAllVisible(heap_page) &&
- ((*vmbits & VISIBILITYMAP_VALID_BITS) != 0))
- {
- ereport(WARNING,
- (errcode(ERRCODE_DATA_CORRUPTED),
- errmsg("page is not marked all-visible but visibility map bit is set in relation \"%s\" page %u",
- RelationGetRelationName(rel), heap_blk)));
-
- visibilitymap_clear(rel, heap_blk, vmbuffer,
- VISIBILITYMAP_VALID_BITS);
- *vmbits = 0;
- }
-
- /*
- * It's possible for the value returned by
- * GetOldestNonRemovableTransactionId() to move backwards, so it's not
- * wrong for us to see tuples that appear to not be visible to everyone
- * yet, while PD_ALL_VISIBLE is already set. The real safe xmin value
- * never moves backwards, but GetOldestNonRemovableTransactionId() is
- * conservative and sometimes returns a value that's unnecessarily small,
- * so if we see that contradiction it just means that the tuples that we
- * think are not visible to everyone yet actually are, and the
- * PD_ALL_VISIBLE flag is correct.
- *
- * There should never be LP_DEAD items on a page with PD_ALL_VISIBLE set,
- * however.
- */
- else if (PageIsAllVisible(heap_page) && nlpdead_items > 0)
- {
- ereport(WARNING,
- (errcode(ERRCODE_DATA_CORRUPTED),
- errmsg("page containing LP_DEAD items is marked as all-visible in relation \"%s\" page %u",
- RelationGetRelationName(rel), heap_blk)));
-
- PageClearAllVisible(heap_page);
- MarkBufferDirty(heap_buffer);
- visibilitymap_clear(rel, heap_blk, vmbuffer,
- VISIBILITYMAP_VALID_BITS);
- *vmbits = 0;
- }
-}
-
/*
* lazy_scan_prune() -- lazy_scan_heap() pruning and freezing.
*
@@ -2066,13 +1986,12 @@ lazy_scan_prune(LVRelState *vacrel,
PruneFreezeParams params = {
.relation = rel,
.buffer = buf,
+ .vmbuffer = vmbuffer,
.reason = PRUNE_VACUUM_SCAN,
- .options = HEAP_PAGE_PRUNE_FREEZE,
+ .options = HEAP_PAGE_PRUNE_FREEZE | HEAP_PAGE_PRUNE_UPDATE_VM,
.vistest = vacrel->vistest,
.cutoffs = &vacrel->cutoffs,
};
- uint8 old_vmbits = 0;
- uint8 new_vmbits = 0;
Assert(BufferGetBlockNumber(buf) == blkno);
@@ -2159,6 +2078,14 @@ lazy_scan_prune(LVRelState *vacrel,
}
/* Finally, add page-local counts to whole-VACUUM counts */
+ vacrel->vm_new_visible_pages += presult.new_all_visible_pages;
+ vacrel->vm_new_visible_frozen_pages += presult.new_all_visible_frozen_pages;
+ vacrel->vm_new_frozen_pages += presult.new_all_frozen_pages;
+
+ /* Capture if the page was newly set frozen */
+ *vm_page_frozen = presult.new_all_visible_frozen_pages > 0 ||
+ presult.new_all_frozen_pages > 0;
+
vacrel->tuples_deleted += presult.ndeleted;
vacrel->tuples_frozen += presult.nfrozen;
vacrel->lpdead_items += presult.lpdead_items;
@@ -2172,76 +2099,6 @@ lazy_scan_prune(LVRelState *vacrel,
/* Did we find LP_DEAD items? */
*has_lpdead_items = (presult.lpdead_items > 0);
- Assert(!presult.all_visible || !(*has_lpdead_items));
- Assert(!presult.all_frozen || presult.all_visible);
-
- old_vmbits = visibilitymap_get_status(vacrel->rel, blkno, &vmbuffer);
-
- identify_and_fix_vm_corruption(vacrel->rel, buf, blkno, page,
- presult.lpdead_items, vmbuffer,
- &old_vmbits);
-
- if (!presult.all_visible)
- return presult.ndeleted;
-
- /* Set the visibility map and page visibility hint */
- new_vmbits = VISIBILITYMAP_ALL_VISIBLE;
-
- if (presult.all_frozen)
- new_vmbits |= VISIBILITYMAP_ALL_FROZEN;
-
- /* Nothing to do */
- if (old_vmbits == new_vmbits)
- return presult.ndeleted;
-
- /*
- * It should never be the case that the visibility map page is set while
- * the page-level bit is clear (and if so, we cleared it above), but the
- * reverse is allowed (if checksums are not enabled). Regardless, set both
- * bits so that we get back in sync.
- *
- * The heap buffer must be marked dirty before adding it to the WAL chain
- * when setting the VM. We don't worry about unnecessarily dirtying the
- * heap buffer if PD_ALL_VISIBLE is already set, though. It is extremely
- * rare to have a clean heap buffer with PD_ALL_VISIBLE already set and
- * the VM bits clear, so there is no point in optimizing it.
- */
- PageSetAllVisible(page);
- MarkBufferDirty(buf);
-
- /*
- * If the page is being set all-frozen, we pass InvalidTransactionId as
- * the cutoff_xid, since a snapshot conflict horizon sufficient to make
- * everything safe for REDO was logged when the page's tuples were frozen.
- */
- Assert(!presult.all_frozen ||
- !TransactionIdIsValid(presult.vm_conflict_horizon));
-
- visibilitymap_set(vacrel->rel, blkno, buf,
- InvalidXLogRecPtr,
- vmbuffer, presult.vm_conflict_horizon,
- new_vmbits);
-
- /*
- * If the page wasn't already set all-visible and/or all-frozen in the VM,
- * count it as newly set for logging.
- */
- if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
- {
- vacrel->vm_new_visible_pages++;
- if (presult.all_frozen)
- {
- vacrel->vm_new_visible_frozen_pages++;
- *vm_page_frozen = true;
- }
- }
- else if ((old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0 &&
- presult.all_frozen)
- {
- vacrel->vm_new_frozen_pages++;
- *vm_page_frozen = true;
- }
-
return presult.ndeleted;
}
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index 3c0961ab36b..df9fa17a6f9 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -42,6 +42,7 @@
/* "options" flag bits for heap_page_prune_and_freeze */
#define HEAP_PAGE_PRUNE_MARK_UNUSED_NOW (1 << 0)
#define HEAP_PAGE_PRUNE_FREEZE (1 << 1)
+#define HEAP_PAGE_PRUNE_UPDATE_VM (1 << 2)
typedef struct BulkInsertStateData *BulkInsertState;
typedef struct GlobalVisState GlobalVisState;
@@ -238,6 +239,12 @@ typedef struct PruneFreezeParams
Relation relation; /* relation containing buffer to be pruned */
Buffer buffer; /* buffer to be pruned */
+ /*
+ * If we will consider updating the visibility map, vmbuffer should
+ * contain the correct block of the visibility map and be pinned.
+ */
+ Buffer vmbuffer;
+
/*
* The reason pruning was performed. It is used to set the WAL record
* opcode which is used for debugging and analysis purposes.
@@ -252,6 +259,9 @@ typedef struct PruneFreezeParams
*
* HEAP_PAGE_PRUNE_FREEZE indicates that we will also freeze tuples, and
* will return 'all_visible', 'all_frozen' flags to the caller.
+ *
+ * HEAP_PAGE_PRUNE_UPDATE_VM indicates that we will set the page's status
+ * in the VM.
*/
int options;
@@ -299,6 +309,16 @@ typedef struct PruneFreezeResult
bool all_frozen;
TransactionId vm_conflict_horizon;
+ /*
+ * Whether or not the page was newly set all-visible and all-frozen during
+ * phase I of vacuuming.
+ *
+ * These are only set if the HEAP_PAGE_PRUNE_UPDATE_VM option is set.
+ */
+ BlockNumber new_all_visible_pages;
+ BlockNumber new_all_visible_frozen_pages;
+ BlockNumber new_all_frozen_pages;
+
/*
* Whether or not the page makes rel truncation unsafe. This is set to
* 'true', even if the page contains LP_DEAD items. VACUUM will remove
--
2.43.0
[text/x-patch] v34-0003-Move-VM-assert-into-prune-freeze-code-and-simpli.patch (9.4K, 4-v34-0003-Move-VM-assert-into-prune-freeze-code-and-simpli.patch)
download | inline diff:
From e591bf061ee673f3750d1180673e1ab48be43bb8 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Tue, 27 Jan 2026 16:53:11 -0500
Subject: [PATCH v34 03/14] Move VM assert into prune/freeze code and simplify
returned values
After pruning and freezing, we do an assert-only validatation that the
page's visibility status matches what we found during the pruning and
freezing pass over the page.
There's no reason to wait until lazy_scan_prune() to do this validation,
as all of the VM setting logic has already been moved to
heap_page_prune_and_freeze().
Doing so also allows us to remove some fields of PruneFreezeResult,
narrowing the scope of values the caller has to think about.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/7ib3sa55sapwjlaz4sijbiq7iezna27kjvvvar4dpgkmadml6t%40gfpkkwmdnepx
---
src/backend/access/heap/pruneheap.c | 65 +++++++++++++++++++---------
src/backend/access/heap/vacuumlazy.c | 35 +--------------
src/include/access/heapam.h | 26 ++++-------
3 files changed, 54 insertions(+), 72 deletions(-)
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index 014c3c92d6c..192df9a2218 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -991,6 +991,7 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
Buffer vmbuffer = params->vmbuffer;
Page page = BufferGetPage(buffer);
BlockNumber blockno = BufferGetBlockNumber(buffer);
+ TransactionId vm_conflict_horizon = InvalidTransactionId;
PruneState prstate;
bool do_freeze;
bool do_prune;
@@ -1149,23 +1150,8 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
presult->nfrozen = prstate.nfrozen;
presult->live_tuples = prstate.live_tuples;
presult->recently_dead_tuples = prstate.recently_dead_tuples;
- presult->all_visible = prstate.all_visible;
- presult->all_frozen = prstate.all_frozen;
presult->hastup = prstate.hastup;
- /*
- * For callers planning to update the visibility map, the conflict horizon
- * for that record must be the newest xmin on the page. However, if the
- * page is completely frozen, there can be no conflict and the
- * vm_conflict_horizon should remain InvalidTransactionId. This includes
- * the case that we just froze all the tuples; the prune-freeze record
- * included the conflict XID already so the caller doesn't need it.
- */
- if (presult->all_frozen)
- presult->vm_conflict_horizon = InvalidTransactionId;
- else
- presult->vm_conflict_horizon = prstate.visibility_cutoff_xid;
-
presult->lpdead_items = prstate.lpdead_items;
/* the presult->deadoffsets array was already filled in */
@@ -1183,6 +1169,46 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
}
}
+ /*
+ * If updating the visibility map, the conflict horizon for that record
+ * must be the newest xmin on the page. However, if the page is
+ * completely frozen, there can be no conflict and the vm_conflict_horizon
+ * should remain InvalidTransactionId. This includes the case that we
+ * just froze all the tuples; the prune-freeze record included the
+ * conflict XID already so we don't need to again.
+ */
+ if (prstate.all_frozen)
+ vm_conflict_horizon = InvalidTransactionId;
+ else
+ vm_conflict_horizon = prstate.visibility_cutoff_xid;
+
+ /*
+ * During its second pass over the heap, VACUUM calls
+ * heap_page_would_be_all_visible() to determine whether a page is
+ * all-visible and all-frozen. The logic here is similar. After completing
+ * pruning and freezing, use an assertion to verify that our results
+ * remain consistent with heap_page_would_be_all_visible().
+ */
+#ifdef USE_ASSERT_CHECKING
+ if (prstate.all_visible)
+ {
+ TransactionId debug_cutoff;
+ bool debug_all_frozen;
+
+ Assert(presult->lpdead_items == 0);
+
+ Assert(heap_page_is_all_visible(params->relation, buffer,
+ prstate.cutoffs->OldestXmin,
+ &debug_all_frozen,
+ &debug_cutoff, off_loc));
+
+ Assert(prstate.all_frozen == debug_all_frozen);
+
+ Assert(!TransactionIdIsValid(debug_cutoff) ||
+ debug_cutoff == vm_conflict_horizon);
+ }
+#endif
+
/* Now update the visibility map and PD_ALL_VISIBLE hint */
Assert(!prstate.all_visible || (prstate.lpdead_items == 0));
@@ -1229,22 +1255,21 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
* make everything safe for REDO was logged when the page's tuples
* were frozen.
*/
- Assert(!prstate.all_frozen ||
- !TransactionIdIsValid(presult->vm_conflict_horizon));
+ Assert(!prstate.all_frozen || !TransactionIdIsValid(vm_conflict_horizon));
visibilitymap_set(params->relation, blockno, buffer,
InvalidXLogRecPtr,
- vmbuffer, presult->vm_conflict_horizon,
+ vmbuffer, vm_conflict_horizon,
new_vmbits);
if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
{
presult->new_all_visible_pages = 1;
- if (presult->all_frozen)
+ if (prstate.all_frozen)
presult->new_all_visible_frozen_pages = 1;
}
else if ((old_vmbits & VISIBILITYMAP_ALL_FROZEN) == 0 &&
- presult->all_frozen)
+ prstate.all_frozen)
presult->new_all_frozen_pages = 1;
}
}
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 323b8e3dde3..90e94b2ac3f 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -456,13 +456,6 @@ static void dead_items_add(LVRelState *vacrel, BlockNumber blkno, OffsetNumber *
static void dead_items_reset(LVRelState *vacrel);
static void dead_items_cleanup(LVRelState *vacrel);
-#ifdef USE_ASSERT_CHECKING
-static bool heap_page_is_all_visible(Relation rel, Buffer buf,
- TransactionId OldestXmin,
- bool *all_frozen,
- TransactionId *visibility_cutoff_xid,
- OffsetNumber *logging_offnum);
-#endif
static bool heap_page_would_be_all_visible(Relation rel, Buffer buf,
TransactionId OldestXmin,
OffsetNumber *deadoffsets,
@@ -2032,32 +2025,6 @@ lazy_scan_prune(LVRelState *vacrel,
vacrel->new_frozen_tuple_pages++;
}
- /*
- * VACUUM will call heap_page_is_all_visible() during the second pass over
- * the heap to determine all_visible and all_frozen for the page -- this
- * is a specialized version of the logic from this function. Now that
- * we've finished pruning and freezing, make sure that we're in total
- * agreement with heap_page_is_all_visible() using an assertion.
- */
-#ifdef USE_ASSERT_CHECKING
- if (presult.all_visible)
- {
- TransactionId debug_cutoff;
- bool debug_all_frozen;
-
- Assert(presult.lpdead_items == 0);
-
- Assert(heap_page_is_all_visible(vacrel->rel, buf,
- vacrel->cutoffs.OldestXmin, &debug_all_frozen,
- &debug_cutoff, &vacrel->offnum));
-
- Assert(presult.all_frozen == debug_all_frozen);
-
- Assert(!TransactionIdIsValid(debug_cutoff) ||
- debug_cutoff == presult.vm_conflict_horizon);
- }
-#endif
-
/*
* Now save details of the LP_DEAD items from the page in vacrel
*/
@@ -3511,7 +3478,7 @@ dead_items_cleanup(LVRelState *vacrel)
* that expect no LP_DEAD on the page. Currently assert-only, but there is no
* reason not to use it outside of asserts.
*/
-static bool
+bool
heap_page_is_all_visible(Relation rel, Buffer buf,
TransactionId OldestXmin,
bool *all_frozen,
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index df9fa17a6f9..a0f7974942e 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -257,8 +257,7 @@ typedef struct PruneFreezeParams
* HEAP_PAGE_PRUNE_MARK_UNUSED_NOW indicates that dead items can be set
* LP_UNUSED during pruning.
*
- * HEAP_PAGE_PRUNE_FREEZE indicates that we will also freeze tuples, and
- * will return 'all_visible', 'all_frozen' flags to the caller.
+ * HEAP_PAGE_PRUNE_FREEZE indicates that we will also freeze tuples.
*
* HEAP_PAGE_PRUNE_UPDATE_VM indicates that we will set the page's status
* in the VM.
@@ -294,21 +293,6 @@ typedef struct PruneFreezeResult
int live_tuples;
int recently_dead_tuples;
- /*
- * all_visible and all_frozen indicate if the all-visible and all-frozen
- * bits in the visibility map can be set for this page, after pruning.
- *
- * vm_conflict_horizon is the newest xmin of live tuples on the page. The
- * caller can use it as the conflict horizon when setting the VM bits. It
- * is only valid if we froze some tuples (nfrozen > 0), and all_frozen is
- * true.
- *
- * These are only set if the HEAP_PAGE_PRUNE_FREEZE option is set.
- */
- bool all_visible;
- bool all_frozen;
- TransactionId vm_conflict_horizon;
-
/*
* Whether or not the page was newly set all-visible and all-frozen during
* phase I of vacuuming.
