From: Nikolay Borisov <nborisov@suse.com>
To: linux-btrfs@vger.kernel.org, David Sterba <dsterba@suse.com>
Subject: Re: [PATCH] btrfs: Use transid for DIR_ITEM/DIR_INDEX's location
Date: Mon, 2 Nov 2020 16:17:37 +0200 [thread overview]
Message-ID: <cc07a185-b00a-c1cd-fe69-49b3991e1609@suse.com> (raw)
In-Reply-To: <20200828132010.27886-1-nborisov@suse.com>
On 28.08.20 г. 16:20 ч., Nikolay Borisov wrote:
> When a snapshot is created its root item is inserted in the root tree
> with the 'offset' field set to the transaction id when the snapshot
> was created. Immediately afterwards the offset is set to -1 so when
> the same key is used to create DIR_ITEM/DIR_INDEX in the destination
> file tree its offset is really set to -1. Root tree item:
>
> item 13 key (258 ROOT_ITEM 7) itemoff 12744 itemsize 439
> generation 7 root_dirid 256 bytenr 30703616 level 0 refs 1
> lastsnap 7 byte_limit 0 bytes_used 16384 flags 0x0(none)
> uuid f13abf0d-b1f5-f34b-a179-fd1c2f89e762
> parent_uuid 51a74677-a077-4c21-bd87-2141a147ff85
> ctransid 7 otransid 7 stransid 0 rtransid 0
> ctime 1598441149.466822752 (2020-08-26 11:25:49)
> otime 1598441149.467474846 (2020-08-26 11:25:49)
> drop key (0 UNKNOWN.0 0) level 0
>
> DIR_INDEX item for the same rooti in the destination fs tree:
>
> item 5 key (256 DIR_INDEX 9) itemoff 15967 itemsize 39
> location key (258 ROOT_ITEM 18446744073709551615) type DIR
> transid 7 data_len 0 name_len 9
> name: snapshot1
>
> The location key is generally used to read the root. This is not a
> problem per-se since the function dealing with root searching
> (btrfs_find_root) is well equipped to deal with offset being -1, namely:
>
> If ->offset of 'search_key' is -1ULL, it means we are not sure the
> offset of the search key, just lookup the root with the highest
> offset for a given objectid.
>
> However this is a needless inconcistency in the way internal data
> structures are being created. This patch modifies the behavior so that
> DIR_INDEX/DIR_ITEM will have the offset field of the location key set
> to the transid. While this results in a change of the on-disk metadata,
> it doesn't constitute a functional change since older kernels can cope
> with both '-1' and transid as values of the offset field so no INCOMPAT
> flags are needed.
>
> Finally while at it also move the initialization of the key in
> create_pending_snapshot closer to where it's being used for the first
> time.
>
> Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Ping on that, it's been reviewed by Josef but seems to have been missed.
next prev parent reply other threads:[~2020-11-02 14:17 UTC|newest]
Thread overview: 5+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-08-28 13:20 [PATCH] btrfs: Use transid for DIR_ITEM/DIR_INDEX's location Nikolay Borisov
2020-08-31 15:39 ` Josef Bacik
2020-11-02 14:17 ` Nikolay Borisov [this message]
2020-12-18 8:58 ` Nikolay Borisov
2021-02-25 15:52 ` David Sterba
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=cc07a185-b00a-c1cd-fe69-49b3991e1609@suse.com \
--to=nborisov@suse.com \
--cc=dsterba@suse.com \
--cc=linux-btrfs@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox