[RTF] iwmc3200wifi: fix device registration ordering for cfg80211

[RTF] iwmc3200wifi: fix device registration ordering for cfg80211

[Luis R. Rodriguez]

Without this you can get a race against udev and userspace daemons
which will try to poke the device even before netdev ops and friends
are available. Device initialization will fail without this.

Cc: stable@kernel.org
Cc: Naveen Singh <nsingh@atheros.com>
Cc: Wey-Yi Guy <wey-yi.w.guy@intel.com>
Cc: Intel Linux Wireless <ilw@linux.intel.com>
Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
---

I'm pretty sure this is correct, I noticed ath6kl followed the same path as
iwmc3200wifi and I spotted a bug on ath6kl with this path.. so I'm pretty damn
sure this bug is also present on iwmc3200wifi. Please test.

 drivers/net/wireless/iwmc3200wifi/cfg80211.c |    9 ---------
 drivers/net/wireless/iwmc3200wifi/netdev.c   |    6 ++++++
 2 files changed, 6 insertions(+), 9 deletions(-)

diff --git a/drivers/net/wireless/iwmc3200wifi/cfg80211.c b/drivers/net/wireless/iwmc3200wifi/cfg80211.c
index ed57e44..2f48f72 100644
--- a/drivers/net/wireless/iwmc3200wifi/cfg80211.c
+++ b/drivers/net/wireless/iwmc3200wifi/cfg80211.c
@@ -837,17 +837,8 @@ struct wireless_dev *iwm_wdev_alloc(int sizeof_bus, struct device *dev)
 	wdev->wiphy->cipher_suites = cipher_suites;
 	wdev->wiphy->n_cipher_suites = ARRAY_SIZE(cipher_suites);
 
-	ret = wiphy_register(wdev->wiphy);
-	if (ret < 0) {
-		dev_err(dev, "Couldn't register wiphy device\n");
-		goto out_err_register;
-	}
-
 	return wdev;
 
- out_err_register:
-	wiphy_free(wdev->wiphy);
-
  out_err_new:
 	kfree(wdev);
 
diff --git a/drivers/net/wireless/iwmc3200wifi/netdev.c b/drivers/net/wireless/iwmc3200wifi/netdev.c
index 5091d77..731058e 100644
--- a/drivers/net/wireless/iwmc3200wifi/netdev.c
+++ b/drivers/net/wireless/iwmc3200wifi/netdev.c
@@ -145,6 +145,12 @@ void *iwm_if_alloc(int sizeof_bus, struct device *dev,
 
 	iwm_init_default_profile(iwm, iwm->umac_profile);
 
+	ret = wiphy_register(wdev->wiphy);
+	if (ret < 0) {
+		dev_err(dev, "Couldn't register wiphy device\n");
+		goto out_profile;
+	}
+
 	return iwm;
 
  out_profile:
-- 
1.7.4.15.g7811d

Re: [RTF] iwmc3200wifi: fix device registration ordering for cfg80211

[Luis R. Rodriguez]

On Thu, Mar 17, 2011 at 8:10 PM, Luis R. Rodriguez
<lrodriguez@atheros.com> wrote:
> Without this you can get a race against udev and userspace daemons
> which will try to poke the device even before netdev ops and friends
> are available. Device initialization will fail without this.
>
> Cc: stable@kernel.org
> Cc: Naveen Singh <nsingh@atheros.com>
> Cc: Wey-Yi Guy <wey-yi.w.guy@intel.com>
> Cc: Intel Linux Wireless <ilw@linux.intel.com>
> Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
> ---

Oops sorry, this was supposed to go to Linville.

  LUis

Re: [RTF] iwmc3200wifi: fix device registration ordering for cfg80211

[John W. Linville]

On Thu, Mar 17, 2011 at 08:11:17PM -0700, Luis R. Rodriguez wrote:
> On Thu, Mar 17, 2011 at 8:10 PM, Luis R. Rodriguez
> <lrodriguez@atheros.com> wrote:
> > Without this you can get a race against udev and userspace daemons
> > which will try to poke the device even before netdev ops and friends
> > are available. Device initialization will fail without this.
> >
> > Cc: stable@kernel.org
> > Cc: Naveen Singh <nsingh@atheros.com>
> > Cc: Wey-Yi Guy <wey-yi.w.guy@intel.com>
> > Cc: Intel Linux Wireless <ilw@linux.intel.com>
> > Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
> > ---
> 
> Oops sorry, this was supposed to go to Linville.

And probably "[RFT]" instead of "[RTF]" too. :-)

Does anyone even have this hardware?

John
-- 
John W. Linville		Someday the world will need a hero, and you
linville@tuxdriver.com			might be all we have.  Be ready.

Re: [RTF] iwmc3200wifi: fix device registration ordering for cfg80211

[Luis R. Rodriguez]

On Fri, Mar 18, 2011 at 6:50 AM, John W. Linville
<linville@tuxdriver.com> wrote:
> On Thu, Mar 17, 2011 at 08:11:17PM -0700, Luis R. Rodriguez wrote:
>> On Thu, Mar 17, 2011 at 8:10 PM, Luis R. Rodriguez
>> <lrodriguez@atheros.com> wrote:
>> > Without this you can get a race against udev and userspace daemons
>> > which will try to poke the device even before netdev ops and friends
>> > are available. Device initialization will fail without this.
>> >
>> > Cc: stable@kernel.org
>> > Cc: Naveen Singh <nsingh@atheros.com>
>> > Cc: Wey-Yi Guy <wey-yi.w.guy@intel.com>
>> > Cc: Intel Linux Wireless <ilw@linux.intel.com>
>> > Signed-off-by: Luis R. Rodriguez <lrodriguez@atheros.com>
>> > ---
>>
>> