From: viro@parcelfarce.linux.theplanet.co.uk
To: Jonathan Corbet <corbet@lwn.net>
Cc: linux-kernel@vger.kernel.org
Subject: Re: [PATCH] RFC: Attributes in /sys/cdev
Date: Mon, 22 Sep 2003 22:00:21 +0100 [thread overview]
Message-ID: <20030922210021.GH7665@parcelfarce.linux.theplanet.co.uk> (raw)
In-Reply-To: <20030922164609.5929.qmail@lwn.net>
On Mon, Sep 22, 2003 at 10:46:09AM -0600, Jonathan Corbet wrote:
> > > I have no idea whether this follows the original plan for /sys/cdev.
> >
> > They are more of a side effect.
>
> So...to be sure I have things straight.../sys/cdev isn't really meant to be
> there, and char devices wanting to do things in sysfs should be working
> under /sys/devices or /sys/class or /sys/somethingelse?
<nod>
For block devices we really have non-trivial properties associated with them -
struct gendisk and request queue, for one thing. And these are uniform across
the entire set.
Character devices per se have *no* meaningful properties common to the entire
class. To put it another way, while block devices do form a natural class,
character ones do not.
> > * driver has embedded struct cdev in its data structures
>
> > * ->open() can use ->i_cdev to get whatever data structure driver
> > had intended and avoid any lookups of its own
>
> I noticed there's no "private" member in the cdev structure. So drivers
> should embed the structure and use container_of to get their real structure
> of interest?
Yes, if they bother with cdev of their own in the first place.
> [Forgive my ignorance here...] If I embed a struct cdev within my own
> device structure, how do I know when I can safely free said device
> structure? Will there be a release method that gets exposed at the driver
> level, or am I missing something obvious again?
Umm... Any kobject has ->release() method, obviously. NOTE: as always,
embedding kobject into one of your objects requires serious thinking about
lifetime rules for your objects. Until they are cleaned up you'd better
*not* mess with that. OTOH, in a lot of cases cleanup is needed anyway -
e.g. if you declare per-object sysctls/procfs entries/whatnot, you have
the same sort of problems already (plus you need to deal with the same
sort of races in your code that finds object by major:minor; many drivers
are FUBAR in that area).
next prev parent reply other threads:[~2003-09-22 21:00 UTC|newest]
Thread overview: 10+ messages / expand[flat|nested] mbox.gz Atom feed top
2003-09-19 23:10 Jonathan Corbet
2003-09-19 23:32 ` Greg KH
2003-09-20 1:28 ` viro
2003-09-20 2:47 ` Greg KH
2003-09-22 16:29 ` Patrick Mochel
2003-09-22 16:46 ` Jonathan Corbet
2003-09-22 21:00 ` viro [this message]
2003-09-22 21:15 ` Jonathan Corbet
2003-09-22 22:37 ` viro
2003-09-24 22:22 ` Greg KH
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=20030922210021.GH7665@parcelfarce.linux.theplanet.co.uk \
--to=viro@parcelfarce.linux.theplanet.co.uk \
--cc=corbet@lwn.net \
--cc=linux-kernel@vger.kernel.org \
/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
all inboxes | Powered by JetHome®