linux-fsdevel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Al Viro <viro@ZenIV.linux.org.uk>
To: Miklos Szeredi <miklos@szeredi.hu>
Cc: Miklos Szeredi <mszeredi@redhat.com>,
	overlayfs <linux-unionfs@vger.kernel.org>,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
	Linus Torvalds <torvalds@linux-foundation.org>
Subject: Re: [PATCH 14/39] ovl: stack file ops
Date: Tue, 12 Jun 2018 03:29:26 +0100	[thread overview]
Message-ID: <20180612022926.GY30522@ZenIV.linux.org.uk> (raw)
In-Reply-To: <CAJfpegtEGuBqmJ+6Kzd0456o0xyebe3LBk66xWQcuzsenqKVeg@mail.gmail.com>

On Mon, Jun 11, 2018 at 09:09:04AM +0200, Miklos Szeredi wrote:

[context: opening files in layers with unholy mix of overlayfs
->f_path and layer's ->f_inode/->f_op]

> I'd really like to get there some time but...
> 
> List of basic requirements:
> 
>  - Private mmap of overlay file shares page cache with lower file (and
> hence with all other overlays using the same lower file).
> 
>  - /proc/PID/maps shows correct path.
> 
> Thought about setting f_mapping/i_mapping of overlay file to that of
> underlying file.  But that breaks when doing a copy-up.  We can't just
> go and change those mapping pointers, assumption is that those remain
> constant (we'd need READ_ONCE() for all cases where we use the mapping
> more than once).  It's probably doable, but it's a large and fragile
> change.

We are really asking for trouble here - anything with e.g. ->read_iter()
using dentry will get in trouble with that kind of games.  Consider something
like

foo_read_iter(struct kiocb *iocb, struct iov_iter *to)
{
	struct file *file = iocb->ki_filp;
	struct foo_data *p = file->f_path.dentry->d_fsdata;
	...
}

which will work just fine for files on foofs, where we have ->d_fsdata set
on lookup.  Now, try to use foofs as a layer; suddenly, you get foofs
files with ->f_path.dentry being *overlayfs* dentry, with ->d_fsdata
being nothing like struct foo_data *.

Better yet, consider

foo_open(struct inode *inode, struct file *file)
{
	struct dentry *dentry = file->f_path.dentry;
	...
	foo_add_splat(dentry, splat);
	...
}
where foo_add_splat() inserts struct foo_splat into an hlist starting
in dentry->d_fsdata.  That's not a pure theory - we *do* have ->open()
instances doing things of that sort.  That'll bugger overlayfs quite
badly, not to mention that foofs methods won't be happy with overlayfs
dentries.

It might (or might not) work for the filesystems you'd been testing
on, but it's a lot of trouble waiting to happen.  Hell, try and use
ecryptfs as lower layer, see how fast it'll blow up.  Sure, it's
a dumb testcase, but I don't see how to check if something more
realistic is trouble-free.

I'd been trying to come up with some way to salvage that kludge of yours,
but I don't see any solutions.  We don't have good proxies for "this
filesystem might be unsafe as lower layer" ;-/

Frankly, it might be saner and safer to teach procfs (and similar
places) to do more than just use ->vm_file->f_path.  _That_ at least
is much more local in impact.

  reply	other threads:[~2018-06-12  2:29 UTC|newest]

