From: Stephen Hemminger <shemminger@osdl.org>
To: "David S. Miller" <davem@redhat.com>
Cc: netdev@oss.sgi.com, kuznet@ms2.inr.ac.ru
Subject: Re: dev->destructor
Date: Wed, 30 Apr 2003 09:33:03 -0700 [thread overview]
Message-ID: <20030430093303.2e80e7ad.shemminger@osdl.org> (raw)
In-Reply-To: <20030429.232631.68131803.davem@redhat.com>
On Tue, 29 Apr 2003 23:26:31 -0700 (PDT)
"David S. Miller" <davem@redhat.com> wrote:
>
> Stephen, you're right about the dev->destructor problem.
> I misread your postings, and I'm very sorry about that.
> We were talking about two different things, and admittedly
> I had forgotten how some of this stuff works.
>
> Alexey, currently dev->{open,close} are what does get/put
> of device module reference.
>
> However, device unregister can explode if dev->destructor is
> present. Unlike in dev->destructor==NULL case, we do not
> wait for remnant dev->refcnt to go away. Therefore we could
> invoke dev->destructor() after module is unloaded.
>
> I guess there are two ways to address this problem:
>
> 1) dev_get() gets module reference and dev_put() puts is.
> Ugly, as this means dev_get() can fail, but this does
> cover all the possible cases.
>
> 2) Make unregister_netdev() wait for refcount to reach 1
> regardless of whether dev->destructor is NULL or not.
>
> I don't like #1. Do you see some holes in #2?
>
> As Stephen brought up, this also means we should do something
> about that NETDEV_UNREGISTER code in dst_dev_event() :-(
There are other (not nice possibilities).
A) Require driver have a wait queue per device and wait after unregister.
Ugly and requires repeated work.
B) Put wait queue and wakeup logic in netdevice/unregister_netdev.
Adds more to already overloaded netdevice structure, but
could cleanup existing polling stuff.
C) Audit unregister notifier callbacks to ensure they all dev_put,
all references to device. This works for
normal IPv4 except for the dst cache. The following patch causes
the dst cache to do this.
--- linux-2.5/net/core/dst.c 2003-04-29 11:54:39.000000000 -0700
+++ linux-2.5-bridge/net/core/dst.c 2003-04-29 10:17:15.000000000 -0700
@@ -228,7 +228,7 @@
_race_ _condition_.
*/
if (event!=NETDEV_DOWN &&
- dev->destructor == NULL &&
+ (dev->destructor == NULL || dev->owner) &&
dst->output == dst_blackhole) {
dst->dev = &loopback_dev;
dev_put(dev);
D) Your option #2 only needs to be done if device is a module
otherwise, can just let destructor run later.
--- linux-2.5/net/core/dev.c 2003-04-29 11:54:39.000000000 -0700
+++ linux-2.5-bridge/net/core/dev.c 2003-04-30 09:28:34.000000000 -0700
@@ -2737,7 +2737,7 @@
free_divert_blk(dev);
#endif
- if (dev->destructor != NULL) {
+ if (dev->destructor != NULL && dev->owner == NULL) {
#ifdef NET_REFCNT_DEBUG
if (atomic_read(&dev->refcnt) != 1)
printk(KERN_DEBUG "unregister_netdevice: holding %s "
next prev parent reply other threads:[~2003-04-30 16:33 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2003-04-30 6:26 dev->destructor David S. Miller
2003-04-30 16:33 ` Stephen Hemminger [this message]
2003-05-01 1:10 ` dev->destructor kuznet
2003-05-01 7:00 ` dev->destructor David S. Miller
2003-05-01 12:01 ` dev->destructor Rusty Russell
2003-05-01 11:09 ` dev->destructor David S. Miller
2003-05-01 17:51 ` dev->destructor Arnaldo Carvalho de Melo
2003-05-01 16:55 ` dev->destructor David S. Miller
2003-05-01 17:28 ` dev->destructor Arnaldo Carvalho de Melo
2003-05-02 4:06 ` dev->destructor kuznet
2003-05-02 5:25 ` dev->destructor Rusty Russell
2003-05-02 20:48 ` dev->destructor David S. Miller
2003-05-03 4:07 ` dev->destructor Rusty Russell
2003-05-03 3:46 ` dev->destructor David S. Miller
2003-05-05 5:18 ` dev->destructor Rusty Russell
2003-05-03 4:00 ` dev->destructor David S. Miller
2003-05-05 16:08 ` dev->destructor Stephen Hemminger
2003-05-06 14:25 ` dev->destructor David S. Miller
2003-05-07 2:54 ` dev->destructor Rusty Russell
2003-05-05 20:00 ` dev->destructor Stephen Hemminger
2003-05-06 4:18 ` dev->destructor Rusty Russell
2003-05-06 14:23 ` dev->destructor David S. Miller
2003-05-06 22:32 ` [PATCH 2.5.69] IPV4 should use dev_hold Stephen Hemminger
2003-05-07 7:32 ` David S. Miller
2003-05-07 2:50 ` dev->destructor Rusty Russell
2003-05-07 3:58 ` dev->destructor David S. Miller
2003-05-06 22:35 ` dev->destructor Stephen Hemminger
2003-05-06 23:51 ` [RFC] Experiment with dev and module ref counts Stephen Hemminger
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=20030430093303.2e80e7ad.shemminger@osdl.org \
--to=shemminger@osdl.org \
--cc=davem@redhat.com \
--cc=kuznet@ms2.inr.ac.ru \
--cc=netdev@oss.sgi.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
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.