linux-kernel - Re: [PATCH 3/3] libnvdimm/ioctl: Switch to cdev_register

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CAPcyv4h74NjqcuUjv4zFKHAxio_bV0bngLoxP=ACw=JvMfq-UA@mail.gmail.com>
Date:   Thu, 21 Jan 2021 09:02:53 -0800
From:   Dan Williams <dan.j.williams@...el.com>
To:     Greg KH <gregkh@...uxfoundation.org>
Cc:     Vishal Verma <vishal.l.verma@...el.com>,
        Dave Jiang <dave.jiang@...el.com>,
        Ira Weiny <ira.weiny@...el.com>,
        Logan Gunthorpe <logang@...tatee.com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        linux-nvdimm <linux-nvdimm@...ts.01.org>
Subject: Re: [PATCH 3/3] libnvdimm/ioctl: Switch to cdev_register_queued()

On Thu, Jan 21, 2021 at 12:16 AM Greg KH <gregkh@...uxfoundation.org> wrote:
>
> On Wed, Jan 20, 2021 at 11:39:08AM -0800, Dan Williams wrote:
> > The ioctl implementation in libnvdimm is a case study in what can be
> > cleaned up when the cdev core handles synchronizing in-flight ioctls
> > with device removal. Switch to cdev_register_queued() which allows for
> > the ugly context lookup and activity tracking implementation to be
> > dropped, among other cleanups.
>
> I'm confused, the cdev handles the filesystem access from /dev/ which
> handles the ioctl.  Any use of a cdev with relationship to a struct
> device that might go away is independent, so we really should not tie
> these together in any way.

Oh, no, there's no object lifetime ties here, that problem was already
solved with Logan's addition of cdev_device_add(). Instead, this about
file_operations liveness. Look at procfs and debugfs that Christoph
pointed out have open coded the same facility. When the driver is
unbound ongoing file operations should be flushed and blocked
currently the cdev api is forcing every driver to open code this
themselves.

This seems a pattern that is asking to be unified as a general
"managed fops" concept.