@@ -453,7 +437,13 @@ extern void log_heap_prune_and_freeze(Relation relation, Buffer buffer,
/* in heap/vacuumlazy.c */
extern void heap_vacuum_rel(Relation rel,
const VacuumParams params, BufferAccessStrategy bstrategy);
-
+#ifdef USE_ASSERT_CHECKING
+extern bool heap_page_is_all_visible(Relation rel, Buffer buf,
+ TransactionId OldestXmin,
+ bool *all_frozen,
+ TransactionId *visibility_cutoff_xid,
+ OffsetNumber *logging_offnum);
+#endif
/* in heap/heapam_visibility.c */
extern bool HeapTupleSatisfiesVisibility(HeapTuple htup, Snapshot snapshot,
Buffer buffer);
--
2.43.0
[text/x-patch] v34-0004-Eliminate-XLOG_HEAP2_VISIBLE-from-vacuum-phase-I.patch (14.1K, 5-v34-0004-Eliminate-XLOG_HEAP2_VISIBLE-from-vacuum-phase-I.patch)
download | inline diff:
From a94267babeedec6705fd7f3b43242c6ba0e458c0 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Tue, 2 Dec 2025 16:16:22 -0500
Subject: [PATCH v34 04/14] Eliminate XLOG_HEAP2_VISIBLE from vacuum phase I
prune/freeze
Vacuum no longer emits a separate WAL record for each page set
all-visible or all-frozen during phase I. Instead, visibility map
updates are now included in the XLOG_HEAP2_PRUNE_VACUUM_SCAN record that
is already emitted for pruning and freezing.
Previously, heap_page_prune_and_freeze() determined whether a page was
all-visible, but the corresponding VM bits were only set later in
lazy_scan_prune(). Now the VM is updated immediately in
heap_page_prune_and_freeze(), at the same time as the heap
modifications.
This change applies only to vacuum phase I, not to pruning performed
during normal page access.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Robert Haas <[email protected]>
Reviewed-by: Kirill Reshke <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/flat/CAAKRu_ZMw6Npd_qm2KM%2BFwQ3cMOMx1Dh3VMhp8-V7SOLxdK9-g%40mail.gmail.com
---
src/backend/access/heap/pruneheap.c | 266 ++++++++++++++++------------
1 file changed, 152 insertions(+), 114 deletions(-)
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index 192df9a2218..b8ba5b7a681 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -205,6 +205,11 @@ static bool heap_page_will_set_vm(PruneState *prstate,
int nlpdead_items,
uint8 *old_vmbits,
uint8 *new_vmbits);
+static TransactionId get_conflict_xid(bool do_prune, bool do_freeze, bool do_set_vm,
+ uint8 old_vmbits, uint8 new_vmbits,
+ TransactionId latest_xid_removed,
+ TransactionId frz_conflict_horizon,
+ TransactionId visibility_cutoff_xid);
/*
@@ -804,6 +809,62 @@ heap_page_will_freeze(Relation relation, Buffer buffer,
return do_freeze;
}
+/*
+ * Calculate the conflict horizon for the whole XLOG_HEAP2_PRUNE_VACUUM_SCAN
+ * or XLOG_HEAP2_PRUNE_ON_ACCESS record.
+ */
+static TransactionId
+get_conflict_xid(bool do_prune, bool do_freeze, bool do_set_vm,
+ uint8 old_vmbits, uint8 new_vmbits,
+ TransactionId latest_xid_removed, TransactionId frz_conflict_horizon,
+ TransactionId visibility_cutoff_xid)
+{
+ TransactionId conflict_xid;
+
+ /*
+ * We can omit the snapshot conflict horizon if we are not pruning or
+ * freezing any tuples and are setting an already all-visible page
+ * all-frozen in the VM. In this case, all of the tuples on the page must
+ * already be seen as frozen by all MVCC snapshots on the standby.
+ */
+ if (!do_prune &&
+ !do_freeze &&
+ (old_vmbits & VISIBILITYMAP_ALL_VISIBLE) != 0 &&
+ (new_vmbits & VISIBILITYMAP_ALL_FROZEN) != 0)
+ return InvalidTransactionId;
+
+ /*
+ * The snapshot conflict horizon for the whole record should be the most
+ * conservative of all the horizons calculated for any of the possible
+ * modifications. If this record will prune tuples, any transactions on
+ * the standby older than the youngest xmax of the most recently removed
+ * tuple this record will prune will conflict. If this record will freeze
+ * tuples, any transactions on the standby with xids older than the
+ * youngest tuple this record will freeze will conflict.
+ *
+ * If we are setting the VM, the conflict horizon is almost always the
+ * visibility cutoff XID, except in the situation described above.
+ *
+ * By picking the newest of all of those, we can ensure that all changes
+ * in the record have been taken into account.
+ */
+ if (do_set_vm)
+ conflict_xid = visibility_cutoff_xid;
+ else if (do_freeze)
+ conflict_xid = frz_conflict_horizon;
+ else
+ conflict_xid = InvalidTransactionId;
+
+ /*
+ * If we are removing tuples with a younger xmax than our so far
+ * calculated conflict_xid, we must use this as our horizon.
+ */
+ if (TransactionIdFollows(latest_xid_removed, conflict_xid))
+ conflict_xid = latest_xid_removed;
+
+ return conflict_xid;
+}
+
/*
* Helper to correct any corruption detected on a heap page and its
* corresponding visibility map page after pruning but before setting the
@@ -991,7 +1052,6 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
Buffer vmbuffer = params->vmbuffer;
Page page = BufferGetPage(buffer);
BlockNumber blockno = BufferGetBlockNumber(buffer);
- TransactionId vm_conflict_horizon = InvalidTransactionId;
PruneState prstate;
bool do_freeze;
bool do_prune;
@@ -999,6 +1059,7 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
bool do_set_vm;
bool did_tuple_hint_fpi;
int64 fpi_before = pgWalUsage.wal_fpi;
+ TransactionId conflict_xid;
uint8 new_vmbits;
uint8 old_vmbits;
@@ -1063,6 +1124,37 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
prstate.all_visible = prstate.all_frozen = false;
Assert(!prstate.all_frozen || prstate.all_visible);
+ Assert(!prstate.all_visible || (prstate.lpdead_items == 0));
+
+ /*
+ * Decide whether to set the VM bits based on information from the VM and
+ * the all_visible/all_frozen flags.
+ */
+ do_set_vm = heap_page_will_set_vm(&prstate,
+ params->relation,
+ blockno,
+ buffer,
+ page,
+ vmbuffer,
+ prstate.lpdead_items,
+ &old_vmbits,
+ &new_vmbits);
+
+ /*
+ * new_vmbits should be 0 regardless of whether or not the page is
+ * all-visible if we do not intend to set the VM.
+ */
+ Assert(do_set_vm || new_vmbits == 0);
+
+ conflict_xid = get_conflict_xid(do_prune, do_freeze, do_set_vm,
+ old_vmbits, new_vmbits,
+ prstate.latest_xid_removed,
+ prstate.frz_conflict_horizon,
+ prstate.visibility_cutoff_xid);
+
+ /* Lock vmbuffer before entering a critical section */
+ if (do_set_vm)
+ LockBuffer(vmbuffer, BUFFER_LOCK_EXCLUSIVE);
/* Any error while applying the changes is critical */
START_CRIT_SECTION();
@@ -1084,14 +1176,17 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
/*
* If that's all we had to do to the page, this is a non-WAL-logged
- * hint. If we are going to freeze or prune the page, we will mark
- * the buffer dirty below.
+ * hint. If we are going to freeze or prune the page or set
+ * PD_ALL_VISIBLE, we will mark the buffer dirty below.
+ *
+ * Setting PD_ALL_VISIBLE is fully WAL-logged because it is forbidden
+ * for the VM to be set and PD_ALL_VISIBLE to be clear.
*/
- if (!do_freeze && !do_prune)
+ if (!do_freeze && !do_prune && !do_set_vm)
MarkBufferDirtyHint(buffer, true);
}
- if (do_prune || do_freeze)
+ if (do_prune || do_freeze || do_set_vm)
{
/* Apply the planned item changes and repair page fragmentation. */
if (do_prune)
@@ -1105,6 +1200,26 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
if (do_freeze)
heap_freeze_prepared_tuples(buffer, prstate.frozen, prstate.nfrozen);
+ /* Set the visibility map and page visibility hint */
+ if (do_set_vm)
+ {
+ /*
+ * While it is valid for PD_ALL_VISIBLE to be set when the
+ * corresponding VM bit is clear, we strongly prefer to keep them
+ * in sync.
+ *
+ * The heap buffer must be marked dirty before adding it to the
+ * WAL chain when setting the VM. We don't worry about
+ * unnecessarily dirtying the heap buffer if PD_ALL_VISIBLE is
+ * already set, though. It is extremely rare to have a clean heap
+ * buffer with PD_ALL_VISIBLE already set and the VM bits clear,
+ * so there is no point in optimizing it.
+ */
+ PageSetAllVisible(page);
+ visibilitymap_set_vmbits(blockno, vmbuffer, new_vmbits,
+ params->relation->rd_locator);
+ }
+
MarkBufferDirty(buffer);
/*
@@ -1112,29 +1227,12 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
*/
if (RelationNeedsWAL(params->relation))
{
- /*
- * The snapshotConflictHorizon for the whole record should be the
- * most conservative of all the horizons calculated for any of the
- * possible modifications. If this record will prune tuples, any
- * transactions on the standby older than the youngest xmax of the
- * most recently removed tuple this record will prune will
- * conflict. If this record will freeze tuples, any transactions
- * on the standby with xids older than the youngest tuple this
- * record will freeze will conflict.
- */
- TransactionId conflict_xid;
-
- if (TransactionIdFollows(prstate.frz_conflict_horizon,
- prstate.latest_xid_removed))
- conflict_xid = prstate.frz_conflict_horizon;
- else
- conflict_xid = prstate.latest_xid_removed;
-
log_heap_prune_and_freeze(params->relation, buffer,
- InvalidBuffer, /* vmbuffer */
- 0, /* vmflags */
+ do_set_vm ? vmbuffer : InvalidBuffer,
+ do_set_vm ? new_vmbits : 0,
conflict_xid,
- true, params->reason,
+ true, /* cleanup lock */
+ params->reason,
prstate.frozen, prstate.nfrozen,
prstate.redirected, prstate.nredirected,
prstate.nowdead, prstate.ndead,
@@ -1144,43 +1242,8 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
END_CRIT_SECTION();
- /* Copy information back for caller */
- presult->ndeleted = prstate.ndeleted;
- presult->nnewlpdead = prstate.ndead;
- presult->nfrozen = prstate.nfrozen;
- presult->live_tuples = prstate.live_tuples;
- presult->recently_dead_tuples = prstate.recently_dead_tuples;
- presult->hastup = prstate.hastup;
-
- presult->lpdead_items = prstate.lpdead_items;
- /* the presult->deadoffsets array was already filled in */
-
- if (prstate.attempt_freeze)
- {
- if (presult->nfrozen > 0)
- {
- *new_relfrozen_xid = prstate.pagefrz.FreezePageRelfrozenXid;
- *new_relmin_mxid = prstate.pagefrz.FreezePageRelminMxid;
- }
- else
- {
- *new_relfrozen_xid = prstate.pagefrz.NoFreezePageRelfrozenXid;
- *new_relmin_mxid = prstate.pagefrz.NoFreezePageRelminMxid;
- }
- }
-
- /*
- * If updating the visibility map, the conflict horizon for that record
- * must be the newest xmin on the page. However, if the page is
- * completely frozen, there can be no conflict and the vm_conflict_horizon
- * should remain InvalidTransactionId. This includes the case that we
- * just froze all the tuples; the prune-freeze record included the
- * conflict XID already so we don't need to again.
- */
- if (prstate.all_frozen)
- vm_conflict_horizon = InvalidTransactionId;
- else
- vm_conflict_horizon = prstate.visibility_cutoff_xid;
+ if (do_set_vm)
+ LockBuffer(vmbuffer, BUFFER_LOCK_UNLOCK);
/*
* During its second pass over the heap, VACUUM calls
@@ -1195,7 +1258,8 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
TransactionId debug_cutoff;
bool debug_all_frozen;
- Assert(presult->lpdead_items == 0);
+ Assert(prstate.lpdead_items == 0);
+ Assert(prstate.cutoffs);
Assert(heap_page_is_all_visible(params->relation, buffer,
prstate.cutoffs->OldestXmin,
@@ -1205,63 +1269,23 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
Assert(prstate.all_frozen == debug_all_frozen);
Assert(!TransactionIdIsValid(debug_cutoff) ||
- debug_cutoff == vm_conflict_horizon);
+ debug_cutoff == prstate.visibility_cutoff_xid);
}
#endif
- /* Now update the visibility map and PD_ALL_VISIBLE hint */
- Assert(!prstate.all_visible || (prstate.lpdead_items == 0));
-
- do_set_vm = heap_page_will_set_vm(&prstate,
- params->relation,
- blockno,
- buffer,
- page,
- vmbuffer,
- prstate.lpdead_items,
- &old_vmbits,
- &new_vmbits);
+ /* Copy information back for caller */
+ presult->ndeleted = prstate.ndeleted;
+ presult->nnewlpdead = prstate.ndead;
+ presult->nfrozen = prstate.nfrozen;
+ presult->live_tuples = prstate.live_tuples;
+ presult->recently_dead_tuples = prstate.recently_dead_tuples;
+ presult->hastup = prstate.hastup;
- /*
- * new_vmbits should be 0 regardless of whether or not the page is
- * all-visible if we do not intend to set the VM.
- */
- Assert(do_set_vm || new_vmbits == 0);
+ presult->lpdead_items = prstate.lpdead_items;
+ /* the presult->deadoffsets array was already filled in */
- /* Set the visibility map and page visibility hint, if relevant */
if (do_set_vm)
{
- Assert(prstate.all_visible);
-
- /*
- * It should never be the case that the visibility map page is set
- * while the page-level bit is clear (and if so, we cleared it above),
- * but the reverse is allowed (if checksums are not enabled).
- * Regardless, set both bits so that we get back in sync.
- *
- * The heap buffer must be marked dirty before adding it to the WAL
- * chain when setting the VM. We don't worry about unnecessarily
- * dirtying the heap buffer if PD_ALL_VISIBLE is already set, though.
- * It is extremely rare to have a clean heap buffer with
- * PD_ALL_VISIBLE already set and the VM bits clear, so there is no
- * point in optimizing it.
- */
- PageSetAllVisible(page);
- MarkBufferDirty(buffer);
-
- /*
- * If the page is being set all-frozen, we pass InvalidTransactionId
- * as the cutoff_xid, since a snapshot conflict horizon sufficient to
- * make everything safe for REDO was logged when the page's tuples
- * were frozen.
- */
- Assert(!prstate.all_frozen || !TransactionIdIsValid(vm_conflict_horizon));
-
- visibilitymap_set(params->relation, blockno, buffer,
- InvalidXLogRecPtr,
- vmbuffer, vm_conflict_horizon,
- new_vmbits);
-
if ((old_vmbits & VISIBILITYMAP_ALL_VISIBLE) == 0)
{
presult->new_all_visible_pages = 1;
@@ -1272,6 +1296,20 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
prstate.all_frozen)
presult->new_all_frozen_pages = 1;
}
+
+ if (prstate.attempt_freeze)
+ {
+ if (presult->nfrozen > 0)
+ {
+ *new_relfrozen_xid = prstate.pagefrz.FreezePageRelfrozenXid;
+ *new_relmin_mxid = prstate.pagefrz.FreezePageRelminMxid;
+ }
+ else
+ {
+ *new_relfrozen_xid = prstate.pagefrz.NoFreezePageRelfrozenXid;
+ *new_relmin_mxid = prstate.pagefrz.NoFreezePageRelminMxid;
+ }
+ }
}
--
2.43.0
[text/x-patch] v34-0005-Eliminate-XLOG_HEAP2_VISIBLE-from-empty-page-vac.patch (2.6K, 6-v34-0005-Eliminate-XLOG_HEAP2_VISIBLE-from-empty-page-vac.patch)
download | inline diff:
From dda6ceffc924333fd7587c92c336a01932967532 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Sat, 27 Sep 2025 11:55:21 -0400
Subject: [PATCH v34 05/14] Eliminate XLOG_HEAP2_VISIBLE from empty-page vacuum
As part of removing XLOG_HEAP2_VISIBLE records, phase I of VACUUM now
marks empty pages all-visible in a XLOG_HEAP2_PRUNE_VACUUM_SCAN record.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Robert Haas <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Chao Li <[email protected]>
---
src/backend/access/heap/vacuumlazy.c | 35 +++++++++++++++++++++++-----
1 file changed, 29 insertions(+), 6 deletions(-)
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 90e94b2ac3f..86b3155717e 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -1900,9 +1900,12 @@ lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf, BlockNumber blkno,
*/
if (!PageIsAllVisible(page))
{
+ /* Lock vmbuffer before entering critical section */
+ LockBuffer(vmbuffer, BUFFER_LOCK_EXCLUSIVE);
+
START_CRIT_SECTION();
- /* mark buffer dirty before writing a WAL record */
+ /* Mark buffer dirty before writing any WAL records */
MarkBufferDirty(buf);
/*
@@ -1919,13 +1922,33 @@ lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf, BlockNumber blkno,
log_newpage_buffer(buf, true);
PageSetAllVisible(page);
- visibilitymap_set(vacrel->rel, blkno, buf,
- InvalidXLogRecPtr,
- vmbuffer, InvalidTransactionId,
- VISIBILITYMAP_ALL_VISIBLE |
- VISIBILITYMAP_ALL_FROZEN);
+ visibilitymap_set_vmbits(blkno,
+ vmbuffer,
+ VISIBILITYMAP_ALL_VISIBLE |
+ VISIBILITYMAP_ALL_FROZEN,
+ vacrel->rel->rd_locator);
+
+ /*
+ * Emit WAL for setting PD_ALL_VISIBLE on the heap page and
+ * setting the VM.
