Linux-CIFS Archive on lore.kernel.org
 help / color / Atom feed
From: Murphy Zhou <jencce.kernel@gmail.com>
To: Pavel Shilovsky <piastryyy@gmail.com>
Cc: Murphy Zhou <jencce.kernel@gmail.com>,
	Jeff Layton <jlayton@kernel.org>,
	linux-cifs <linux-cifs@vger.kernel.org>,
	Steve French <smfrench@gmail.com>
Subject: Re: [PATCH] CIFS: unlock file across process
Date: Wed, 19 Feb 2020 10:10:39 +0800
Message-ID: <20200219021039.3mpkrmvipd6z3wes@xzhoux.usersys.redhat.com> (raw)
In-Reply-To: <CAKywueRV8+8qVP6e5nsvbpMQtwDU5mQGw5h51w=5rOsCN+Oj0w@mail.gmail.com>

On Fri, Feb 14, 2020 at 11:03:00AM -0800, Pavel Shilovsky wrote:
> Also, please make sure that resulting patch works against Windows file
> share since the locking semantics may be different there.

OK.

> 
> Depending on a kind of lease we have on a file, locks may be cached or
> not. We probably don't want to have different behavior for cached and
> non-cached locks. Especially given the fact that a lease may be broken
> in the middle of app execution and the different behavior will be
> applied immediately.

Testing new patch with and without cache=none option, both samba
and Win2019 server.

Thanks very much for reviewing!

Murphy
> 
> --
> Best regards,
> Pavel Shilovsky
> 
> пт, 14 февр. 2020 г. в 06:30, Murphy Zhou <jencce.kernel@gmail.com>:
> >
> > On Fri, Feb 14, 2020 at 07:26:46AM -0500, Jeff Layton wrote:
> > > On Fri, 2020-02-14 at 12:35 +0800, Murphy Zhou wrote:
> > > > Now child can't unlock the same file that has been locked by
> > > > parent. Fix this by not skipping unlock if requesting from
> > > > different process.
> > > >
> > > > Patch tested by LTP and xfstests using samba server.
> > > >
> > > > Signed-off-by: Murphy Zhou <jencce.kernel@gmail.com>
> > > > ---
> > > >  fs/cifs/smb2file.c | 2 --
> > > >  1 file changed, 2 deletions(-)
> > > >
> > > > diff --git a/fs/cifs/smb2file.c b/fs/cifs/smb2file.c
> > > > index afe1f03aabe3..b5bca0e13d51 100644
> > > > --- a/fs/cifs/smb2file.c
> > > > +++ b/fs/cifs/smb2file.c
> > > > @@ -151,8 +151,6 @@ smb2_unlock_range(struct cifsFileInfo *cfile, struct file_lock *flock,
> > > >                 (flock->fl_start + length) <
> > > >                 (li->offset + li->length))
> > > >                     continue;
> > > > -           if (current->tgid != li->pid)
> > > > -                   continue;
> > > >             if (cinode->can_cache_brlcks) {
> > > >                     /*
> > > >                      * We can cache brlock requests - simply remove a lock
> > >
> > > I'm not as familiar with this code as I once was, but...
> > >
> > > From fork(2) manpage:
> > >
> > >        *  The  child does not inherit process-associated record locks from its
> > >           parent (fcntl(2)).  (On the other hand,  it  does  inherit  fcntl(2)
> > >           open file description locks and flock(2) locks from its parent.)
> > >
> > > It looks like cifs_setlk calls mand_unlock_range, and that gets called
> > > from both fcntl and flock codepaths.
> > >
> > > So, I'm not sure about just removing this. It seems like the pid check
> > > is probably correct for traditional posix locks, but probably not for
> > > OFD and flock locks? What ensures that completely unrelated tasks can't
> > > unlock your locks?
> >
> > You are right Jeff. Just removing this is not right. We need to handle
> > at least 3 types of locks: posix, OFD and flock.
> >
> > Thanks very much for reviewing! I'll try to sort this out.
> > > --
> > > Jeff Layton <jlayton@kernel.org>
> > >

  reply index

Thread overview: 9+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-02-14  4:35 Murphy Zhou
2020-02-14  5:32 ` Steve French
2020-02-14 12:26 ` Jeff Layton
2020-02-14 14:28   ` Murphy Zhou
2020-02-14 19:03     ` Pavel Shilovsky
2020-02-19  2:10       ` Murphy Zhou [this message]
2020-02-24 19:39         ` Pavel Shilovsky
2020-02-25  5:15           ` Murphy Zhou
2020-02-25 19:21             ` Pavel Shilovsky

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=20200219021039.3mpkrmvipd6z3wes@xzhoux.usersys.redhat.com \
    --to=jencce.kernel@gmail.com \
    --cc=jlayton@kernel.org \
    --cc=linux-cifs@vger.kernel.org \
    --cc=piastryyy@gmail.com \
    --cc=smfrench@gmail.com \
    /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

Linux-CIFS Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/linux-cifs/0 linux-cifs/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 linux-cifs linux-cifs/ https://lore.kernel.org/linux-cifs \
		linux-cifs@vger.kernel.org
	public-inbox-index linux-cifs

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.linux-cifs


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git