Thread overview: 83+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2018-05-29 14:43 [PATCH 00/39] overlayfs: stack file operations Miklos Szeredi
2018-05-29 14:43 ` [PATCH 01/39] vfs: dedpue: return loff_t Miklos Szeredi
2018-06-04  8:43   ` Christoph Hellwig
2018-06-05  8:33     ` Miklos Szeredi
2018-06-06 15:09       ` Darrick J. Wong
2018-06-18 20:08         ` Miklos Szeredi
2018-05-29 14:43 ` [PATCH 02/39] vfs: dedupe: rationalize args Miklos Szeredi
2018-06-06 15:02   ` Darrick J. Wong
2018-05-29 14:43 ` [PATCH 03/39] vfs: dedupe: extract helper for a single dedup Miklos Szeredi
2018-05-29 15:41   ` Amir Goldstein
2018-05-29 16:04     ` Amir Goldstein
2018-06-04  8:44   ` Christoph Hellwig
2018-05-29 14:43 ` [PATCH 04/39] vfs: add path_open() Miklos Szeredi
2018-06-04  8:46   ` Christoph Hellwig
2018-06-10  4:36     ` Al Viro
2018-05-29 14:43 ` [PATCH 05/39] vfs: optionally don't account file in nr_files Miklos Szeredi
2018-06-04  8:47   ` Christoph Hellwig
2018-06-04  8:57     ` Miklos Szeredi
2018-06-10  4:41   ` Al Viro
2018-05-29 14:43 ` [PATCH 06/39] vfs: add f_op->pre_mmap() Miklos Szeredi
2018-06-04  8:48   ` Christoph Hellwig
2018-06-05 11:36     ` Miklos Szeredi
2018-05-29 14:43 ` [PATCH 07/39] vfs: export vfs_ioctl() to modules Miklos Szeredi
2018-06-04  8:49   ` Christoph Hellwig
2018-06-10  4:57     ` Al Viro
2018-06-11  7:19       ` Miklos Szeredi
2018-06-11 16:24         ` Christoph Hellwig
2018-06-19 14:04           ` Miklos Szeredi
2018-06-19 14:24             ` Christoph Hellwig
2018-06-19 14:34               ` Miklos Szeredi
2018-06-19 14:54                 ` Al Viro
2018-05-29 14:43 ` [PATCH 08/39] vfs: export vfs_dedupe_file_range_one() " Miklos Szeredi
2018-05-29 14:43 ` [PATCH 09/39] ovl: copy up times Miklos Szeredi
2018-05-29 14:43 ` [PATCH 10/39] ovl: copy up inode flags Miklos Szeredi
2018-05-29 14:43 ` [PATCH 11/39] Revert "Revert "ovl: get_write_access() in truncate"" Miklos Szeredi
2018-05-29 14:43 ` [PATCH 12/39] ovl: copy up file size as well Miklos Szeredi
2018-05-29 14:43 ` [PATCH 13/39] ovl: deal with overlay files in ovl_d_real() Miklos Szeredi
2018-05-29 14:43 ` [PATCH 14/39] ovl: stack file ops Miklos Szeredi
2018-06-10  4:13   ` Al Viro
2018-06-11  7:09     ` Miklos Szeredi
2018-06-12  2:29       ` Al Viro [this message]
2018-06-12  2:40         ` Al Viro
2018-06-12  9:24           ` Miklos Szeredi
2018-06-12 18:24             ` Al Viro
2018-06-12 18:31               ` Al Viro
2018-06-13  9:21                 ` Miklos Szeredi
2018-06-15  5:47                   ` Al Viro
2018-06-18 11:50                     ` Miklos Szeredi
2018-06-13 11:56               ` J. R. Okajima
2018-05-29 14:43 ` [PATCH 15/39] ovl: add helper to return real file Miklos Szeredi
2018-06-10  5:42   ` Al Viro
2018-06-11  8:11     ` Miklos Szeredi
2018-05-29 14:43 ` [PATCH 16/39] ovl: add ovl_read_iter() Miklos Szeredi
2018-05-29 14:43 ` [PATCH 17/39] ovl: add ovl_write_iter() Miklos Szeredi
2018-05-29 14:43 ` [PATCH 18/39] ovl: add ovl_fsync() Miklos Szeredi
2018-05-29 14:43 ` [PATCH 19/39] ovl: add ovl_mmap() Miklos Szeredi
2018-06-10  5:24   ` Al Viro
2018-06-11  7:58     ` Miklos Szeredi
2018-05-29 14:43 ` [PATCH 20/39] ovl: add ovl_fallocate() Miklos Szeredi
2018-05-29 14:43 ` [PATCH 21/39] ovl: add lsattr/chattr support Miklos Szeredi
2018-05-29 14:43 ` [PATCH 22/39] ovl: add ovl_fiemap() Miklos Szeredi
2018-05-29 14:43 ` [PATCH 23/39] ovl: add O_DIRECT support Miklos Szeredi
2018-06-10  5:31   ` Al Viro
2018-06-11  8:08     ` Miklos Szeredi
2018-05-29 14:43 ` [PATCH 24/39] ovl: add reflink/copyfile/dedup support Miklos Szeredi
2018-05-29 14:43 ` [PATCH 25/39] vfs: don't open real Miklos Szeredi
2018-05-29 14:43 ` [PATCH 26/39] ovl: copy-up on MAP_SHARED Miklos Szeredi
2018-05-29 14:43 ` [PATCH 27/39] ovl: obsolete "check_copy_up" module option Miklos Szeredi
2018-05-29 15:13   ` Amir Goldstein
2018-05-30  8:26     ` Miklos Szeredi
2018-05-29 14:43 ` [PATCH 28/39] ovl: fix documentation of non-standard behavior Miklos Szeredi
2018-05-29 14:43 ` [PATCH 29/39] vfs: simplify dentry_open() Miklos Szeredi
2018-05-29 14:43 ` [PATCH 30/39] Revert "ovl: fix may_write_real() for overlayfs directories" Miklos Szeredi
2018-05-29 14:43 ` [PATCH 31/39] Revert "ovl: don't allow writing ioctl on lower layer" Miklos Szeredi
2018-05-29 14:43 ` [PATCH 32/39] vfs: fix freeze protection in mnt_want_write_file() for overlayfs Miklos Szeredi
2018-06-04  8:50   ` Christoph Hellwig
2018-05-29 14:43 ` [PATCH 33/39] Revert "ovl: fix relatime for directories" Miklos Szeredi
2018-05-29 14:43 ` [PATCH 34/39] Revert "vfs: update ovl inode before relatime check" Miklos Szeredi
2018-05-29 14:43 ` [PATCH 35/39] Revert "vfs: add flags to d_real()" Miklos Szeredi
2018-05-29 14:43 ` [PATCH 36/39] Revert "vfs: do get_write_access() on upper layer of overlayfs" Miklos Szeredi
2018-05-29 14:43 ` [PATCH 37/39] Partially revert "locks: fix file locking on overlayfs" Miklos Szeredi
2018-05-29 14:43 ` [PATCH 38/39] Revert "fsnotify: support overlayfs" Miklos Szeredi
2018-05-29 14:43 ` [PATCH 39/39] vfs: remove open_flags from d_real() Miklos Szeredi

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=20180612022926.GY30522@ZenIV.linux.org.uk \
    --to=viro@zeniv.linux.org.uk \
    --cc=linux-fsdevel@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-unionfs@vger.kernel.org \
    --cc=miklos@szeredi.hu \
    --cc=mszeredi@redhat.com \
    --cc=torvalds@linux-foundation.org \
    --subject='Re: [PATCH 14/39] ovl: stack file ops' \
    /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

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).