+ */
+ if (RelationNeedsWAL(vacrel->rel))
+ log_heap_prune_and_freeze(vacrel->rel, buf,
+ vmbuffer,
+ VISIBILITYMAP_ALL_VISIBLE |
+ VISIBILITYMAP_ALL_FROZEN,
+ InvalidTransactionId, /* conflict xid */
+ false, /* cleanup lock */
+ PRUNE_VACUUM_SCAN, /* reason */
+ NULL, 0,
+ NULL, 0,
+ NULL, 0,
+ NULL, 0);
+
END_CRIT_SECTION();
+ LockBuffer(vmbuffer, BUFFER_LOCK_UNLOCK);
+
/* Count the newly all-frozen pages for logging */
vacrel->vm_new_visible_pages++;
vacrel->vm_new_visible_frozen_pages++;
--
2.43.0
[text/x-patch] v34-0006-Remove-XLOG_HEAP2_VISIBLE-entirely.patch (24.9K, 7-v34-0006-Remove-XLOG_HEAP2_VISIBLE-entirely.patch)
download | inline diff:
From cfe59761babde9cf9aeab5d5ca0b50220a41b76d Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Sat, 27 Sep 2025 11:55:36 -0400
Subject: [PATCH v34 06/14] Remove XLOG_HEAP2_VISIBLE entirely
As no remaining users emit XLOG_HEAP2_VISIBLE records.
This includes deleting the xl_heap_visible struct and all functions
responsible for emitting or replaying XLOG_HEAP2_VISIBLE records.
This changes the visibility map API, so any external users/consumers of
the VM-only WAL record will need to change.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andrey Borodin <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Chao Li <[email protected]>
---
src/backend/access/common/bufmask.c | 4 +-
src/backend/access/heap/heapam.c | 54 +-------
src/backend/access/heap/heapam_xlog.c | 155 ++---------------------
src/backend/access/heap/pruneheap.c | 4 +-
src/backend/access/heap/vacuumlazy.c | 16 +--
src/backend/access/heap/visibilitymap.c | 110 +---------------
src/backend/access/rmgrdesc/heapdesc.c | 10 --
src/backend/replication/logical/decode.c | 1 -
src/backend/storage/ipc/standby.c | 12 +-
src/include/access/heapam_xlog.h | 20 ---
src/include/access/visibilitymap.h | 13 +-
src/include/access/visibilitymapdefs.h | 9 --
src/tools/pgindent/typedefs.list | 1 -
13 files changed, 38 insertions(+), 371 deletions(-)
diff --git a/src/backend/access/common/bufmask.c b/src/backend/access/common/bufmask.c
index 1a9e7bea5d2..bce767d7b71 100644
--- a/src/backend/access/common/bufmask.c
+++ b/src/backend/access/common/bufmask.c
@@ -56,8 +56,8 @@ mask_page_hint_bits(Page page)
/*
* During replay, if the page LSN has advanced past our XLOG record's LSN,
- * we don't mark the page all-visible. See heap_xlog_visible() for
- * details.
+ * we don't mark the page all-visible. See heap_xlog_prune_and_freeze()
+ * for more details.
*/
PageClearAllVisible(page);
}
diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index f30a56ecf55..1a3fa8a76aa 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -2586,11 +2586,11 @@ heap_multi_insert(Relation relation, TupleTableSlot **slots, int ntuples,
else if (all_frozen_set)
{
PageSetAllVisible(page);
- visibilitymap_set_vmbits(BufferGetBlockNumber(buffer),
- vmbuffer,
- VISIBILITYMAP_ALL_VISIBLE |
- VISIBILITYMAP_ALL_FROZEN,
- relation->rd_locator);
+ visibilitymap_set(BufferGetBlockNumber(buffer),
+ vmbuffer,
+ VISIBILITYMAP_ALL_VISIBLE |
+ VISIBILITYMAP_ALL_FROZEN,
+ relation->rd_locator);
}
/*
@@ -8875,50 +8875,6 @@ bottomup_sort_and_shrink(TM_IndexDeleteOp *delstate)
return nblocksfavorable;
}
-/*
- * Perform XLogInsert for a heap-visible operation. 'block' is the block
- * being marked all-visible, and vm_buffer is the buffer containing the
- * corresponding visibility map block. Both should have already been modified
- * and dirtied.
- *
- * snapshotConflictHorizon comes from the largest xmin on the page being
- * marked all-visible. REDO routine uses it to generate recovery conflicts.
- *
- * If checksums or wal_log_hints are enabled, we may also generate a full-page
- * image of heap_buffer. Otherwise, we optimize away the FPI (by specifying
- * REGBUF_NO_IMAGE for the heap buffer), in which case the caller should *not*
- * update the heap page's LSN.
- */
-XLogRecPtr
-log_heap_visible(Relation rel, Buffer heap_buffer, Buffer vm_buffer,
- TransactionId snapshotConflictHorizon, uint8 vmflags)
-{
- xl_heap_visible xlrec;
- XLogRecPtr recptr;
- uint8 flags;
-
- Assert(BufferIsValid(heap_buffer));
- Assert(BufferIsValid(vm_buffer));
-
- xlrec.snapshotConflictHorizon = snapshotConflictHorizon;
- xlrec.flags = vmflags;
- if (RelationIsAccessibleInLogicalDecoding(rel))
- xlrec.flags |= VISIBILITYMAP_XLOG_CATALOG_REL;
- XLogBeginInsert();
- XLogRegisterData(&xlrec, SizeOfHeapVisible);
-
- XLogRegisterBuffer(0, vm_buffer, 0);
-
- flags = REGBUF_STANDARD;
- if (!XLogHintBitIsNeeded())
- flags |= REGBUF_NO_IMAGE;
- XLogRegisterBuffer(1, heap_buffer, flags);
-
- recptr = XLogInsert(RM_HEAP2_ID, XLOG_HEAP2_VISIBLE);
-
- return recptr;
-}
-
/*
* Perform XLogInsert for a heap-update operation. Caller must already
* have modified the buffer(s) and marked them dirty.
diff --git a/src/backend/access/heap/heapam_xlog.c b/src/backend/access/heap/heapam_xlog.c
index f765345e9e4..9a29fda3601 100644
--- a/src/backend/access/heap/heapam_xlog.c
+++ b/src/backend/access/heap/heapam_xlog.c
@@ -236,7 +236,7 @@ heap_xlog_prune_freeze(XLogReaderState *record)
if (PageIsNew(vmpage))
PageInit(vmpage, BLCKSZ, 0);
- visibilitymap_set_vmbits(blkno, vmbuffer, vmflags, rlocator);
+ visibilitymap_set(blkno, vmbuffer, vmflags, rlocator);
Assert(BufferIsDirty(vmbuffer));
PageSetLSN(vmpage, lsn);
@@ -249,142 +249,6 @@ heap_xlog_prune_freeze(XLogReaderState *record)
XLogRecordPageWithFreeSpace(rlocator, blkno, freespace);
}
-/*
- * Replay XLOG_HEAP2_VISIBLE records.
- *
- * The critical integrity requirement here is that we must never end up with
- * a situation where the visibility map bit is set, and the page-level
- * PD_ALL_VISIBLE bit is clear. If that were to occur, then a subsequent
- * page modification would fail to clear the visibility map bit.
- */
-static void
-heap_xlog_visible(XLogReaderState *record)
-{
- XLogRecPtr lsn = record->EndRecPtr;
- xl_heap_visible *xlrec = (xl_heap_visible *) XLogRecGetData(record);
- Buffer vmbuffer = InvalidBuffer;
- Buffer buffer;
- Page page;
- RelFileLocator rlocator;
- BlockNumber blkno;
- XLogRedoAction action;
-
- Assert((xlrec->flags & VISIBILITYMAP_XLOG_VALID_BITS) == xlrec->flags);
-
- XLogRecGetBlockTag(record, 1, &rlocator, NULL, &blkno);
-
- /*
- * If there are any Hot Standby transactions running that have an xmin
- * horizon old enough that this page isn't all-visible for them, they
- * might incorrectly decide that an index-only scan can skip a heap fetch.
- *
- * NB: It might be better to throw some kind of "soft" conflict here that
- * forces any index-only scan that is in flight to perform heap fetches,
- * rather than killing the transaction outright.
- */
- if (InHotStandby)
- ResolveRecoveryConflictWithSnapshot(xlrec->snapshotConflictHorizon,
- xlrec->flags & VISIBILITYMAP_XLOG_CATALOG_REL,
- rlocator);
-
- /*
- * Read the heap page, if it still exists. If the heap file has dropped or
- * truncated later in recovery, we don't need to update the page, but we'd
- * better still update the visibility map.
- */
- action = XLogReadBufferForRedo(record, 1, &buffer);
- if (action == BLK_NEEDS_REDO)
- {
- /*
- * We don't bump the LSN of the heap page when setting the visibility
- * map bit (unless checksums or wal_hint_bits is enabled, in which
- * case we must). This exposes us to torn page hazards, but since
- * we're not inspecting the existing page contents in any way, we
- * don't care.
- */
- page = BufferGetPage(buffer);
-
- PageSetAllVisible(page);
-
- if (XLogHintBitIsNeeded())
- PageSetLSN(page, lsn);
-
- MarkBufferDirty(buffer);
- }
- else if (action == BLK_RESTORED)
- {
- /*
- * If heap block was backed up, we already restored it and there's
- * nothing more to do. (This can only happen with checksums or
- * wal_log_hints enabled.)
- */
- }
-
- if (BufferIsValid(buffer))
- {
- Size space = PageGetFreeSpace(BufferGetPage(buffer));
-
- UnlockReleaseBuffer(buffer);
-
- /*
- * Since FSM is not WAL-logged and only updated heuristically, it
- * easily becomes stale in standbys. If the standby is later promoted
- * and runs VACUUM, it will skip updating individual free space
- * figures for pages that became all-visible (or all-frozen, depending
- * on the vacuum mode,) which is troublesome when FreeSpaceMapVacuum
- * propagates too optimistic free space values to upper FSM layers;
- * later inserters try to use such pages only to find out that they
- * are unusable. This can cause long stalls when there are many such
- * pages.
- *
- * Forestall those problems by updating FSM's idea about a page that
- * is becoming all-visible or all-frozen.
- *
- * Do this regardless of a full-page image being applied, since the
- * FSM data is not in the page anyway.
- */
- if (xlrec->flags & VISIBILITYMAP_VALID_BITS)
- XLogRecordPageWithFreeSpace(rlocator, blkno, space);
- }
-
- /*
- * Even if we skipped the heap page update due to the LSN interlock, it's
- * still safe to update the visibility map. Any WAL record that clears
- * the visibility map bit does so before checking the page LSN, so any
- * bits that need to be cleared will still be cleared.
- */
- if (XLogReadBufferForRedoExtended(record, 0, RBM_ZERO_ON_ERROR, false,
- &vmbuffer) == BLK_NEEDS_REDO)
- {
- Page vmpage = BufferGetPage(vmbuffer);
- Relation reln;
- uint8 vmbits;
-
- /* initialize the page if it was read as zeros */
- if (PageIsNew(vmpage))
- PageInit(vmpage, BLCKSZ, 0);
-
- /* remove VISIBILITYMAP_XLOG_* */
- vmbits = xlrec->flags & VISIBILITYMAP_VALID_BITS;
-
- /*
- * XLogReadBufferForRedoExtended locked the buffer. But
- * visibilitymap_set will handle locking itself.
- */
- LockBuffer(vmbuffer, BUFFER_LOCK_UNLOCK);
-
- reln = CreateFakeRelcacheEntry(rlocator);
-
- visibilitymap_set(reln, blkno, InvalidBuffer, lsn, vmbuffer,
- xlrec->snapshotConflictHorizon, vmbits);
-
- ReleaseBuffer(vmbuffer);
- FreeFakeRelcacheEntry(reln);
- }
- else if (BufferIsValid(vmbuffer))
- UnlockReleaseBuffer(vmbuffer);
-}
-
/*
* Given an "infobits" field from an XLog record, set the correct bits in the
* given infomask and infomask2 for the tuple touched by the record.
@@ -762,8 +626,8 @@ heap_xlog_multi_insert(XLogReaderState *record)
*
* During recovery, however, no concurrent writers exist. Therefore,
* updating the VM without holding the heap page lock is safe enough. This
- * same approach is taken when replaying xl_heap_visible records (see
- * heap_xlog_visible()).
+ * same approach is taken when replaying XLOG_HEAP2_PRUNE* records (see
+ * heap_xlog_prune_and_freeze()).
*/
if ((xlrec->flags & XLH_INSERT_ALL_FROZEN_SET) &&
XLogReadBufferForRedoExtended(record, 1, RBM_ZERO_ON_ERROR, false,
@@ -775,11 +639,11 @@ heap_xlog_multi_insert(XLogReaderState *record)
if (PageIsNew(vmpage))
PageInit(vmpage, BLCKSZ, 0);
- visibilitymap_set_vmbits(blkno,
- vmbuffer,
- VISIBILITYMAP_ALL_VISIBLE |
- VISIBILITYMAP_ALL_FROZEN,
- rlocator);
+ visibilitymap_set(blkno,
+ vmbuffer,
+ VISIBILITYMAP_ALL_VISIBLE |
+ VISIBILITYMAP_ALL_FROZEN,
+ rlocator);
Assert(BufferIsDirty(vmbuffer));
PageSetLSN(vmpage, lsn);
@@ -1360,9 +1224,6 @@ heap2_redo(XLogReaderState *record)
case XLOG_HEAP2_PRUNE_VACUUM_CLEANUP:
heap_xlog_prune_freeze(record);
break;
- case XLOG_HEAP2_VISIBLE:
- heap_xlog_visible(record);
- break;
case XLOG_HEAP2_MULTI_INSERT:
heap_xlog_multi_insert(record);
break;
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index b8ba5b7a681..2ba863be07c 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -1216,8 +1216,8 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
* so there is no point in optimizing it.
*/
PageSetAllVisible(page);
- visibilitymap_set_vmbits(blockno, vmbuffer, new_vmbits,
- params->relation->rd_locator);
+ visibilitymap_set(blockno, vmbuffer, new_vmbits,
+ params->relation->rd_locator);
}
MarkBufferDirty(buffer);
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 86b3155717e..d4624010123 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -1922,11 +1922,11 @@ lazy_scan_new_or_empty(LVRelState *vacrel, Buffer buf, BlockNumber blkno,
log_newpage_buffer(buf, true);
PageSetAllVisible(page);
- visibilitymap_set_vmbits(blkno,
- vmbuffer,
- VISIBILITYMAP_ALL_VISIBLE |
- VISIBILITYMAP_ALL_FROZEN,
- vacrel->rel->rd_locator);
+ visibilitymap_set(blkno,
+ vmbuffer,
+ VISIBILITYMAP_ALL_VISIBLE |
+ VISIBILITYMAP_ALL_FROZEN,
+ vacrel->rel->rd_locator);
/*
* Emit WAL for setting PD_ALL_VISIBLE on the heap page and
@@ -2789,9 +2789,9 @@ lazy_vacuum_heap_page(LVRelState *vacrel, BlockNumber blkno, Buffer buffer,
* set PD_ALL_VISIBLE.
*/
PageSetAllVisible(page);
- visibilitymap_set_vmbits(blkno,
- vmbuffer, vmflags,
- vacrel->rel->rd_locator);
+ visibilitymap_set(blkno,
+ vmbuffer, vmflags,
+ vacrel->rel->rd_locator);
conflict_xid = visibility_cutoff_xid;
}
diff --git a/src/backend/access/heap/visibilitymap.c b/src/backend/access/heap/visibilitymap.c
index 3047bd46def..fc74e39e069 100644
--- a/src/backend/access/heap/visibilitymap.c
+++ b/src/backend/access/heap/visibilitymap.c
@@ -14,8 +14,7 @@
* visibilitymap_clear - clear bits for one page in the visibility map
* visibilitymap_pin - pin a map page for setting a bit
* visibilitymap_pin_ok - check whether correct map page is already pinned
- * visibilitymap_set - set bit(s) in a previously pinned page and log
- * visibilitymap_set_vmbits - set bit(s) in a pinned page
+ * visibilitymap_set - set bit(s) in a previously pinned page
* visibilitymap_get_status - get status of bits
* visibilitymap_count - count number of bits set in visibility map
* visibilitymap_prepare_truncate -
@@ -220,112 +219,11 @@ visibilitymap_pin_ok(BlockNumber heapBlk, Buffer vmbuf)
return BufferIsValid(vmbuf) && BufferGetBlockNumber(vmbuf) == mapBlock;
}
-/*
- * visibilitymap_set - set bit(s) on a previously pinned page
- *
- * recptr is the LSN of the XLOG record we're replaying, if we're in recovery,
- * or InvalidXLogRecPtr in normal running. The VM page LSN is advanced to the
- * one provided; in normal running, we generate a new XLOG record and set the
- * page LSN to that value (though the heap page's LSN may *not* be updated;
- * see below). cutoff_xid is the largest xmin on the page being marked
- * all-visible; it is needed for Hot Standby, and can be InvalidTransactionId
- * if the page contains no tuples. It can also be set to InvalidTransactionId
- * when a page that is already all-visible is being marked all-frozen.
- *
- * Caller is expected to set the heap page's PD_ALL_VISIBLE bit before calling
- * this function. Except in recovery, caller should also pass the heap
- * buffer. When checksums are enabled and we're not in recovery, we must add
- * the heap buffer to the WAL chain to protect it from being torn.
- *
- * You must pass a buffer containing the correct map page to this function.
- * Call visibilitymap_pin first to pin the right one. This function doesn't do
- * any I/O.
