linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Al Viro <viro@zeniv.linux.org.uk>
To: Amir Goldstein <amir73il@gmail.com>
Cc: Miklos Szeredi <miklos@szeredi.hu>,
	Christian Brauner <brauner@kernel.org>,
	linux-fsdevel@vger.kernel.org, linux-unionfs@vger.kernel.org
Subject: Re: [PATCH 1/4] ovl: do not open non-data lower file for fsync
Date: Fri, 4 Oct 2024 23:28:11 +0100	[thread overview]
Message-ID: <20241004222811.GU4017910@ZenIV> (raw)
In-Reply-To: <20241004221625.GR4017910@ZenIV>

On Fri, Oct 04, 2024 at 11:16:25PM +0100, Al Viro wrote:
> On Fri, Oct 04, 2024 at 12:23:39PM +0200, Amir Goldstein wrote:
> > ovl_fsync() with !datasync opens a backing file from the top most dentry
> > in the stack, checks if this dentry is non-upper and skips the fsync.
> > 
> > In case of an overlay dentry stack with lower data and lower metadata
> > above it, but without an upper metadata above it, the backing file is
> > opened from the top most lower metadata dentry and never used.
> > 
> > Fix the helper ovl_real_fdget_meta() to return an empty struct fd in
> > that case to avoid the unneeded backing file open.
> 
> Umm...  Won't that screw the callers of ovl_real_fd()?
> 
> I mean, here
> 
> > @@ -395,7 +397,7 @@ static int ovl_fsync(struct file *file, loff_t start, loff_t end, int datasync)
> >  		return ret;
> >  
> >  	ret = ovl_real_fdget_meta(file, &real, !datasync);
> > -	if (ret)
> > +	if (ret || fd_empty(real))
> >  		return ret;
> >  
> 
> you are taking account of that, but what of e.g.
>         ret = ovl_real_fdget(file, &real);
>         if (ret)
>                 return ret;
> 
>         /*
>          * Overlay file f_pos is the master copy that is preserved
>          * through copy up and modified on read/write, but only real
>          * fs knows how to SEEK_HOLE/SEEK_DATA and real fs may impose
>          * limitations that are more strict than ->s_maxbytes for specific
>          * files, so we use the real file to perform seeks.
>          */
>         ovl_inode_lock(inode);
>         fd_file(real)->f_pos = file->f_pos;
> in ovl_llseek()?  Get ovl_real_fdget_meta() called by ovl_real_fdget() and
> have it return 0 with NULL in fd_file(real), and you've got an oops right
> there, don't you?

I see... so you rely upon that thing never happening when the last argument of
ovl_real_fdget_meta() is false, including the call from ovl_real_fdget().

I still don't like the calling conventions, TBH.  Let me think a bit...

  reply	other threads:[~2024-10-04 22:28 UTC|newest]

Thread overview: 14+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-10-04 10:23 [PATCH 0/4] Stash overlay real upper file in backing_file Amir Goldstein
2024-10-04 10:23 ` [PATCH 1/4] ovl: do not open non-data lower file for fsync Amir Goldstein
2024-10-04 22:16   ` Al Viro
2024-10-04 22:28     ` Al Viro [this message]
2024-10-05  1:35       ` Al Viro
2024-10-05  6:30         ` Amir Goldstein
2024-10-05 19:49           ` Al Viro
2024-10-06  8:03             ` Amir Goldstein
2024-10-04 10:23 ` [PATCH 2/4] ovl: stash upper real file in backing_file struct Amir Goldstein
2024-10-04 10:23 ` [PATCH 3/4] ovl: convert ovl_real_fdget_meta() callers to ovl_real_file_meta() Amir Goldstein
2024-10-04 22:23   ` Al Viro
2024-10-05 12:37     ` Amir Goldstein
2024-10-04 10:23 ` [PATCH 4/4] ovl: convert ovl_real_fdget() callers to ovl_real_file() Amir Goldstein
2024-10-04 22:25   ` Al Viro

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=20241004222811.GU4017910@ZenIV \
    --to=viro@zeniv.linux.org.uk \
    --cc=amir73il@gmail.com \
    --cc=brauner@kernel.org \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-unionfs@vger.kernel.org \
    --cc=miklos@szeredi.hu \
    /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;
as well as URLs for NNTP newsgroup(s).