On Fri, Mar 24, 2017 at 9:46 AM, Ashutosh Sharma <ashu.coe...@gmail.com> wrote:
> On Fri, Mar 24, 2017 at 9:21 AM, Amit Kapila <amit.kapil...@gmail.com> wrote:
>> On Thu, Mar 23, 2017 at 6:46 PM, Ashutosh Sharma <ashu.coe...@gmail.com>
>> wrote:
>>>>
>>>> Oh, okay, but my main objection was that we should not check hash page
>>>> type (hasho_flag) without ensuring whether it is a hash page. Can you
>>>> try to adjust the above code so that this check can be moved after
>>>> hasho_page_id check?
>>>
>>> Yes, I got your point. I have done that but then i had to remove the
>>> check for PageIsEmpty(). Anyways, I think PageIsEmpty() condition will
>>> only be true for one page in entire hash index table and can be
>>> ignored. If you wish, I could mention about it in the documentation.
>>>
>>
>> Yeah, I think it is worth adding a Note in docs, but we can do that
>> separately if required.
>>
>>>>
>>>>> To avoid
>>>>> this, at the start of verify_hash_page function itself if we recognise
>>>>> page as UNUSED page we return immediately.
>>>>>
>>>>>>
>>>>>> 2.
>>>>>> + /* Check if it is an empty hash page. */
>>>>>> + if (PageIsEmpty(page))
>>>>>> + ereport(ERROR,
>>>>>> + (errcode(ERRCODE_INDEX_CORRUPTED),
>>>>>> + errmsg("index table contains empty page")));
>>>>>>
>>>>>>
>>>>>> Do we want to give a separate message for EMPTY and NEW pages? Isn't
>>>>>> it better that the same error message can be given for both of them as
>>>>>> from user perspective there is not much difference between both the
>>>>>> messages?
>>>>>
>>>>> I think we should show separate message because they are two different
>>>>> type of pages. In the sense like, one is initialised whereas other is
>>>>> completely zero.
>>>>>
>>>>
>>>> I understand your point, but not sure if it makes any difference to user.
>>>>
>>>
>>> okay, I have now anyways removed the check for PageIsEmpty. Please
>>> refer to the attached '0002
>>> allow_pageinspect_handle_UNUSED_hash_pages.patch'
>>>
>>
>> +
>> if (pageopaque->hasho_page_id != HASHO_PAGE_ID)
>> ereport(ERROR,
>>
>> spurious white space.
>>
>>>
>>> Also, I have attached
>>> '0001-Mark-freed-overflow-page-as-UNUSED-pagetype-v2.patch' that
>>> handles your comment mentioned in [1].
>>>
>>
>> In general, we have to initialize prevblock with max_bucket, but here
>> it is okay, because we anyway initialize it after this page is
>> allocated as overflow page.
>>
>> Your patches look good to me except for small white space change.
>
> Thanks for reviewing my patch. I have removed the extra white space.
> Attached are both the patches.
Sorry, I have mistakenly attached wrong patch. Here are the correct
set of patches.
--
With Regards,
Ashutosh Sharma
EnterpriseDB:http://www.enterprisedb.com
From 86eb711ea80b495c9d2f5228558682d0e95c41eb Mon Sep 17 00:00:00 2001
From: ashu <ashu@localhost.localdomain>
Date: Thu, 23 Mar 2017 16:47:17 +0530
Subject: [PATCH] Mark freed overflow page as UNUSED pagetype v2
---
src/backend/access/hash/hash_xlog.c | 9 +++++++++
src/backend/access/hash/hashovfl.c | 13 ++++++++++++-
2 files changed, 21 insertions(+), 1 deletion(-)
diff --git a/src/backend/access/hash/hash_xlog.c b/src/backend/access/hash/hash_xlog.c
index d9ac42c..2ccaf46 100644
--- a/src/backend/access/hash/hash_xlog.c
+++ b/src/backend/access/hash/hash_xlog.c
@@ -697,11 +697,20 @@ hash_xlog_squeeze_page(XLogReaderState *record)
if (XLogReadBufferForRedo(record, 2, &ovflbuf) == BLK_NEEDS_REDO)
{
Page ovflpage;
+ HashPageOpaque ovflopaque;
ovflpage = BufferGetPage(ovflbuf);
_hash_pageinit(ovflpage, BufferGetPageSize(ovflbuf));
+ ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+
+ ovflopaque->hasho_prevblkno = InvalidBlockNumber;
+ ovflopaque->hasho_nextblkno = InvalidBlockNumber;
+ ovflopaque->hasho_bucket = -1;
+ ovflopaque->hasho_flag = LH_UNUSED_PAGE;
+ ovflopaque->hasho_page_id = HASHO_PAGE_ID;
+
PageSetLSN(ovflpage, lsn);
MarkBufferDirty(ovflbuf);
}
diff --git a/src/backend/access/hash/hashovfl.c b/src/backend/access/hash/hashovfl.c
index a3cae21..d3eaee0 100644
--- a/src/backend/access/hash/hashovfl.c
+++ b/src/backend/access/hash/hashovfl.c
@@ -591,9 +591,20 @@ _hash_freeovflpage(Relation rel, Buffer bucketbuf, Buffer ovflbuf,
/*
* Initialize the freed overflow page. Just zeroing the page won't work,
* because WAL replay routines expect pages to be initialized. See
- * explanation of RBM_NORMAL mode atop XLogReadBufferExtended.
+ * explanation of RBM_NORMAL mode atop XLogReadBufferExtended. Also, mark
+ * the page as UNUSED type and retain it's page id. This allows the tools
+ * like pageinspect to consider it as a hash page.
*/
_hash_pageinit(ovflpage, BufferGetPageSize(ovflbuf));
+
+ ovflopaque = (HashPageOpaque) PageGetSpecialPointer(ovflpage);
+
+ ovflopaque->hasho_prevblkno = InvalidBlockNumber;
+ ovflopaque->hasho_nextblkno = InvalidBlockNumber;
+ ovflopaque->hasho_bucket = -1;
+ ovflopaque->hasho_flag = LH_UNUSED_PAGE;
+ ovflopaque->hasho_page_id = HASHO_PAGE_ID;
+
MarkBufferDirty(ovflbuf);
if (BufferIsValid(prevbuf))
--
1.8.3.1
diff --git a/contrib/pageinspect/hashfuncs.c b/contrib/pageinspect/hashfuncs.c
index 812a03f..bc65c4b 100644
--- a/contrib/pageinspect/hashfuncs.c
+++ b/contrib/pageinspect/hashfuncs.c
@@ -77,6 +77,10 @@ verify_hash_page(bytea *raw_page, int flags)
errdetail("Expected %08x, got %08x.",
HASHO_PAGE_ID, pageopaque->hasho_page_id)));
+ /* Check if it is an unused hash page. */
+ if (pageopaque->hasho_flag == LH_UNUSED_PAGE)
+ return page;
+
/* Check that page type is sane. */
pagetype = pageopaque->hasho_flag & LH_PAGE_TYPE;
if (pagetype != LH_OVERFLOW_PAGE && pagetype != LH_BUCKET_PAGE &&
--
Sent via pgsql-hackers mailing list (pgsql-hackers@postgresql.org)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-hackers