- */
-void
-visibilitymap_set(Relation rel, BlockNumber heapBlk, Buffer heapBuf,
- XLogRecPtr recptr, Buffer vmBuf, TransactionId cutoff_xid,
- uint8 flags)
-{
- BlockNumber mapBlock = HEAPBLK_TO_MAPBLOCK(heapBlk);
- uint32 mapByte = HEAPBLK_TO_MAPBYTE(heapBlk);
- uint8 mapOffset = HEAPBLK_TO_OFFSET(heapBlk);
- Page page;
- uint8 *map;
- uint8 status;
-
-#ifdef TRACE_VISIBILITYMAP
- elog(DEBUG1, "vm_set flags 0x%02X for %s %d",
- flags, RelationGetRelationName(rel), heapBlk);
-#endif
-
- Assert(InRecovery || !XLogRecPtrIsValid(recptr));
- Assert(InRecovery || PageIsAllVisible(BufferGetPage(heapBuf)));
- Assert((flags & VISIBILITYMAP_VALID_BITS) == flags);
-
- /* Must never set all_frozen bit without also setting all_visible bit */
- Assert(flags != VISIBILITYMAP_ALL_FROZEN);
-
- /* Check that we have the right heap page pinned, if present */
- if (BufferIsValid(heapBuf) && BufferGetBlockNumber(heapBuf) != heapBlk)
- elog(ERROR, "wrong heap buffer passed to visibilitymap_set");
-
- Assert(!BufferIsValid(heapBuf) ||
- BufferIsLockedByMeInMode(heapBuf, BUFFER_LOCK_EXCLUSIVE));
-
- /* Check that we have the right VM page pinned */
- if (!BufferIsValid(vmBuf) || BufferGetBlockNumber(vmBuf) != mapBlock)
- elog(ERROR, "wrong VM buffer passed to visibilitymap_set");
-
- page = BufferGetPage(vmBuf);
- map = (uint8 *) PageGetContents(page);
- LockBuffer(vmBuf, BUFFER_LOCK_EXCLUSIVE);
-
- status = (map[mapByte] >> mapOffset) & VISIBILITYMAP_VALID_BITS;
- if (flags != status)
- {
- START_CRIT_SECTION();
-
- map[mapByte] |= (flags << mapOffset);
- MarkBufferDirty(vmBuf);
-
- if (RelationNeedsWAL(rel))
- {
- if (!XLogRecPtrIsValid(recptr))
- {
- Assert(!InRecovery);
- recptr = log_heap_visible(rel, heapBuf, vmBuf, cutoff_xid, flags);
-
- /*
- * If data checksums are enabled (or wal_log_hints=on), we
- * need to protect the heap page from being torn.
- *
- * If not, then we must *not* update the heap page's LSN. In
- * this case, the FPI for the heap page was omitted from the
- * WAL record inserted above, so it would be incorrect to
- * update the heap page's LSN.
- */
- if (XLogHintBitIsNeeded())
- {
- Page heapPage = BufferGetPage(heapBuf);
-
- PageSetLSN(heapPage, recptr);
- }
- }
- PageSetLSN(page, recptr);
- }
-
- END_CRIT_SECTION();
- }
-
- LockBuffer(vmBuf, BUFFER_LOCK_UNLOCK);
-}
-
/*
* Set VM (visibility map) flags in the VM block in vmBuf.
*
* This function is intended for callers that log VM changes together
* with the heap page modifications that rendered the page all-visible.
- * Callers that log VM changes separately should use visibilitymap_set().
*
* vmBuf must be pinned and exclusively locked, and it must cover the VM bits
* corresponding to heapBlk.
@@ -341,9 +239,9 @@ visibilitymap_set(Relation rel, BlockNumber heapBlk, Buffer heapBuf,
* rlocator is used only for debugging messages.
*/
void
-visibilitymap_set_vmbits(BlockNumber heapBlk,
- Buffer vmBuf, uint8 flags,
- const RelFileLocator rlocator)
+visibilitymap_set(BlockNumber heapBlk,
+ Buffer vmBuf, uint8 flags,
+ const RelFileLocator rlocator)
{
BlockNumber mapBlock = HEAPBLK_TO_MAPBLOCK(heapBlk);
uint32 mapByte = HEAPBLK_TO_MAPBYTE(heapBlk);
diff --git a/src/backend/access/rmgrdesc/heapdesc.c b/src/backend/access/rmgrdesc/heapdesc.c
index 02ae91653c1..75ae6f9d375 100644
--- a/src/backend/access/rmgrdesc/heapdesc.c
+++ b/src/backend/access/rmgrdesc/heapdesc.c
@@ -349,13 +349,6 @@ heap2_desc(StringInfo buf, XLogReaderState *record)
}
}
}
- else if (info == XLOG_HEAP2_VISIBLE)
- {
- xl_heap_visible *xlrec = (xl_heap_visible *) rec;
-
- appendStringInfo(buf, "snapshotConflictHorizon: %u, flags: 0x%02X",
- xlrec->snapshotConflictHorizon, xlrec->flags);
- }
else if (info == XLOG_HEAP2_MULTI_INSERT)
{
xl_heap_multi_insert *xlrec = (xl_heap_multi_insert *) rec;
@@ -461,9 +454,6 @@ heap2_identify(uint8 info)
case XLOG_HEAP2_PRUNE_VACUUM_CLEANUP:
id = "PRUNE_VACUUM_CLEANUP";
break;
- case XLOG_HEAP2_VISIBLE:
- id = "VISIBLE";
- break;
case XLOG_HEAP2_MULTI_INSERT:
id = "MULTI_INSERT";
break;
diff --git a/src/backend/replication/logical/decode.c b/src/backend/replication/logical/decode.c
index e25dd6bc366..f7ddb56fc30 100644
--- a/src/backend/replication/logical/decode.c
+++ b/src/backend/replication/logical/decode.c
@@ -449,7 +449,6 @@ heap2_decode(LogicalDecodingContext *ctx, XLogRecordBuffer *buf)
case XLOG_HEAP2_PRUNE_ON_ACCESS:
case XLOG_HEAP2_PRUNE_VACUUM_SCAN:
case XLOG_HEAP2_PRUNE_VACUUM_CLEANUP:
- case XLOG_HEAP2_VISIBLE:
case XLOG_HEAP2_LOCK_UPDATED:
break;
default:
diff --git a/src/backend/storage/ipc/standby.c b/src/backend/storage/ipc/standby.c
index afffab77106..f8681dcc9c7 100644
--- a/src/backend/storage/ipc/standby.c
+++ b/src/backend/storage/ipc/standby.c
@@ -475,12 +475,12 @@ ResolveRecoveryConflictWithSnapshot(TransactionId snapshotConflictHorizon,
* If we get passed InvalidTransactionId then we do nothing (no conflict).
*
* This can happen when replaying already-applied WAL records after a
- * standby crash or restart, or when replaying an XLOG_HEAP2_VISIBLE
- * record that marks as frozen a page which was already all-visible. It's
- * also quite common with records generated during index deletion
- * (original execution of the deletion can reason that a recovery conflict
- * which is sufficient for the deletion operation must take place before
- * replay of the deletion record itself).
+ * standby crash or restart, or when replaying a record that marks as
+ * frozen a page which was already marked all-visible in the visibility
+ * map. It's also quite common with records generated during index
+ * deletion (original execution of the deletion can reason that a recovery
+ * conflict which is sufficient for the deletion operation must take place
+ * before replay of the deletion record itself).
*/
if (!TransactionIdIsValid(snapshotConflictHorizon))
return;
diff --git a/src/include/access/heapam_xlog.h b/src/include/access/heapam_xlog.h
index ce3566ba949..5eed567a8e5 100644
--- a/src/include/access/heapam_xlog.h
+++ b/src/include/access/heapam_xlog.h
@@ -60,7 +60,6 @@
#define XLOG_HEAP2_PRUNE_ON_ACCESS 0x10
#define XLOG_HEAP2_PRUNE_VACUUM_SCAN 0x20
#define XLOG_HEAP2_PRUNE_VACUUM_CLEANUP 0x30
-#define XLOG_HEAP2_VISIBLE 0x40
#define XLOG_HEAP2_MULTI_INSERT 0x50
#define XLOG_HEAP2_LOCK_UPDATED 0x60
#define XLOG_HEAP2_NEW_CID 0x70
@@ -443,20 +442,6 @@ typedef struct xl_heap_inplace
#define MinSizeOfHeapInplace (offsetof(xl_heap_inplace, nmsgs) + sizeof(int))
-/*
- * This is what we need to know about setting a visibility map bit
- *
- * Backup blk 0: visibility map buffer
- * Backup blk 1: heap buffer
- */
-typedef struct xl_heap_visible
-{
- TransactionId snapshotConflictHorizon;
- uint8 flags;
-} xl_heap_visible;
-
-#define SizeOfHeapVisible (offsetof(xl_heap_visible, flags) + sizeof(uint8))
-
typedef struct xl_heap_new_cid
{
/*
@@ -500,11 +485,6 @@ extern void heap2_desc(StringInfo buf, XLogReaderState *record);
extern const char *heap2_identify(uint8 info);
extern void heap_xlog_logical_rewrite(XLogReaderState *r);
-extern XLogRecPtr log_heap_visible(Relation rel, Buffer heap_buffer,
- Buffer vm_buffer,
- TransactionId snapshotConflictHorizon,
- uint8 vmflags);
-
/* in heapdesc.c, so it can be shared between frontend/backend code */
extern void heap_xlog_deserialize_prune_and_freeze(char *cursor, uint16 flags,
int *nplans, xlhp_freeze_plan **plans,
diff --git a/src/include/access/visibilitymap.h b/src/include/access/visibilitymap.h
index a0166c5b410..001afb037f3 100644
--- a/src/include/access/visibilitymap.h
+++ b/src/include/access/visibilitymap.h
@@ -15,7 +15,6 @@
#define VISIBILITYMAP_H
#include "access/visibilitymapdefs.h"
-#include "access/xlogdefs.h"
#include "storage/block.h"
#include "storage/buf.h"
#include "storage/relfilelocator.h"
@@ -32,15 +31,9 @@ extern bool visibilitymap_clear(Relation rel, BlockNumber heapBlk,
extern void visibilitymap_pin(Relation rel, BlockNumber heapBlk,
Buffer *vmbuf);
extern bool visibilitymap_pin_ok(BlockNumber heapBlk, Buffer vmbuf);
-extern void visibilitymap_set(Relation rel,
- BlockNumber heapBlk, Buffer heapBuf,
- XLogRecPtr recptr,
- Buffer vmBuf,
- TransactionId cutoff_xid,
- uint8 flags);
-extern void visibilitymap_set_vmbits(BlockNumber heapBlk,
- Buffer vmBuf, uint8 flags,
- const RelFileLocator rlocator);
+extern void visibilitymap_set(BlockNumber heapBlk,
+ Buffer vmBuf, uint8 flags,
+ const RelFileLocator rlocator);
extern uint8 visibilitymap_get_status(Relation rel, BlockNumber heapBlk, Buffer *vmbuf);
extern void visibilitymap_count(Relation rel, BlockNumber *all_visible, BlockNumber *all_frozen);
extern BlockNumber visibilitymap_prepare_truncate(Relation rel,
diff --git a/src/include/access/visibilitymapdefs.h b/src/include/access/visibilitymapdefs.h
index 89153b3cd9a..e5794c8559e 100644
--- a/src/include/access/visibilitymapdefs.h
+++ b/src/include/access/visibilitymapdefs.h
@@ -21,14 +21,5 @@
#define VISIBILITYMAP_ALL_FROZEN 0x02
#define VISIBILITYMAP_VALID_BITS 0x03 /* OR of all valid visibilitymap
* flags bits */
-/*
- * To detect recovery conflicts during logical decoding on a standby, we need
- * to know if a table is a user catalog table. For that we add an additional
- * bit into xl_heap_visible.flags, in addition to the above.
- *
- * NB: VISIBILITYMAP_XLOG_* may not be passed to visibilitymap_set().
- */
-#define VISIBILITYMAP_XLOG_CATALOG_REL 0x04
-#define VISIBILITYMAP_XLOG_VALID_BITS (VISIBILITYMAP_VALID_BITS | VISIBILITYMAP_XLOG_CATALOG_REL)
#endif /* VISIBILITYMAPDEFS_H */
diff --git a/src/tools/pgindent/typedefs.list b/src/tools/pgindent/typedefs.list
index ddbe4c64971..a85b41e006b 100644
--- a/src/tools/pgindent/typedefs.list
+++ b/src/tools/pgindent/typedefs.list
@@ -4342,7 +4342,6 @@ xl_heap_prune
xl_heap_rewrite_mapping
xl_heap_truncate
xl_heap_update
-xl_heap_visible
xl_invalid_page
xl_invalid_page_key
xl_invalidations
--
2.43.0
[text/x-patch] v34-0007-Simplify-heap_page_would_be_all_visible-visibili.patch (1.9K, 8-v34-0007-Simplify-heap_page_would_be_all_visible-visibili.patch)
download | inline diff:
From a64707f1f2fa88d7292f7a2f2a760c613eea4950 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Wed, 17 Dec 2025 13:57:16 -0500
Subject: [PATCH v34 07/14] Simplify heap_page_would_be_all_visible visibility
check
heap_page_would_be_all_visible() doesn't care about the distinction
between HEAPTUPLE_RECENTLY_DEAD and HEAPTUPLE_DEAD tuples -- any tuple
that is not HEAPTUPLE_LIVE means the page is not all-visible and causes
us to return false.
Therefore, we don't need to call HeapTupleSatisfiesVacuum(), which
includes an extra step to distinguish between dead and recently dead
tuples using OldestXmin. Replace it with the more minimal
HeapTupleSatisfiesVacuumHorizon().
This has the added benefit of making it easier to replace uses of
OldestXmin in heap_page_would_be_all_visible() in the future.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Kirill Reshke <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/CALdSSPjvhGXihT_9f-GJabYU%3D_PjrFDUxYaURuTbfLyQM6TErg%40mail.gmail.com
---
src/backend/access/heap/vacuumlazy.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index d4624010123..5d7ec5c0240 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -3585,6 +3585,7 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
{
ItemId itemid;
HeapTupleData tuple;
+ TransactionId dead_after;
/*
* Set the offset number so that we can display it along with any
@@ -3624,7 +3625,7 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
/* Visibility checks may do IO or allocate memory */
Assert(CritSectionCount == 0);
- switch (HeapTupleSatisfiesVacuum(&tuple, OldestXmin, buf))
+ switch (HeapTupleSatisfiesVacuumHorizon(&tuple, buf, &dead_after))
{
case HEAPTUPLE_LIVE:
{
--
2.43.0
[text/x-patch] v34-0008-Remove-table_scan_analyze_next_tuple-unneeded-pa.patch (4.4K, 9-v34-0008-Remove-table_scan_analyze_next_tuple-unneeded-pa.patch)
download | inline diff:
From 85ab0d4eb681eaba4668ee23602d425c27f56d07 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Mon, 22 Dec 2025 10:46:45 -0500
Subject: [PATCH v34 08/14] Remove table_scan_analyze_next_tuple unneeded
parameter OldestXmin
heapam_scan_analyze_next_tuple() doesn't distinguish between dead and
recently dead tuples when counting them, so it doesn't need OldestXmin.
Looking at other table AMs implementing table_scan_analyze_next_tuple(),
it appears most do not use OldestXmin either.
Suggested-by: Kirill Reshke <[email protected]>
Reviewed-by: Kirill Reshke <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/CALdSSPjvhGXihT_9f-GJabYU%3D_PjrFDUxYaURuTbfLyQM6TErg%40mail.gmail.com
---
src/backend/access/heap/heapam_handler.c | 8 +++++---
src/backend/commands/analyze.c | 6 +-----
src/include/access/tableam.h | 5 ++---
3 files changed, 8 insertions(+), 11 deletions(-)
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index cbef73e5d4b..da4ae236ca8 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -1040,7 +1040,7 @@ heapam_scan_analyze_next_block(TableScanDesc scan, ReadStream *stream)
}
static bool
-heapam_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
+heapam_scan_analyze_next_tuple(TableScanDesc scan,
double *liverows, double *deadrows,
TupleTableSlot *slot)
{
@@ -1061,6 +1061,7 @@ heapam_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
ItemId itemid;
HeapTuple targtuple = &hslot->base.tupdata;
bool sample_it = false;
+ TransactionId dead_after;
itemid = PageGetItemId(targpage, hscan->rs_cindex);
@@ -1083,8 +1084,9 @@ heapam_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
targtuple->t_data = (HeapTupleHeader) PageGetItem(targpage, itemid);
targtuple->t_len = ItemIdGetLength(itemid);
- switch (HeapTupleSatisfiesVacuum(targtuple, OldestXmin,
- hscan->rs_cbuf))
+ switch (HeapTupleSatisfiesVacuumHorizon(targtuple,
+ hscan->rs_cbuf,
+ &dead_after))
{
case HEAPTUPLE_LIVE:
sample_it = true;
diff --git a/src/backend/commands/analyze.c b/src/backend/commands/analyze.c
index a483424152c..53adac9139b 100644
--- a/src/backend/commands/analyze.c
+++ b/src/backend/commands/analyze.c
@@ -1213,7 +1213,6 @@ acquire_sample_rows(Relation onerel, int elevel,
double rowstoskip = -1; /* -1 means not set yet */
uint32 randseed; /* Seed for block sampler(s) */
BlockNumber totalblocks;
- TransactionId OldestXmin;
BlockSamplerData bs;
ReservoirStateData rstate;
TupleTableSlot *slot;
@@ -1226,9 +1225,6 @@ acquire_sample_rows(Relation onerel, int elevel,
totalblocks = RelationGetNumberOfBlocks(onerel);
- /* Need a cutoff xmin for HeapTupleSatisfiesVacuum */
- OldestXmin = GetOldestNonRemovableTransactionId(onerel);
-
/* Prepare for sampling block numbers */
randseed = pg_prng_uint32(&pg_global_prng_state);
nblocks = BlockSampler_Init(&bs, totalblocks, targrows, randseed);
@@ -1261,7 +1257,7 @@ acquire_sample_rows(Relation onerel, int elevel,
{
vacuum_delay_point(true);
- while (table_scan_analyze_next_tuple(scan, OldestXmin, &liverows, &deadrows, slot))
+ while (table_scan_analyze_next_tuple(scan, &liverows, &deadrows, slot))
{
/*
* The first targrows sample rows are simply copied into the
diff --git a/src/include/access/tableam.h b/src/include/access/tableam.h
index e2ec5289d4d..c9fa9f259cd 100644
--- a/src/include/access/tableam.h
+++ b/src/include/access/tableam.h
@@ -683,7 +683,6 @@ typedef struct TableAmRoutine
* callback).
