archive mirror
 help / color / mirror / Atom feed
From: Miklos Szeredi <>
To: Amir Goldstein <>
Cc: Al Viro <>,
	overlayfs <>,
	linux-fsdevel <>
Subject: Re: ovl: set I_CREATING on inode being created
Date: Wed, 22 Aug 2018 10:41:21 +0200	[thread overview]
Message-ID: <> (raw)
In-Reply-To: <>

On Fri, Aug 17, 2018 at 1:16 PM, Amir Goldstein <> wrote:
> Miklos,
> I have 2 questions/comments w.r.t. commit 9475938ce8cf
> ("ovl: set I_CREATING on inode being created") in overlayfs-next.
> 1. insert_inode_locked4() sets I_CREATING not inside
> spinlock and your patch sets it inside spinlock.
> technically, I guess the spinlock taken inside inode_insert5()
> to set I_NEW and insert to hash probably provides the needed
> barriers.

I was thinking along the same lines.  However, the spinlock is already
in L1 cache (due to new_inode_pseudo() pulling it in) and is obviously
uncontended, so it's basically free (at least compared to the actual
creation of a new object, which follows).

The lock here is for documentation purposes.   Using the common
pattern also makes possible future changes to i_state rules simpler.

> Do you think we should choose one of the practices
> and stick with it??

The only lockless modification of i_state is insert_inode_locked4(),
AFAICS.  I think it should either be converted to being locked or a
comment added explaining why that is not done.

> 2. I find it nicer if ovl_new_inode() would return an I_CREATING
> inode, to conform with the users of inode_insert_locked4() (e.g.
> btrfs_new_inode()) this will conform with the pattern:
> inode = XXXfs_new_inode(...);
> d_instantiate_new(dentry, inode);
> The call site d_make_root(ovl_new_inode(sb, S_IFDIR, 0));
> could either use a variant of ovl_new_inode() or we could
> let ovl_inode_init() clear I_CREATING.

Normal filesystems already know the hash key when allocating the
in-core inode.  That's not the case for overlay, that's why we need
the unusual sequence of calls.  Basically before the inode is inserted
it just doesn't matter when I_CREATING is set, we could just as well
set it right before calling inode_insert5().


  reply	other threads:[~2018-08-22 12:05 UTC|newest]

Thread overview: 3+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-08-17 11:16 ovl: set I_CREATING on inode being created Amir Goldstein
2018-08-22  8:41 ` Miklos Szeredi [this message]
2018-08-23  0:27   ` Dave Chinner

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:

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \ \ \ \ \ \ \

* 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;
as well as URLs for NNTP newsgroup(s).