All of lore.kernel.org
 help / color / mirror / Atom feed
From: Dave Chinner <david@fromorbit.com>
To: Amir Goldstein <amir73il@gmail.com>
Cc: Allison Henderson <allison.henderson@oracle.com>,
	linux-xfs <linux-xfs@vger.kernel.org>
Subject: Re: [PATCH 00/17] Parent Pointers V3
Date: Mon, 23 Oct 2017 19:40:33 +1100	[thread overview]
Message-ID: <20171023084032.GW3666@dastard> (raw)
In-Reply-To: <CAOQ4uxgH8oza283kL=cUvE7a0SA5wPsD1gyVApBqd2hWHE8y3w@mail.gmail.com>

On Mon, Oct 23, 2017 at 09:48:24AM +0300, Amir Goldstein wrote:
> On Mon, Oct 23, 2017 at 8:32 AM, Dave Chinner <david@fromorbit.com> wrote:
> > On Mon, Oct 23, 2017 at 07:30:01AM +0300, Amir Goldstein wrote:
> >> If there is no such reason I would like to propose to store parent
> >> pointers of directory inodes with an xattr name that is diroffset agnostic,
> >> namely:
> >>
> >> struct xfs_dir_parent_name_rec {
> >>       __be64  p_ino;
> >>       __be32  p_gen;
> >> } __attribute__((packed));
> >>
> >> It makes sense for metadata consistency anyway, but the reason I am
> >> proposing this is so xfs_fs_get_name() could be implemented at O(1).

[...]

> But you missed my question completely. I understand why diroffset
> is needed for non-dir. But directories are not allowed to have 2 differnt
> parents nor 2 entries in the parent dir, so I was asking if we can
> special case parent pointer stored in *directory inodes*.

I did answer it - not directly, but the answers are there.

> Am I missing something?

Because you are just looking at it from a "reverse lookup"
perspective, I suspect you're not seeing the "detect and
reconstruct" broken directory structure side of the picture.

The directory offset is redundant information which needed to fully
cross-check and, if necessary, reconstruct broken direcotry
structures. E.g. if a directory is missing a block due to a bad
sector, we can reconstruct that block exactly from the parent
pointer information because we know exactly what inodes had dirents
in the block that was lost, and we know exactly what order they
appeared in the directory block...

If we don't have that info for all child inodes (directory or
regular file), we can't tell the difference between "lost/stale
child that we should ignore" and "child that was referenced by the
block we lost". So even directories need to have the diroffset in
their parent pointer....

Cheers,

Dave.
-- 
Dave Chinner
david@fromorbit.com

  reply	other threads:[~2017-10-23  8:40 UTC|newest]

Thread overview: 61+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2017-10-18 22:55 [PATCH 00/17] Parent Pointers V3 Allison Henderson
2017-10-18 22:55 ` [PATCH 01/17] Add helper functions xfs_attr_set_args and xfs_attr_remove_args Allison Henderson
2017-10-19 20:03   ` Darrick J. Wong
2017-10-21  1:14     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 02/17] Set up infastructure for deferred attribute operations Allison Henderson
2017-10-19 19:02   ` Darrick J. Wong
2017-10-21  1:08     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 03/17] Add xfs_attr_set_defered and xfs_attr_remove_defered Allison Henderson
2017-10-19 19:13   ` Darrick J. Wong
2017-10-21  1:08     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 04/17] Remove all strlen calls in all xfs_attr_* functions for attr names Allison Henderson
2017-10-19 19:15   ` Darrick J. Wong
2017-10-21  1:10     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 05/17] xfs: get directory offset when adding directory name Allison Henderson
2017-10-18 22:55 ` [PATCH 06/17] xfs: get directory offset when removing " Allison Henderson
2017-10-19 19:17   ` Darrick J. Wong
2017-10-21  1:11     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 07/17] xfs: get directory offset when replacing a " Allison Henderson
2017-10-18 22:55 ` [PATCH 08/17] xfs: add parent pointer support to attribute code Allison Henderson
2017-10-18 22:55 ` [PATCH 09/17] xfs: define parent pointer xattr format Allison Henderson
2017-10-18 22:55 ` [PATCH 10/17] :xfs: extent transaction reservations for parent attributes Allison Henderson
2017-10-19 18:24   ` Darrick J. Wong
     [not found]     ` <8680e0c1-ada8-06e3-e397-61a5076030be@oracle.com>
2017-10-20 23:45       ` Darrick J. Wong
2017-10-21  0:12         ` Allison Henderson
2017-10-21  1:07     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 11/17] Add the extra space requirements for parent pointer attributes when calculating the minimum log size during mkfs Allison Henderson
2017-10-19 18:13   ` Darrick J. Wong
2017-10-21  1:07     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 12/17] xfs: parent pointer attribute creation Allison Henderson
2017-10-19 19:36   ` Darrick J. Wong
     [not found]     ` <9185d3e8-4b41-b2d8-294b-934f7d3409f0@oracle.com>
2017-10-21  0:03       ` Darrick J. Wong
2017-10-21  1:11     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 13/17] xfs: add parent attributes to link Allison Henderson
2017-10-19 19:40   ` Darrick J. Wong
2017-10-21  1:12     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 14/17] xfs: remove parent pointers in unlink Allison Henderson
2017-10-19 19:43   ` Darrick J. Wong
2017-10-21  1:12     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 15/17] xfs_bmap_add_attrfork(): re-add error handling from set_attrforkoff() call Allison Henderson
2017-10-19 19:43   ` Darrick J. Wong
2017-10-21  1:13     ` Allison Henderson
2017-10-18 22:55 ` [PATCH 16/17] Add parent pointers to rename Allison Henderson
2017-10-18 22:55 ` [PATCH 17/17] Add the parent pointer support to the superblock version 5 Allison Henderson
2017-10-19  3:57   ` Amir Goldstein
2017-10-19 20:06     ` Darrick J. Wong
2017-10-20  3:18       ` Amir Goldstein
2017-10-19 19:45   ` Darrick J. Wong
2017-10-21  1:13     ` Allison Henderson
2017-10-19  4:11 ` [PATCH 00/17] Parent Pointers V3 Amir Goldstein
2017-10-20  3:22   ` Amir Goldstein
2017-10-21  1:06     ` Allison Henderson
2017-10-20 22:41   ` Dave Chinner
2017-10-21  7:34     ` Amir Goldstein
2017-10-22 23:27       ` Dave Chinner
2017-10-23  4:30         ` Amir Goldstein
2017-10-23  5:32           ` Dave Chinner
2017-10-23  6:48             ` Amir Goldstein
2017-10-23  8:40               ` Dave Chinner [this message]
2017-10-23  9:06                 ` Amir Goldstein
2017-10-23 17:14                   ` Darrick J. Wong
2017-10-23 19:20                     ` Amir Goldstein

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=20171023084032.GW3666@dastard \
    --to=david@fromorbit.com \
    --cc=allison.henderson@oracle.com \
    --cc=amir73il@gmail.com \
    --cc=linux-xfs@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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.