*/
bool (*scan_analyze_next_tuple) (TableScanDesc scan,
- TransactionId OldestXmin,
double *liverows,
double *deadrows,
TupleTableSlot *slot);
@@ -1714,11 +1713,11 @@ table_scan_analyze_next_block(TableScanDesc scan, ReadStream *stream)
* tuples.
*/
static inline bool
-table_scan_analyze_next_tuple(TableScanDesc scan, TransactionId OldestXmin,
+table_scan_analyze_next_tuple(TableScanDesc scan,
double *liverows, double *deadrows,
TupleTableSlot *slot)
{
- return scan->rs_rd->rd_tableam->scan_analyze_next_tuple(scan, OldestXmin,
+ return scan->rs_rd->rd_tableam->scan_analyze_next_tuple(scan,
liverows, deadrows,
slot);
}
--
2.43.0
[text/x-patch] v34-0009-Use-GlobalVisState-in-vacuum-to-determine-page-l.patch (12.4K, 10-v34-0009-Use-GlobalVisState-in-vacuum-to-determine-page-l.patch)
download | inline diff:
From 8d350868206456f631883a40a955dff480e408d3 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Wed, 17 Dec 2025 16:51:05 -0500
Subject: [PATCH v34 09/14] Use GlobalVisState in vacuum to determine page
level visibility
During vacuum's first and third phases, we examine tuples' visibility
to determine if we can set the page all-visible in the visibility map.
Previously, this check compared tuple xmins against a single XID chosen at
the start of vacuum (OldestXmin). We now use GlobalVisState, which also
enables future work to set the VM during on-access pruning, since ordinary
queries have access to GlobalVisState but not OldestXmin.
This also benefits vacuum: in some cases, GlobalVisState may advance
during a vacuum, allowing more pages to become considered all-visible.
And, in the future, we could easily add a heuristic to update
GlobalVisState more frequently during vacuums of large tables.
OldestXmin is still used for freezing and as a backstop to ensure we
don't freeze a dead tuple that wasn't yet prunable according to
GlobalVisState in the rare occurrences where GlobalVisState moves
backwards.
Because comparing a transaction ID against GlobalVisState is more
expensive than comparing against a single XID, we defer this check until
after scanning all tuples on the page. If visibility_cutoff_xid was
maintained, we perform the GlobalVisState check only once per page.
This is safe because visibility_cutoff_xid records the newest xmin on
the page; if it is globally visible, then the entire page is all-visible.
This approach may result in examining more tuple xmins than before,
since with OldestXmin we could sometimes rule out the page being
all-visible earlier. However, profiling shows the additional cost is not
significant.
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Chao Li <[email protected]>
---
src/backend/access/heap/heapam_visibility.c | 22 ++++++++
src/backend/access/heap/pruneheap.c | 57 +++++++++++----------
src/backend/access/heap/vacuumlazy.c | 44 ++++++++++------
src/include/access/heapam.h | 4 +-
4 files changed, 85 insertions(+), 42 deletions(-)
diff --git a/src/backend/access/heap/heapam_visibility.c b/src/backend/access/heap/heapam_visibility.c
index 75ae268d753..aee88947393 100644
--- a/src/backend/access/heap/heapam_visibility.c
+++ b/src/backend/access/heap/heapam_visibility.c
@@ -1060,6 +1060,28 @@ HeapTupleSatisfiesVacuum(HeapTuple htup, TransactionId OldestXmin,
return res;
}
+/*
+ * Wrapper around GlobalVisTestIsRemovableXid() for use when examining live
+ * tuples. Returns true if the given XID may be considered running by at least
+ * one snapshot.
+ *
+ * This function alone is insufficient to determine tuple visibility; callers
+ * must also consider the XID's commit status. Its purpose is purely semantic:
+ * when applied to live tuples, GlobalVisTestIsRemovableXid() is checking
+ * whether the inserting transaction is still considered running, not whether
+ * the tuple is removable. Live tuples are, by definition, not removable, but
+ * the snapshot criteria for “transaction still running” are identical to
+ * those used for removal XIDs.
+ *
+ * See the comment above GlobalVisTestIsRemovable[Full]Xid() for details on the
+ * required preconditions for calling this function.
+ */
+bool
+GlobalVisTestXidMaybeRunning(GlobalVisState *state, TransactionId xid)
+{
+ return !GlobalVisTestIsRemovableXid(state, xid);
+}
+
/*
* Work horse for HeapTupleSatisfiesVacuum and similar routines.
*
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index 2ba863be07c..c5bf0899c89 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -461,11 +461,12 @@ prune_freeze_setup(PruneFreezeParams *params,
/*
* The visibility cutoff xid is the newest xmin of live, committed tuples
- * older than OldestXmin on the page. This field is only kept up-to-date
- * if the page is all-visible. As soon as a tuple is encountered that is
- * not visible to all, this field is unmaintained. As long as it is
- * maintained, it can be used to calculate the snapshot conflict horizon
- * when updating the VM and/or freezing all the tuples on the page.
+ * on the page older than the visibility horizon represented in the
+ * GlobalVisState. This field is only kept up-to-date if the page is
+ * all-visible. It is invalid if there are any tuples on the page that are
+ * not visible to all. As long as it is maintained, it can be used to
+ * calculate the snapshot conflict horizon when updating the VM and/or
+ * freezing all the tuples on the page.
*/
prstate->visibility_cutoff_xid = InvalidTransactionId;
}
@@ -1077,6 +1078,24 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
prune_freeze_plan(RelationGetRelid(params->relation),
buffer, &prstate, off_loc);
+ /*
+ * After processing all the live tuples on the page, if the newest xmin
+ * amongst them may be considered running by any snapshot, the page cannot
+ * be all-visible.
+ */
+ if (prstate.all_visible &&
+ TransactionIdIsNormal(prstate.visibility_cutoff_xid) &&
+ GlobalVisTestXidMaybeRunning(prstate.vistest, prstate.visibility_cutoff_xid))
+ {
+ prstate.all_visible = prstate.all_frozen = false;
+
+ /*
+ * We won't try to use this if all_visible is false, but better to be
+ * safe and invalidate it.
+ */
+ prstate.visibility_cutoff_xid = InvalidTransactionId;
+ }
+
/*
* If checksums are enabled, calling heap_prune_satisfies_vacuum() while
* checking tuple visibility information in prune_freeze_plan() may have
@@ -1259,10 +1278,9 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
bool debug_all_frozen;
Assert(prstate.lpdead_items == 0);
- Assert(prstate.cutoffs);
Assert(heap_page_is_all_visible(params->relation, buffer,
- prstate.cutoffs->OldestXmin,
+ prstate.vistest,
&debug_all_frozen,
&debug_cutoff, off_loc));
@@ -1794,28 +1812,15 @@ heap_prune_record_unchanged_lp_normal(Page page, PruneState *prstate, OffsetNumb
}
/*
- * The inserter definitely committed. But is it old enough
- * that everyone sees it as committed? A FrozenTransactionId
- * is seen as committed to everyone. Otherwise, we check if
- * there is a snapshot that considers this xid to still be
- * running, and if so, we don't consider the page all-visible.
+ * The inserter definitely committed. But we don't know if it
+ * is old enough that everyone sees it as committed. Later,
+ * after processing all the tuples on the page, we'll check if
+ * there is any snapshot that still considers the newest xid
+ * on the page to be running. If so, we don't consider the
+ * page all-visible.
*/
xmin = HeapTupleHeaderGetXmin(htup);
- /*
- * For now always use prstate->cutoffs for this test, because
- * we only update 'all_visible' and 'all_frozen' when freezing
- * is requested. We could use GlobalVisTestIsRemovableXid
- * instead, if a non-freezing caller wanted to set the VM bit.
- */
- Assert(prstate->cutoffs);
- if (!TransactionIdPrecedes(xmin, prstate->cutoffs->OldestXmin))
- {
- prstate->all_visible = false;
- prstate->all_frozen = false;
- break;
- }
-
/* Track newest xmin on page. */
if (TransactionIdFollows(xmin, prstate->visibility_cutoff_xid) &&
TransactionIdIsNormal(xmin))
diff --git a/src/backend/access/heap/vacuumlazy.c b/src/backend/access/heap/vacuumlazy.c
index 5d7ec5c0240..96ed31c9570 100644
--- a/src/backend/access/heap/vacuumlazy.c
+++ b/src/backend/access/heap/vacuumlazy.c
@@ -457,7 +457,7 @@ static void dead_items_reset(LVRelState *vacrel);
static void dead_items_cleanup(LVRelState *vacrel);
static bool heap_page_would_be_all_visible(Relation rel, Buffer buf,
- TransactionId OldestXmin,
+ GlobalVisState *vistest,
OffsetNumber *deadoffsets,
int ndeadoffsets,
bool *all_frozen,
@@ -2743,7 +2743,7 @@ lazy_vacuum_heap_page(LVRelState *vacrel, BlockNumber blkno, Buffer buffer,
* done outside the critical section.
*/
if (heap_page_would_be_all_visible(vacrel->rel, buffer,
- vacrel->cutoffs.OldestXmin,
+ vacrel->vistest,
deadoffsets, num_offsets,
&all_frozen, &visibility_cutoff_xid,
&vacrel->offnum))
@@ -3503,14 +3503,14 @@ dead_items_cleanup(LVRelState *vacrel)
*/
bool
heap_page_is_all_visible(Relation rel, Buffer buf,
- TransactionId OldestXmin,
+ GlobalVisState *vistest,
bool *all_frozen,
TransactionId *visibility_cutoff_xid,
OffsetNumber *logging_offnum)
{
return heap_page_would_be_all_visible(rel, buf,
- OldestXmin,
+ vistest,
NULL, 0,
all_frozen,
visibility_cutoff_xid,
@@ -3531,7 +3531,7 @@ heap_page_is_all_visible(Relation rel, Buffer buf,
* Returns true if the page is all-visible other than the provided
* deadoffsets and false otherwise.
*
- * OldestXmin is used to determine visibility.
+ * vistest is used to determine visibility.
*
* Output parameters:
*
@@ -3550,7 +3550,7 @@ heap_page_is_all_visible(Relation rel, Buffer buf,
*/
static bool
heap_page_would_be_all_visible(Relation rel, Buffer buf,
- TransactionId OldestXmin,
+ GlobalVisState *vistest,
OffsetNumber *deadoffsets,
int ndeadoffsets,
bool *all_frozen,
@@ -3631,7 +3631,7 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
{
TransactionId xmin;
- /* Check comments in lazy_scan_prune. */
+ /* Check heap_prune_record_unchanged_lp_normal comments */
if (!HeapTupleHeaderXminCommitted(tuple.t_data))
{
all_visible = false;
@@ -3640,16 +3640,17 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
}
/*
- * The inserter definitely committed. But is it old enough
- * that everyone sees it as committed?
+ * The inserter definitely committed. But we don't know if
+ * it is old enough that everyone sees it as committed.
+ * Don't check that now.
+ *
+ * If we scan all tuples without finding one that prevents
+ * the page from being all-visible, we then check whether
+ * any snapshot still considers the newest XID on the page
+ * to be running. In that case, the page is not considered
+ * all-visible.
*/
xmin = HeapTupleHeaderGetXmin(tuple.t_data);
- if (!TransactionIdPrecedes(xmin, OldestXmin))
- {
- all_visible = false;
- *all_frozen = false;
- break;
- }
/* Track newest xmin on page. */
if (TransactionIdFollows(xmin, *visibility_cutoff_xid) &&
@@ -3678,6 +3679,19 @@ heap_page_would_be_all_visible(Relation rel, Buffer buf,
}
} /* scan along page */
+ /*
+ * After processing all the live tuples on the page, if the newest xmin
+ * among them may still be considered running by any snapshot, the page
+ * cannot be all-visible.
+ */
+ if (all_visible &&
+ TransactionIdIsNormal(*visibility_cutoff_xid) &&
+ GlobalVisTestXidMaybeRunning(vistest, *visibility_cutoff_xid))
+ {
+ all_visible = false;
+ *all_frozen = false;
+ }
+
/* Clear the offset information once we have processed the given page. */
*logging_offnum = InvalidOffsetNumber;
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index a0f7974942e..05ef8d8cd5e 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -439,7 +439,7 @@ extern void heap_vacuum_rel(Relation rel,
const VacuumParams params, BufferAccessStrategy bstrategy);
#ifdef USE_ASSERT_CHECKING
extern bool heap_page_is_all_visible(Relation rel, Buffer buf,
- TransactionId OldestXmin,
+ GlobalVisState *vistest,
bool *all_frozen,
TransactionId *visibility_cutoff_xid,
OffsetNumber *logging_offnum);
@@ -451,6 +451,8 @@ extern TM_Result HeapTupleSatisfiesUpdate(HeapTuple htup, CommandId curcid,
Buffer buffer);
extern HTSV_Result HeapTupleSatisfiesVacuum(HeapTuple htup, TransactionId OldestXmin,
Buffer buffer);
+
+extern bool GlobalVisTestXidMaybeRunning(GlobalVisState *state, TransactionId xid);
extern HTSV_Result HeapTupleSatisfiesVacuumHorizon(HeapTuple htup, Buffer buffer,
TransactionId *dead_after);
extern void HeapTupleSetHintBits(HeapTupleHeader tuple, Buffer buffer,
--
2.43.0
[text/x-patch] v34-0010-Unset-all_visible-sooner-if-not-freezing.patch (2.5K, 11-v34-0010-Unset-all_visible-sooner-if-not-freezing.patch)
download | inline diff:
From 8a0d8c5726f4c5821eb59396c1b48265a474a588 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Tue, 14 Oct 2025 15:22:35 -0400
Subject: [PATCH v34 10/14] Unset all_visible sooner if not freezing
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
In the prune/freeze path, we currently delay clearing all_visible and
all_frozen in the presence of dead items to allow opportunistic
freezing.
However, if no freezing will be attempted, there’s no need to delay.
Clearing the flags earlier avoids extra bookkeeping in
heap_prune_record_unchanged_lp_normal(). This currently has no runtime
effect because all callers that consider setting the VM also prepare
freeze plans, but upcoming changes will allow on-access pruning to set
the VM without freezing. The extra bookkeeping was noticeable in a
profile of on-access VM setting.
Reviewed-by: Chao Li <[email protected]>
---
src/backend/access/heap/pruneheap.c | 15 +++++++++++++--
1 file changed, 13 insertions(+), 2 deletions(-)
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index c5bf0899c89..cc45728d25e 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -1677,8 +1677,13 @@ heap_prune_record_dead(PruneState *prstate, OffsetNumber offnum,
/*
* Deliberately delay unsetting all_visible and all_frozen until later
* during pruning. Removable dead tuples shouldn't preclude freezing the
- * page.
+ * page. If we won't attempt freezing, just unset all-visible now, though.
*/
+ if (!prstate->attempt_freeze)
+ {
+ prstate->all_visible = false;
+ prstate->all_frozen = false;
+ }
/* Record the dead offset for vacuum */
prstate->deadoffsets[prstate->lpdead_items++] = offnum;
@@ -1938,8 +1943,14 @@ heap_prune_record_unchanged_lp_dead(Page page, PruneState *prstate, OffsetNumber
* Similarly, don't unset all_visible and all_frozen until later, at the
* end of heap_page_prune_and_freeze(). This will allow us to attempt to
* freeze the page after pruning. As long as we unset it before updating
- * the visibility map, this will be correct.
+ * the visibility map, this will be correct. If we won't attempt freezing,
+ * though, just unset all_visible and all_frozen now.
*/
+ if (!prstate->attempt_freeze)
+ {
+ prstate->all_visible = false;
+ prstate->all_frozen = false;
+ }
/* Record the dead offset for vacuum */
prstate->deadoffsets[prstate->lpdead_items++] = offnum;
--
2.43.0
[text/x-patch] v34-0011-Track-which-relations-are-modified-by-a-query.patch (2.6K, 12-v34-0011-Track-which-relations-are-modified-by-a-query.patch)
download | inline diff:
From 38c5a2bb61ce9df0035d01a37e1f5e5e806cb5ff Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Wed, 3 Dec 2025 15:07:24 -0500
Subject: [PATCH v34 11/14] Track which relations are modified by a query
Save the relids in a bitmap in the estate. A later commit will pass this
information down to scan nodes to control whether or not the scan allows
setting the visibility map while on-access pruning. We don't want to set
the visibility map if the query is just going to modify the page
immediately after.
