Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <2505800d-8625-dab0-576a-3a0221954ba3@leemhuis.info> <Y3+jz5CVA9S+h2+b@ZenIV>
In-Reply-To: <Y3+jz5CVA9S+h2+b@ZenIV>
From:   Miklos Szeredi <miklos@szeredi.hu>
Date:   Thu, 24 Nov 2022 21:33:54 +0100
Message-ID: <CAJfpeguqCrEbvv4d4jdOnczQ9VOf4u9sSghvnDOWvUb84fv6OQ@mail.gmail.com>
Subject: Re: [regression, bisected] Bug 216738 - Adding O_APPEND to O_RDWR
 with fcntl(fd, F_SETFL) does not work on overlayfs
To:     Al Viro <viro@zeniv.linux.org.uk>
Cc:     Thorsten Leemhuis <regressions@leemhuis.info>,
        Linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        "regressions@lists.linux.dev" <regressions@lists.linux.dev>,
        LKML <linux-kernel@vger.kernel.org>,
        "Christian Brauner (Microsoft)" <brauner@kernel.org>,
        Pierre Labastie <pierre.labastie@neuf.fr>,
        linux-unionfs@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, 24 Nov 2022 at 18:03, Al Viro <viro@zeniv.linux.org.uk> wrote:

> So basically we have this
> static int ovl_change_flags(struct file *file, unsigned int flags)
> {
>         struct inode *inode = file_inode(file);
>         int err;
>
>         flags &= OVL_SETFL_MASK;
>
>         if (((flags ^ file->f_flags) & O_APPEND) && IS_APPEND(inode))
>                 return -EPERM;
>
>         if ((flags & O_DIRECT) && !(file->f_mode & FMODE_CAN_ODIRECT))
>                 return -EINVAL;
>
>         if (file->f_op->check_flags) {
>                 err = file->f_op->check_flags(flags);
>                 if (err)
>                         return err;
>         }
>
>         spin_lock(&file->f_lock);
>         file->f_flags = (file->f_flags & ~OVL_SETFL_MASK) | flags;
>         spin_unlock(&file->f_lock);
>
>         return 0;
> }
> open-coding what setfl() would've done, without updating ->f_iocb_flags...
> Not hard to deal with...
>
> I could pick it in vfs.git #fixes, or Miklos could put it through his tree.
> Miklos, which way would you prefer that to go?

I'll pick this into #overlayfs-next, as a PR for this cycle is needed anyway.

Thanks,
Miklos