Reviewed-by: Chao Li <[email protected]>
---
src/backend/executor/execMain.c | 4 ++++
src/backend/executor/execUtils.c | 2 ++
src/include/nodes/execnodes.h | 6 ++++++
3 files changed, 12 insertions(+)
diff --git a/src/backend/executor/execMain.c b/src/backend/executor/execMain.c
index bfd3ebc601e..87772de3d33 100644
--- a/src/backend/executor/execMain.c
+++ b/src/backend/executor/execMain.c
@@ -920,6 +920,10 @@ InitPlan(QueryDesc *queryDesc, int eflags)
break;
}
+ /* If it has a rowmark, the relation may be modified */
+ estate->es_modified_relids = bms_add_member(estate->es_modified_relids,
+ rc->rti);
+
/* Check that relation is a legal target for marking */
if (relation)
CheckValidRowMarkRel(relation, rc->markType);
diff --git a/src/backend/executor/execUtils.c b/src/backend/executor/execUtils.c
index a7955e476f9..ac0ed6c68eb 100644
--- a/src/backend/executor/execUtils.c
+++ b/src/backend/executor/execUtils.c
@@ -893,6 +893,8 @@ ExecInitResultRelation(EState *estate, ResultRelInfo *resultRelInfo,
estate->es_result_relations = (ResultRelInfo **)
palloc0(estate->es_range_table_size * sizeof(ResultRelInfo *));
estate->es_result_relations[rti - 1] = resultRelInfo;
+ estate->es_modified_relids = bms_add_member(estate->es_modified_relids,
+ rti);
/*
* Saving in the list allows to avoid needlessly traversing the whole
diff --git a/src/include/nodes/execnodes.h b/src/include/nodes/execnodes.h
index f8053d9e572..1e3cd73cf27 100644
--- a/src/include/nodes/execnodes.h
+++ b/src/include/nodes/execnodes.h
@@ -678,6 +678,12 @@ typedef struct EState
* ExecDoInitialPruning() */
const char *es_sourceText; /* Source text from QueryDesc */
+ /*
+ * RT indexes of relations modified by the query through a
+ * UPDATE/DELETE/INSERT/MERGE or targeted by a SELECT FOR UPDATE.
+ */
+ Bitmapset *es_modified_relids;
+
JunkFilter *es_junkFilter; /* top-level junk filter, if any */
/* If query can insert/delete tuples, the command ID to mark them with */
--
2.43.0
[text/x-patch] v34-0012-Pass-down-information-on-table-modification-to-s.patch (24.8K, 13-v34-0012-Pass-down-information-on-table-modification-to-s.patch)
download | inline diff:
From 8205b2d7da0c3ad3cbc5cead336ced677996b37d Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Wed, 3 Dec 2025 15:12:18 -0500
Subject: [PATCH v34 12/14] Pass down information on table modification to scan
node
Pass down information to sequential scan, index [only] scan, and bitmap
table scan nodes on whether or not the query modifies the relation being
scanned. A later commit will use this information to update the VM
during on-access pruning only if the relation is not modified by the
query.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andrey Borodin <[email protected]>
Reviewed-by: Chao Li <[email protected]>
Discussion: https://postgr.es/m/4379FDA3-9446-4E2C-9C15-32EFE8D4F31B%40yandex-team.ru
---
contrib/pgrowlocks/pgrowlocks.c | 2 +-
src/backend/access/brin/brin.c | 3 ++-
src/backend/access/gin/gininsert.c | 3 ++-
src/backend/access/heap/heapam_handler.c | 7 +++---
src/backend/access/index/genam.c | 4 ++--
src/backend/access/index/indexam.c | 6 +++---
src/backend/access/nbtree/nbtsort.c | 2 +-
src/backend/access/table/tableam.c | 7 +++---
src/backend/commands/constraint.c | 2 +-
src/backend/commands/copyto.c | 2 +-
src/backend/commands/tablecmds.c | 8 +++----
src/backend/commands/typecmds.c | 4 ++--
src/backend/executor/execIndexing.c | 2 +-
src/backend/executor/execReplication.c | 8 +++----
src/backend/executor/nodeBitmapHeapscan.c | 9 +++++++-
src/backend/executor/nodeIndexonlyscan.c | 9 +++++++-
src/backend/executor/nodeIndexscan.c | 18 ++++++++++++++--
src/backend/executor/nodeSeqscan.c | 26 ++++++++++++++++++++---
src/backend/partitioning/partbounds.c | 2 +-
src/backend/utils/adt/selfuncs.c | 2 +-
src/include/access/genam.h | 2 +-
src/include/access/heapam.h | 6 ++++++
src/include/access/tableam.h | 19 ++++++++++-------
23 files changed, 107 insertions(+), 46 deletions(-)
diff --git a/contrib/pgrowlocks/pgrowlocks.c b/contrib/pgrowlocks/pgrowlocks.c
index f88269332b6..27f01d8055f 100644
--- a/contrib/pgrowlocks/pgrowlocks.c
+++ b/contrib/pgrowlocks/pgrowlocks.c
@@ -114,7 +114,7 @@ pgrowlocks(PG_FUNCTION_ARGS)
RelationGetRelationName(rel));
/* Scan the relation */
- scan = table_beginscan(rel, GetActiveSnapshot(), 0, NULL);
+ scan = table_beginscan(rel, GetActiveSnapshot(), 0, NULL, 0);
hscan = (HeapScanDesc) scan;
attinmeta = TupleDescGetAttInMetadata(rsinfo->setDesc);
diff --git a/src/backend/access/brin/brin.c b/src/backend/access/brin/brin.c
index 6887e421442..4d9684b1b19 100644
--- a/src/backend/access/brin/brin.c
+++ b/src/backend/access/brin/brin.c
@@ -2843,7 +2843,8 @@ _brin_parallel_scan_and_build(BrinBuildState *state,
indexInfo->ii_Concurrent = brinshared->isconcurrent;
scan = table_beginscan_parallel(heap,
- ParallelTableScanFromBrinShared(brinshared));
+ ParallelTableScanFromBrinShared(brinshared),
+ 0);
reltuples = table_index_build_scan(heap, index, indexInfo, true, true,
brinbuildCallbackParallel, state, scan);
diff --git a/src/backend/access/gin/gininsert.c b/src/backend/access/gin/gininsert.c
index 0d63fb4ba27..f02d6df40a2 100644
--- a/src/backend/access/gin/gininsert.c
+++ b/src/backend/access/gin/gininsert.c
@@ -2059,7 +2059,8 @@ _gin_parallel_scan_and_build(GinBuildState *state,
indexInfo->ii_Concurrent = ginshared->isconcurrent;
scan = table_beginscan_parallel(heap,
- ParallelTableScanFromGinBuildShared(ginshared));
+ ParallelTableScanFromGinBuildShared(ginshared),
+ 0);
reltuples = table_index_build_scan(heap, index, indexInfo, true, progress,
ginBuildCallbackParallel, state, scan);
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index da4ae236ca8..6ce4350c2c8 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -79,12 +79,13 @@ heapam_slot_callbacks(Relation relation)
*/
static IndexFetchTableData *
-heapam_index_fetch_begin(Relation rel)
+heapam_index_fetch_begin(Relation rel, uint32 flags)
{
IndexFetchHeapData *hscan = palloc0_object(IndexFetchHeapData);
hscan->xs_base.rel = rel;
hscan->xs_cbuf = InvalidBuffer;
+ hscan->modifies_base_rel = !(flags & SO_HINT_REL_READ_ONLY);
return &hscan->xs_base;
}
@@ -753,7 +754,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
tableScan = NULL;
heapScan = NULL;
- indexScan = index_beginscan(OldHeap, OldIndex, SnapshotAny, NULL, 0, 0);
+ indexScan = index_beginscan(OldHeap, OldIndex, SnapshotAny, NULL, 0, 0, 0);
index_rescan(indexScan, NULL, 0, NULL, 0);
}
else
@@ -762,7 +763,7 @@ heapam_relation_copy_for_cluster(Relation OldHeap, Relation NewHeap,
pgstat_progress_update_param(PROGRESS_CLUSTER_PHASE,
PROGRESS_CLUSTER_PHASE_SEQ_SCAN_HEAP);
- tableScan = table_beginscan(OldHeap, SnapshotAny, 0, (ScanKey) NULL);
+ tableScan = table_beginscan(OldHeap, SnapshotAny, 0, (ScanKey) NULL, 0);
heapScan = (HeapScanDesc) tableScan;
indexScan = NULL;
diff --git a/src/backend/access/index/genam.c b/src/backend/access/index/genam.c
index a29be6f467b..5ac7d22e49f 100644
--- a/src/backend/access/index/genam.c
+++ b/src/backend/access/index/genam.c
@@ -447,7 +447,7 @@ systable_beginscan(Relation heapRelation,
}
sysscan->iscan = index_beginscan(heapRelation, irel,
- snapshot, NULL, nkeys, 0);
+ snapshot, NULL, nkeys, 0, 0);
index_rescan(sysscan->iscan, idxkey, nkeys, NULL, 0);
sysscan->scan = NULL;
@@ -708,7 +708,7 @@ systable_beginscan_ordered(Relation heapRelation,
}
sysscan->iscan = index_beginscan(heapRelation, indexRelation,
- snapshot, NULL, nkeys, 0);
+ snapshot, NULL, nkeys, 0, 0);
index_rescan(sysscan->iscan, idxkey, nkeys, NULL, 0);
sysscan->scan = NULL;
diff --git a/src/backend/access/index/indexam.c b/src/backend/access/index/indexam.c
index 4ed0508c605..4df56087841 100644
--- a/src/backend/access/index/indexam.c
+++ b/src/backend/access/index/indexam.c
@@ -257,7 +257,7 @@ index_beginscan(Relation heapRelation,
Relation indexRelation,
Snapshot snapshot,
IndexScanInstrumentation *instrument,
- int nkeys, int norderbys)
+ int nkeys, int norderbys, uint32 flags)
{
IndexScanDesc scan;
@@ -284,7 +284,7 @@ index_beginscan(Relation heapRelation,
scan->instrument = instrument;
/* prepare to fetch index matches from table */
- scan->xs_heapfetch = table_index_fetch_begin(heapRelation);
+ scan->xs_heapfetch = table_index_fetch_begin(heapRelation, flags);
return scan;
}
@@ -615,7 +615,7 @@ index_beginscan_parallel(Relation heaprel, Relation indexrel,
scan->instrument = instrument;
/* prepare to fetch index matches from table */
- scan->xs_heapfetch = table_index_fetch_begin(heaprel);
+ scan->xs_heapfetch = table_index_fetch_begin(heaprel, 0);
return scan;
}
diff --git a/src/backend/access/nbtree/nbtsort.c b/src/backend/access/nbtree/nbtsort.c
index 90ab4e91b56..8ae54217f36 100644
--- a/src/backend/access/nbtree/nbtsort.c
+++ b/src/backend/access/nbtree/nbtsort.c
@@ -1925,7 +1925,7 @@ _bt_parallel_scan_and_sort(BTSpool *btspool, BTSpool *btspool2,
indexInfo = BuildIndexInfo(btspool->index);
indexInfo->ii_Concurrent = btshared->isconcurrent;
scan = table_beginscan_parallel(btspool->heap,
- ParallelTableScanFromBTShared(btshared));
+ ParallelTableScanFromBTShared(btshared), 0);
reltuples = table_index_build_scan(btspool->heap, btspool->index, indexInfo,
true, progress, _bt_build_callback,
&buildstate, scan);
diff --git a/src/backend/access/table/tableam.c b/src/backend/access/table/tableam.c
index 87491796523..2ff29b6e40b 100644
--- a/src/backend/access/table/tableam.c
+++ b/src/backend/access/table/tableam.c
@@ -163,10 +163,11 @@ table_parallelscan_initialize(Relation rel, ParallelTableScanDesc pscan,
}
TableScanDesc
-table_beginscan_parallel(Relation relation, ParallelTableScanDesc pscan)
+table_beginscan_parallel(Relation relation, ParallelTableScanDesc pscan, uint32 flags)
{
Snapshot snapshot;
- uint32 flags = SO_TYPE_SEQSCAN |
+
+ flags |= SO_TYPE_SEQSCAN |
SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE;
Assert(RelFileLocatorEquals(relation->rd_locator, pscan->phs_locator));
@@ -248,7 +249,7 @@ table_index_fetch_tuple_check(Relation rel,
bool found;
slot = table_slot_create(rel, NULL);
- scan = table_index_fetch_begin(rel);
+ scan = table_index_fetch_begin(rel, 0);
found = table_index_fetch_tuple(scan, tid, snapshot, slot, &call_again,
all_dead);
table_index_fetch_end(scan);
diff --git a/src/backend/commands/constraint.c b/src/backend/commands/constraint.c
index cc11c47b6f2..37cfbd63938 100644
--- a/src/backend/commands/constraint.c
+++ b/src/backend/commands/constraint.c
@@ -106,7 +106,7 @@ unique_key_recheck(PG_FUNCTION_ARGS)
*/
tmptid = checktid;
{
- IndexFetchTableData *scan = table_index_fetch_begin(trigdata->tg_relation);
+ IndexFetchTableData *scan = table_index_fetch_begin(trigdata->tg_relation, 0);
bool call_again = false;
if (!table_index_fetch_tuple(scan, &tmptid, SnapshotSelf, slot,
diff --git a/src/backend/commands/copyto.c b/src/backend/commands/copyto.c
index 4ab4a3893d5..4261baf4a41 100644
--- a/src/backend/commands/copyto.c
+++ b/src/backend/commands/copyto.c
@@ -1156,7 +1156,7 @@ CopyRelationTo(CopyToState cstate, Relation rel, Relation root_rel, uint64 *proc
AttrMap *map = NULL;
TupleTableSlot *root_slot = NULL;
- scandesc = table_beginscan(rel, GetActiveSnapshot(), 0, NULL);
+ scandesc = table_beginscan(rel, GetActiveSnapshot(), 0, NULL, 0);
slot = table_slot_create(rel, NULL);
/*
diff --git a/src/backend/commands/tablecmds.c b/src/backend/commands/tablecmds.c
index f976c0e5c7e..eb35dbbc853 100644
--- a/src/backend/commands/tablecmds.c
+++ b/src/backend/commands/tablecmds.c
@@ -6378,7 +6378,7 @@ ATRewriteTable(AlteredTableInfo *tab, Oid OIDNewHeap)
* checking all the constraints.
*/
snapshot = RegisterSnapshot(GetLatestSnapshot());
- scan = table_beginscan(oldrel, snapshot, 0, NULL);
+ scan = table_beginscan(oldrel, snapshot, 0, NULL, 0);
/*
* Switch to per-tuple memory context and reset it for each tuple
@@ -13768,7 +13768,7 @@ validateForeignKeyConstraint(char *conname,
*/
snapshot = RegisterSnapshot(GetLatestSnapshot());
slot = table_slot_create(rel, NULL);
- scan = table_beginscan(rel, snapshot, 0, NULL);
+ scan = table_beginscan(rel, snapshot, 0, NULL, 0);
perTupCxt = AllocSetContextCreate(CurrentMemoryContext,
"validateForeignKeyConstraint",
@@ -22626,7 +22626,7 @@ MergePartitionsMoveRows(List **wqueue, List *mergingPartitions, Relation newPart
/* Scan through the rows. */
snapshot = RegisterSnapshot(GetLatestSnapshot());
- scan = table_beginscan(mergingPartition, snapshot, 0, NULL);
+ scan = table_beginscan(mergingPartition, snapshot, 0, NULL, 0);
/*
* Switch to per-tuple memory context and reset it for each tuple
@@ -23090,7 +23090,7 @@ SplitPartitionMoveRows(List **wqueue, Relation rel, Relation splitRel,
/* Scan through the rows. */
snapshot = RegisterSnapshot(GetLatestSnapshot());
- scan = table_beginscan(splitRel, snapshot, 0, NULL);
+ scan = table_beginscan(splitRel, snapshot, 0, NULL, 0);
/*
* Switch to per-tuple memory context and reset it for each tuple
diff --git a/src/backend/commands/typecmds.c b/src/backend/commands/typecmds.c
index 288edb25f2f..34ff8d041f1 100644
--- a/src/backend/commands/typecmds.c
+++ b/src/backend/commands/typecmds.c
@@ -3179,7 +3179,7 @@ validateDomainNotNullConstraint(Oid domainoid)
/* Scan all tuples in this relation */
snapshot = RegisterSnapshot(GetLatestSnapshot());
- scan = table_beginscan(testrel, snapshot, 0, NULL);
+ scan = table_beginscan(testrel, snapshot, 0, NULL, 0);
slot = table_slot_create(testrel, NULL);
while (table_scan_getnextslot(scan, ForwardScanDirection, slot))
{
@@ -3260,7 +3260,7 @@ validateDomainCheckConstraint(Oid domainoid, const char *ccbin, LOCKMODE lockmod
/* Scan all tuples in this relation */
snapshot = RegisterSnapshot(GetLatestSnapshot());
- scan = table_beginscan(testrel, snapshot, 0, NULL);
+ scan = table_beginscan(testrel, snapshot, 0, NULL, 0);
slot = table_slot_create(testrel, NULL);
while (table_scan_getnextslot(scan, ForwardScanDirection, slot))
{
diff --git a/src/backend/executor/execIndexing.c b/src/backend/executor/execIndexing.c
index 6ae0f959592..6d3e9d2f311 100644
--- a/src/backend/executor/execIndexing.c
+++ b/src/backend/executor/execIndexing.c
@@ -816,7 +816,7 @@ check_exclusion_or_unique_constraint(Relation heap, Relation index,
retry:
conflict = false;
found_self = false;
- index_scan = index_beginscan(heap, index, &DirtySnapshot, NULL, indnkeyatts, 0);
+ index_scan = index_beginscan(heap, index, &DirtySnapshot, NULL, indnkeyatts, 0, 0);
index_rescan(index_scan, scankeys, indnkeyatts, NULL, 0);
while (index_getnext_slot(index_scan, ForwardScanDirection, existing_slot))
diff --git a/src/backend/executor/execReplication.c b/src/backend/executor/execReplication.c
index 72f2bff7708..4c965a35e05 100644
--- a/src/backend/executor/execReplication.c
+++ b/src/backend/executor/execReplication.c
@@ -205,7 +205,7 @@ RelationFindReplTupleByIndex(Relation rel, Oid idxoid,
skey_attoff = build_replindex_scan_key(skey, rel, idxrel, searchslot);
/* Start an index scan. */
- scan = index_beginscan(rel, idxrel, &snap, NULL, skey_attoff, 0);
+ scan = index_beginscan(rel, idxrel, &snap, NULL, skey_attoff, 0, 0);
retry:
found = false;
@@ -383,7 +383,7 @@ RelationFindReplTupleSeq(Relation rel, LockTupleMode lockmode,
/* Start a heap scan. */
InitDirtySnapshot(snap);
- scan = table_beginscan(rel, &snap, 0, NULL);
+ scan = table_beginscan(rel, &snap, 0, NULL, 0);
scanslot = table_slot_create(rel, NULL);
retry:
@@ -602,7 +602,7 @@ RelationFindDeletedTupleInfoSeq(Relation rel, TupleTableSlot *searchslot,
* not yet committed or those just committed prior to the scan are
* excluded in update_most_recent_deletion_info().
*/
- scan = table_beginscan(rel, SnapshotAny, 0, NULL);
+ scan = table_beginscan(rel, SnapshotAny, 0, NULL, 0);
scanslot = table_slot_create(rel, NULL);
table_rescan(scan, NULL);
@@ -666,7 +666,7 @@ RelationFindDeletedTupleInfoByIndex(Relation rel, Oid idxoid,
* not yet committed or those just committed prior to the scan are
* excluded in update_most_recent_deletion_info().
*/
- scan = index_beginscan(rel, idxrel, SnapshotAny, NULL, skey_attoff, 0);
+ scan = index_beginscan(rel, idxrel, SnapshotAny, NULL, skey_attoff, 0, 0);
index_rescan(scan, skey, skey_attoff, NULL, 0);
diff --git a/src/backend/executor/nodeBitmapHeapscan.c b/src/backend/executor/nodeBitmapHeapscan.c
index c68c26cbf38..1017676fce0 100644
--- a/src/backend/executor/nodeBitmapHeapscan.c
+++ b/src/backend/executor/nodeBitmapHeapscan.c
@@ -103,11 +103,18 @@ BitmapTableScanSetup(BitmapHeapScanState *node)
*/
if (!node->ss.ss_currentScanDesc)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ node->ss.ps.state->es_modified_relids))
+ flags = SO_HINT_REL_READ_ONLY;
+
node->ss.ss_currentScanDesc =
table_beginscan_bm(node->ss.ss_currentRelation,
node->ss.ps.state->es_snapshot,
0,
- NULL);
+ NULL,
+ flags);
}
node->ss.ss_currentScanDesc->st.rs_tbmiterator = tbmiterator;
diff --git a/src/backend/executor/nodeIndexonlyscan.c b/src/backend/executor/nodeIndexonlyscan.c
index c2d09374517..2fe724a323f 100644
--- a/src/backend/executor/nodeIndexonlyscan.c
+++ b/src/backend/executor/nodeIndexonlyscan.c
@@ -84,6 +84,12 @@ IndexOnlyNext(IndexOnlyScanState *node)
if (scandesc == NULL)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags = SO_HINT_REL_READ_ONLY;
+
/*
* We reach here if the index only scan is not parallel, or if we're
* serially executing an index only scan that was planned to be
@@ -94,7 +100,8 @@ IndexOnlyNext(IndexOnlyScanState *node)
estate->es_snapshot,
&node->ioss_Instrument,
node->ioss_NumScanKeys,
- node->ioss_NumOrderByKeys);
+ node->ioss_NumOrderByKeys,
+ flags);
node->ioss_ScanDesc = scandesc;
diff --git a/src/backend/executor/nodeIndexscan.c b/src/backend/executor/nodeIndexscan.c
index a616abff04c..8730dab7469 100644
--- a/src/backend/executor/nodeIndexscan.c
+++ b/src/backend/executor/nodeIndexscan.c
@@ -102,6 +102,12 @@ IndexNext(IndexScanState *node)
if (scandesc == NULL)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags = SO_HINT_REL_READ_ONLY;
+
/*
* We reach here if the index scan is not parallel, or if we're
* serially executing an index scan that was planned to be parallel.
@@ -111,7 +117,8 @@ IndexNext(IndexScanState *node)
estate->es_snapshot,
&node->iss_Instrument,
node->iss_NumScanKeys,
- node->iss_NumOrderByKeys);
+ node->iss_NumOrderByKeys,
+ flags);
node->iss_ScanDesc = scandesc;
@@ -198,6 +205,12 @@ IndexNextWithReorder(IndexScanState *node)
if (scandesc == NULL)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags = SO_HINT_REL_READ_ONLY;
+
/*
* We reach here if the index scan is not parallel, or if we're
* serially executing an index scan that was planned to be parallel.
@@ -207,7 +220,8 @@ IndexNextWithReorder(IndexScanState *node)
estate->es_snapshot,
&node->iss_Instrument,
node->iss_NumScanKeys,
- node->iss_NumOrderByKeys);
+ node->iss_NumOrderByKeys,
+ flags);
node->iss_ScanDesc = scandesc;
diff --git a/src/backend/executor/nodeSeqscan.c b/src/backend/executor/nodeSeqscan.c
index af3c788ce8b..336354922a2 100644
--- a/src/backend/executor/nodeSeqscan.c
+++ b/src/backend/executor/nodeSeqscan.c
@@ -65,13 +65,20 @@ SeqNext(SeqScanState *node)
if (scandesc == NULL)
{
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags = SO_HINT_REL_READ_ONLY;
+
/*
* We reach here if the scan is not parallel, or if we're serially
* executing a scan that was planned to be parallel.
*/
scandesc = table_beginscan(node->ss.ss_currentRelation,
estate->es_snapshot,
- 0, NULL);
+ 0, NULL, flags);
+
node->ss.ss_currentScanDesc = scandesc;
}
@@ -367,14 +374,20 @@ ExecSeqScanInitializeDSM(SeqScanState *node,
{
EState *estate = node->ss.ps.state;
ParallelTableScanDesc pscan;
+ uint32 flags = 0;
pscan = shm_toc_allocate(pcxt->toc, node->pscan_len);
table_parallelscan_initialize(node->ss.ss_currentRelation,
pscan,
estate->es_snapshot);
shm_toc_insert(pcxt->toc, node->ss.ps.plan->plan_node_id, pscan);
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ estate->es_modified_relids))
+ flags = SO_HINT_REL_READ_ONLY;
+
node->ss.ss_currentScanDesc =
- table_beginscan_parallel(node->ss.ss_currentRelation, pscan);
+ table_beginscan_parallel(node->ss.ss_currentRelation, pscan,
+ flags);
}
/* ----------------------------------------------------------------
@@ -404,8 +417,15 @@ ExecSeqScanInitializeWorker(SeqScanState *node,
ParallelWorkerContext *pwcxt)
{
ParallelTableScanDesc pscan;
+ uint32 flags = 0;
+
+ if (!bms_is_member(((Scan *) node->ss.ps.plan)->scanrelid,
+ node->ss.ps.state->es_modified_relids))
+ flags = SO_HINT_REL_READ_ONLY;
pscan = shm_toc_lookup(pwcxt->toc, node->ss.ps.plan->plan_node_id, false);
node->ss.ss_currentScanDesc =
- table_beginscan_parallel(node->ss.ss_currentRelation, pscan);
+ table_beginscan_parallel(node->ss.ss_currentRelation,
+ pscan,
+ flags);
}
diff --git a/src/backend/partitioning/partbounds.c b/src/backend/partitioning/partbounds.c
index 0ca312ac27d..b7c4e6d1071 100644
--- a/src/backend/partitioning/partbounds.c
+++ b/src/backend/partitioning/partbounds.c
@@ -3362,7 +3362,7 @@ check_default_partition_contents(Relation parent, Relation default_rel,
econtext = GetPerTupleExprContext(estate);
snapshot = RegisterSnapshot(GetLatestSnapshot());
tupslot = table_slot_create(part_rel, &estate->es_tupleTable);
- scan = table_beginscan(part_rel, snapshot, 0, NULL);
+ scan = table_beginscan(part_rel, snapshot, 0, NULL, 0);
/*
* Switch to per-tuple memory context and reset it for each tuple
diff --git a/src/backend/utils/adt/selfuncs.c b/src/backend/utils/adt/selfuncs.c
index 29fec655593..ac181853225 100644
--- a/src/backend/utils/adt/selfuncs.c
+++ b/src/backend/utils/adt/selfuncs.c
@@ -7181,7 +7181,7 @@ get_actual_variable_endpoint(Relation heapRel,
index_scan = index_beginscan(heapRel, indexRel,
&SnapshotNonVacuumable, NULL,
- 1, 0);
+ 1, 0, 0);
/* Set it up for index-only scan */
index_scan->xs_want_itup = true;
index_rescan(index_scan, scankeys, 1, NULL, 0);
diff --git a/src/include/access/genam.h b/src/include/access/genam.h
index 4c0429cc613..3934fa44793 100644
--- a/src/include/access/genam.h
+++ b/src/include/access/genam.h
@@ -156,7 +156,7 @@ extern IndexScanDesc index_beginscan(Relation heapRelation,
Relation indexRelation,
Snapshot snapshot,
IndexScanInstrumentation *instrument,
- int nkeys, int norderbys);
+ int nkeys, int norderbys, uint32 flags);
extern IndexScanDesc index_beginscan_bitmap(Relation indexRelation,
Snapshot snapshot,
IndexScanInstrumentation *instrument,
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index 05ef8d8cd5e..6d54781609a 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -119,6 +119,12 @@ typedef struct IndexFetchHeapData
Buffer xs_cbuf; /* current heap buffer in scan, if any */
/* NB: if xs_cbuf is not InvalidBuffer, we hold a pin on that buffer */
+
+ /*
+ * Some optimizations can only be performed if the query does not modify
+ * the underlying relation. Track that here.
+ */
+ bool modifies_base_rel;
} IndexFetchHeapData;
/* Result codes for HeapTupleSatisfiesVacuum */
diff --git a/src/include/access/tableam.h b/src/include/access/tableam.h
index c9fa9f259cd..fb2a54b010d 100644
--- a/src/include/access/tableam.h
+++ b/src/include/access/tableam.h
@@ -63,6 +63,8 @@ typedef enum ScanOptions
/* unregister snapshot at scan end? */
SO_TEMP_SNAPSHOT = 1 << 9,
+ /* set if the query doesn't modify the relation */
+ SO_HINT_REL_READ_ONLY = 1 << 10,
} ScanOptions;
/*
@@ -420,7 +422,7 @@ typedef struct TableAmRoutine
*
* Tuples for an index scan can then be fetched via index_fetch_tuple.
*/
- struct IndexFetchTableData *(*index_fetch_begin) (Relation rel);
+ struct IndexFetchTableData *(*index_fetch_begin) (Relation rel, uint32 flags);
/*
* Reset index fetch. Typically this will release cross index fetch
@@ -873,9 +875,9 @@ extern TupleTableSlot *table_slot_create(Relation relation, List **reglist);
*/
static inline TableScanDesc
table_beginscan(Relation rel, Snapshot snapshot,
- int nkeys, ScanKeyData *key)
+ int nkeys, ScanKeyData *key, uint32 flags)
{
- uint32 flags = SO_TYPE_SEQSCAN |
+ flags |= SO_TYPE_SEQSCAN |
SO_ALLOW_STRAT | SO_ALLOW_SYNC | SO_ALLOW_PAGEMODE;
return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key, NULL, flags);
@@ -918,9 +920,9 @@ table_beginscan_strat(Relation rel, Snapshot snapshot,
*/
static inline TableScanDesc
table_beginscan_bm(Relation rel, Snapshot snapshot,
- int nkeys, ScanKeyData *key)
+ int nkeys, ScanKeyData *key, uint32 flags)
{
- uint32 flags = SO_TYPE_BITMAPSCAN | SO_ALLOW_PAGEMODE;
+ flags |= SO_TYPE_BITMAPSCAN | SO_ALLOW_PAGEMODE;
return rel->rd_tableam->scan_begin(rel, snapshot, nkeys, key,
NULL, flags);
@@ -1127,7 +1129,8 @@ extern void table_parallelscan_initialize(Relation rel,
* Caller must hold a suitable lock on the relation.
*/
extern TableScanDesc table_beginscan_parallel(Relation relation,
- ParallelTableScanDesc pscan);
+ ParallelTableScanDesc pscan,
+ uint32 flags);
/*
* Begin a parallel tid range scan. `pscan` needs to have been initialized
@@ -1163,9 +1166,9 @@ table_parallelscan_reinitialize(Relation rel, ParallelTableScanDesc pscan)
* Tuples for an index scan can then be fetched via table_index_fetch_tuple().
*/
static inline IndexFetchTableData *
-table_index_fetch_begin(Relation rel)
+table_index_fetch_begin(Relation rel, uint32 flags)
{
- return rel->rd_tableam->index_fetch_begin(rel);
+ return rel->rd_tableam->index_fetch_begin(rel, flags);
}
/*
--
2.43.0
[text/x-patch] v34-0013-Allow-on-access-pruning-to-set-pages-all-visible.patch (10.9K, 14-v34-0013-Allow-on-access-pruning-to-set-pages-all-visible.patch)
download | inline diff:
From cef0d7f13467597bdec9dacfb1d586b857c37137 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Wed, 3 Dec 2025 15:24:08 -0500
Subject: [PATCH v34 13/14] Allow on-access pruning to set pages all-visible
Many queries do not modify the underlying relation. For such queries, if
on-access pruning occurs during the scan, we can check whether the page
has become all-visible and update the visibility map accordingly.
Previously, only vacuum and COPY FREEZE marked pages as all-visible or
all-frozen.
This commit implements on-access VM setting for sequential scans as well
as for the underlying heap relation in index scans and bitmap heap
scans.
Author: Melanie Plageman <[email protected]>
Reviewed-by: Andres Freund <[email protected]>
Reviewed-by: Kirill Reshke <[email protected]>
Discussion: https://postgr.es/m/flat/CAAKRu_ZMw6Npd_qm2KM%2BFwQ3cMOMx1Dh3VMhp8-V7SOLxdK9-g%40mail.gmail.com
---
src/backend/access/heap/heapam.c | 15 +++++++-
src/backend/access/heap/heapam_handler.c | 15 +++++++-
src/backend/access/heap/pruneheap.c | 38 ++++++++++++++++++-
src/include/access/heapam.h | 24 ++++++++++--
.../t/035_standby_logical_decoding.pl | 3 +-
5 files changed, 87 insertions(+), 8 deletions(-)
diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index 1a3fa8a76aa..7d22549a290 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -617,6 +617,7 @@ heap_prepare_pagescan(TableScanDesc sscan)
Buffer buffer = scan->rs_cbuf;
BlockNumber block = scan->rs_cblock;
Snapshot snapshot;
+ Buffer *vmbuffer = NULL;
Page page;
int lines;
bool all_visible;
@@ -631,7 +632,9 @@ heap_prepare_pagescan(TableScanDesc sscan)
/*
* Prune and repair fragmentation for the whole page, if possible.
*/
- heap_page_prune_opt(scan->rs_base.rs_rd, buffer);
+ if (sscan->rs_flags & SO_HINT_REL_READ_ONLY)
+ vmbuffer = &scan->rs_vmbuffer;
+ heap_page_prune_opt(scan->rs_base.rs_rd, buffer, vmbuffer);
/*
* We must hold share lock on the buffer content while examining tuple
@@ -1308,6 +1311,7 @@ heap_beginscan(Relation relation, Snapshot snapshot,
sizeof(TBMIterateResult));
}
+ scan->rs_vmbuffer = InvalidBuffer;
return (TableScanDesc) scan;
}
@@ -1346,6 +1350,12 @@ heap_rescan(TableScanDesc sscan, ScanKey key, bool set_params,
scan->rs_cbuf = InvalidBuffer;
}
+ if (BufferIsValid(scan->rs_vmbuffer))
+ {
+ ReleaseBuffer(scan->rs_vmbuffer);
+ scan->rs_vmbuffer = InvalidBuffer;
+ }
+
/*
* SO_TYPE_BITMAPSCAN would be cleaned up here, but it does not hold any
* additional data vs a normal HeapScan
@@ -1378,6 +1388,9 @@ heap_endscan(TableScanDesc sscan)
if (BufferIsValid(scan->rs_cbuf))
ReleaseBuffer(scan->rs_cbuf);
+ if (BufferIsValid(scan->rs_vmbuffer))
+ ReleaseBuffer(scan->rs_vmbuffer);
+
/*
* Must free the read stream before freeing the BufferAccessStrategy.
*/
diff --git a/src/backend/access/heap/heapam_handler.c b/src/backend/access/heap/heapam_handler.c
index 6ce4350c2c8..c9e67514aea 100644
--- a/src/backend/access/heap/heapam_handler.c
+++ b/src/backend/access/heap/heapam_handler.c
@@ -85,6 +85,7 @@ heapam_index_fetch_begin(Relation rel, uint32 flags)
hscan->xs_base.rel = rel;
hscan->xs_cbuf = InvalidBuffer;
+ hscan->xs_vmbuffer = InvalidBuffer;
hscan->modifies_base_rel = !(flags & SO_HINT_REL_READ_ONLY);
return &hscan->xs_base;
@@ -100,6 +101,12 @@ heapam_index_fetch_reset(IndexFetchTableData *scan)
ReleaseBuffer(hscan->xs_cbuf);
hscan->xs_cbuf = InvalidBuffer;
}
+
+ if (BufferIsValid(hscan->xs_vmbuffer))
+ {
+ ReleaseBuffer(hscan->xs_vmbuffer);
+ hscan->xs_vmbuffer = InvalidBuffer;
+ }
}
static void
@@ -139,7 +146,8 @@ heapam_index_fetch_tuple(struct IndexFetchTableData *scan,
* Prune page, but only if we weren't already on this page
*/
if (prev_buf != hscan->xs_cbuf)
- heap_page_prune_opt(hscan->xs_base.rel, hscan->xs_cbuf);
+ heap_page_prune_opt(hscan->xs_base.rel, hscan->xs_cbuf,
+ hscan->modifies_base_rel ? NULL : &hscan->xs_vmbuffer);
}
/* Obtain share-lock on the buffer so we can examine visibility */
@@ -2488,6 +2496,7 @@ BitmapHeapScanNextBlock(TableScanDesc scan,
TBMIterateResult *tbmres;
OffsetNumber offsets[TBM_MAX_TUPLES_PER_PAGE];
int noffsets = -1;
+ Buffer *vmbuffer = NULL;
Assert(scan->rs_flags & SO_TYPE_BITMAPSCAN);
Assert(hscan->rs_read_stream);
@@ -2534,7 +2543,9 @@ BitmapHeapScanNextBlock(TableScanDesc scan,
/*
* Prune and repair fragmentation for the whole page, if possible.
*/
- heap_page_prune_opt(scan->rs_rd, buffer);
+ if (scan->rs_flags & SO_HINT_REL_READ_ONLY)
+ vmbuffer = &hscan->rs_vmbuffer;
+ heap_page_prune_opt(scan->rs_rd, buffer, vmbuffer);
/*
* We must hold share lock on the buffer content while examining tuple
diff --git a/src/backend/access/heap/pruneheap.c b/src/backend/access/heap/pruneheap.c
index cc45728d25e..59f647b7f77 100644
--- a/src/backend/access/heap/pruneheap.c
+++ b/src/backend/access/heap/pruneheap.c
@@ -202,6 +202,8 @@ static bool heap_page_will_set_vm(PruneState *prstate,
Relation relation,
BlockNumber heap_blk, Buffer heap_buffer, Page heap_page,
Buffer vmbuffer,
+ PruneReason reason,
+ bool do_prune, bool do_freeze,
int nlpdead_items,
uint8 *old_vmbits,
uint8 *new_vmbits);
@@ -223,9 +225,13 @@ static TransactionId get_conflict_xid(bool do_prune, bool do_freeze, bool do_set
* if there's not any use in pruning.
*
* Caller must have pin on the buffer, and must *not* have a lock on it.
+ *
+ * If vmbuffer is not NULL, it is okay for pruning to set the visibility map if
+ * the page is all-visible. We will take care of pinning and, if needed,
+ * reading in the page of the visibility map.
*/
void
-heap_page_prune_opt(Relation relation, Buffer buffer)
+heap_page_prune_opt(Relation relation, Buffer buffer, Buffer *vmbuffer)
{
Page page = BufferGetPage(buffer);
TransactionId prune_xid;
@@ -306,6 +312,13 @@ heap_page_prune_opt(Relation relation, Buffer buffer)
.cutoffs = NULL,
};
+ if (vmbuffer)
+ {
+ visibilitymap_pin(relation, BufferGetBlockNumber(buffer), vmbuffer);
+ params.options |= HEAP_PAGE_PRUNE_UPDATE_VM;
+ params.vmbuffer = *vmbuffer;
+ }
+
heap_page_prune_and_freeze(¶ms, &presult, &dummy_off_loc,
NULL, NULL);
@@ -952,6 +965,9 @@ identify_and_fix_vm_corruption(Relation rel, Buffer heap_buffer,
* corrupted, it will fix them by clearing the VM bits and visibility hint.
* This does not need to be done in a critical section.
*
+ * This should be called only after do_freeze has been decided (and do_prune
+ * has been set), as these factor into our heuristic-based decision.
+ *
* Returns true if one or both VM bits should be set, along with returning the
* current value of the VM bits in *old_vmbits and the desired new value of
* the VM bits in *new_vmbits.
@@ -965,6 +981,8 @@ heap_page_will_set_vm(PruneState *prstate,
Relation relation,
BlockNumber heap_blk, Buffer heap_buffer, Page heap_page,
Buffer vmbuffer,
+ PruneReason reason,
+ bool do_prune, bool do_freeze,
int nlpdead_items,
uint8 *old_vmbits,
uint8 *new_vmbits)
@@ -986,6 +1004,22 @@ heap_page_will_set_vm(PruneState *prstate,
if (!prstate->all_visible)
return false;
+ /*
+ * If this is an on-access call and we're not actually pruning, avoid
+ * setting the visibility map if it would newly dirty the heap page or, if
+ * the page is already dirty, if doing so would require including a
+ * full-page image (FPI) of the heap page in the WAL. This situation
+ * should be rare, as on-access pruning is only attempted when
+ * pd_prune_xid is valid.
+ */
+ if (reason == PRUNE_ON_ACCESS && !do_prune && !do_freeze &&
+ (!BufferIsDirty(heap_buffer) || XLogCheckBufferNeedsBackup(heap_buffer)))
+ {
+ prstate->all_visible = false;
+ prstate->all_frozen = false;
+ return false;
+ }
+
*new_vmbits = VISIBILITYMAP_ALL_VISIBLE;
if (prstate->all_frozen)
@@ -1155,6 +1189,8 @@ heap_page_prune_and_freeze(PruneFreezeParams *params,
buffer,
page,
vmbuffer,
+ params->reason,
+ do_prune, do_freeze,
prstate.lpdead_items,
&old_vmbits,
&new_vmbits);
diff --git a/src/include/access/heapam.h b/src/include/access/heapam.h
index 6d54781609a..55ac478dd67 100644
--- a/src/include/access/heapam.h
+++ b/src/include/access/heapam.h
@@ -95,6 +95,13 @@ typedef struct HeapScanDescData
*/
ParallelBlockTableScanWorkerData *rs_parallelworkerdata;
+ /*
+ * For sequential scans and bitmap heap scans. If the relation is not
+ * being modified, on-access pruning may read in the current heap page's
+ * corresponding VM block to this buffer.
+ */
+ Buffer rs_vmbuffer;
+
/* these fields only used in page-at-a-time mode and for bitmap scans */
uint32 rs_cindex; /* current tuple's index in vistuples */
uint32 rs_ntuples; /* number of visible tuples on page */
@@ -117,8 +124,18 @@ typedef struct IndexFetchHeapData
{
IndexFetchTableData xs_base; /* AM independent part of the descriptor */
- Buffer xs_cbuf; /* current heap buffer in scan, if any */
- /* NB: if xs_cbuf is not InvalidBuffer, we hold a pin on that buffer */
+ /*
+ * Current heap buffer in scan, if any. NB: if xs_cbuf is not
+ * InvalidBuffer, we hold a pin on that buffer.
+ */
+ Buffer xs_cbuf;
+
+ /*
+ * For index scans that do not modify the underlying heap table, on-access
+ * pruning may read in the current heap page's corresponding VM block to
+ * this buffer.
+ */
+ Buffer xs_vmbuffer;
/*
* Some optimizations can only be performed if the query does not modify
@@ -419,7 +436,8 @@ extern TransactionId heap_index_delete_tuples(Relation rel,
TM_IndexDeleteOp *delstate);
/* in heap/pruneheap.c */
-extern void heap_page_prune_opt(Relation relation, Buffer buffer);
+extern void heap_page_prune_opt(Relation relation, Buffer buffer,
+ Buffer *vmbuffer);
extern void heap_page_prune_and_freeze(PruneFreezeParams *params,
PruneFreezeResult *presult,
OffsetNumber *off_loc,
diff --git a/src/test/recovery/t/035_standby_logical_decoding.pl b/src/test/recovery/t/035_standby_logical_decoding.pl
index d264a698ff6..a5536ba4ff6 100644
--- a/src/test/recovery/t/035_standby_logical_decoding.pl
+++ b/src/test/recovery/t/035_standby_logical_decoding.pl
@@ -296,6 +296,7 @@ wal_level = 'logical'
max_replication_slots = 4
max_wal_senders = 4
autovacuum = off
+hot_standby_feedback = on
});
$node_primary->dump_info;
$node_primary->start;
@@ -748,7 +749,7 @@ check_pg_recvlogical_stderr($handle,
$logstart = -s $node_standby->logfile;
reactive_slots_change_hfs_and_wait_for_xmins('shared_row_removal_',
- 'no_conflict_', 0, 1);
+ 'no_conflict_', 1, 0);
# This should not trigger a conflict
wait_until_vacuum_can_remove(
--
2.43.0
[text/x-patch] v34-0014-Set-pd_prune_xid-on-insert.patch (6.8K, 15-v34-0014-Set-pd_prune_xid-on-insert.patch)
download | inline diff:
From ba527a1cf0fff9f2b80d64d9d0f80888c6e5db66 Mon Sep 17 00:00:00 2001
From: Melanie Plageman <[email protected]>
Date: Tue, 29 Jul 2025 16:12:56 -0400
Subject: [PATCH v34 14/14] Set pd_prune_xid on insert
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit
Now that visibility map (VM) updates can occur during read-only queries,
it makes sense to also set the page’s pd_prune_xid hint during inserts.
This enables heap_page_prune_and_freeze() to run after a page is
filled with newly inserted tuples the first time it is read.
This change also addresses a long-standing note in heap_insert() and
heap_multi_insert(), which observed that setting pd_prune_xid would
help clean up aborted insertions sooner. Without it, such tuples might
linger until VACUUM, whereas now they can be pruned earlier.
The index killtuples test had to be updated to reflect a larger number
of hits by some accesses. Since the prune_xid is set by the fill/insert
step, on-access pruning can happen during the first access step (before
the DELETE). This is when the VM is extended. After the DELETE, the next
access hits the VM block instead of extending it. Thus, an additional
buffer hit is counted for the table.
Reviewed-by: Chao Li <[email protected]>
---
src/backend/access/heap/heapam.c | 25 +++++++++++++------
src/backend/access/heap/heapam_xlog.c | 15 ++++++++++-
.../modules/index/expected/killtuples.out | 6 ++---
3 files changed, 34 insertions(+), 12 deletions(-)
diff --git a/src/backend/access/heap/heapam.c b/src/backend/access/heap/heapam.c
index 7d22549a290..afafc7e2cb2 100644
--- a/src/backend/access/heap/heapam.c
+++ b/src/backend/access/heap/heapam.c
@@ -2166,6 +2166,7 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
TransactionId xid = GetCurrentTransactionId();
HeapTuple heaptup;
Buffer buffer;
+ Page page;
Buffer vmbuffer = InvalidBuffer;
bool all_visible_cleared = false;
@@ -2225,15 +2226,19 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
}
/*
- * XXX Should we set PageSetPrunable on this page ?
+ * Set pd_prune_xid to trigger heap_page_prune_and_freeze() once the page
+ * is full so that we can set the page all-visible in the VM.
*
- * The inserting transaction may eventually abort thus making this tuple
- * DEAD and hence available for pruning. Though we don't want to optimize
- * for aborts, if no other tuple in this page is UPDATEd/DELETEd, the
- * aborted tuple will never be pruned until next vacuum is triggered.
+ * Setting pd_prune_xid is also handy if the inserting transaction
+ * eventually aborts making this tuple DEAD and hence available for
+ * pruning. If no other tuple in this page is UPDATEd/DELETEd, the aborted
+ * tuple would never otherwise be pruned until next vacuum is triggered.
*
- * If you do add PageSetPrunable here, add it in heap_xlog_insert too.
+ * Don't set it if we are in bootstrap mode, though.
*/
+ page = BufferGetPage(buffer);
+ if (TransactionIdIsNormal(xid))
+ PageSetPrunable(page, xid);
MarkBufferDirty(buffer);
@@ -2243,7 +2248,6 @@ heap_insert(Relation relation, HeapTuple tup, CommandId cid,
xl_heap_insert xlrec;
xl_heap_header xlhdr;
XLogRecPtr recptr;
- Page page = BufferGetPage(buffer);
uint8 info = XLOG_HEAP_INSERT;
int bufflags = 0;
@@ -2607,8 +2611,13 @@ heap_multi_insert(Relation relation, TupleTableSlot **slots, int ntuples,
}
/*
- * XXX Should we set PageSetPrunable on this page ? See heap_insert()
+ * Set pd_prune_xid. See heap_insert() for more on why we do this when
+ * inserting tuples. This only makes sense if we aren't already
+ * setting the page frozen in the VM. We also don't set it in
+ * bootstrap mode.
*/
+ if (!all_frozen_set && TransactionIdIsNormal(xid))
+ PageSetPrunable(page, xid);
MarkBufferDirty(buffer);
diff --git a/src/backend/access/heap/heapam_xlog.c b/src/backend/access/heap/heapam_xlog.c
index 9a29fda3601..49cc83a6479 100644
--- a/src/backend/access/heap/heapam_xlog.c
+++ b/src/backend/access/heap/heapam_xlog.c
@@ -447,6 +447,12 @@ heap_xlog_insert(XLogReaderState *record)
freespace = PageGetHeapFreeSpace(page); /* needed to update FSM below */
+ /*
+ * Set the page prunable to trigger on-access pruning later which may
+ * set the page all-visible in the VM.
+ */
+ PageSetPrunable(page, XLogRecGetXid(record));
+
PageSetLSN(page, lsn);
if (xlrec->flags & XLH_INSERT_ALL_VISIBLE_CLEARED)
@@ -596,9 +602,16 @@ heap_xlog_multi_insert(XLogReaderState *record)
if (xlrec->flags & XLH_INSERT_ALL_VISIBLE_CLEARED)
PageClearAllVisible(page);
- /* XLH_INSERT_ALL_FROZEN_SET implies that all tuples are visible */
+ /*
+ * XLH_INSERT_ALL_FROZEN_SET implies that all tuples are visible. If
+ * we are not setting the page frozen, then set the page's prunable
+ * hint so that we trigger on-access pruning later which may set the
+ * page all-visible in the VM.
+ */
if (xlrec->flags & XLH_INSERT_ALL_FROZEN_SET)
PageSetAllVisible(page);
+ else
+ PageSetPrunable(page, XLogRecGetXid(record));
MarkBufferDirty(buffer);
}
diff --git a/src/test/modules/index/expected/killtuples.out b/src/test/modules/index/expected/killtuples.out
index be7ddd756ef..b29f2434b00 100644
--- a/src/test/modules/index/expected/killtuples.out
+++ b/src/test/modules/index/expected/killtuples.out
@@ -54,7 +54,7 @@ step flush: SELECT FROM pg_stat_force_next_flush();
step result: SELECT heap_blks_read + heap_blks_hit - counter.heap_accesses AS new_heap_accesses FROM counter, pg_statio_all_tables WHERE relname = 'kill_prior_tuple';
new_heap_accesses
-----------------
- 1
+ 2
(1 row)
step measure: UPDATE counter SET heap_accesses = (SELECT heap_blks_read + heap_blks_hit FROM pg_statio_all_tables WHERE relname = 'kill_prior_tuple');
@@ -130,7 +130,7 @@ step flush: SELECT FROM pg_stat_force_next_flush();
step result: SELECT heap_blks_read + heap_blks_hit - counter.heap_accesses AS new_heap_accesses FROM counter, pg_statio_all_tables WHERE relname = 'kill_prior_tuple';
new_heap_accesses
-----------------
- 1
+ 2
(1 row)
step measure: UPDATE counter SET heap_accesses = (SELECT heap_blks_read + heap_blks_hit FROM pg_statio_all_tables WHERE relname = 'kill_prior_tuple');
@@ -283,7 +283,7 @@ step flush: SELECT FROM pg_stat_force_next_flush();
step result: SELECT heap_blks_read + heap_blks_hit - counter.heap_accesses AS new_heap_accesses FROM counter, pg_statio_all_tables WHERE relname = 'kill_prior_tuple';
new_heap_accesses
-----------------
- 1
+ 2
(1 row)
step measure: UPDATE counter SET heap_accesses = (SELECT heap_blks_read + heap_blks_hit FROM pg_statio_all_tables WHERE relname = 'kill_prior_tuple');
--
2.43.0
view thread (17+ messages) latest in thread
reply
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Reply to all the recipients using the --to and --cc options:
reply via email
To: [email protected]
Cc: [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected], [email protected]
Subject: Re: eliminate xl_heap_visible to reduce WAL (and eventually set VM on-access)
In-Reply-To: <CAAKRu_bs+gZ83QDacmBxunPvCGnXJ05hxP2BDPJ3BGwdbGRXzg@mail.gmail.com>
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
This inbox is served by agora; see mirroring instructions
for how to clone and mirror all data and code used for this inbox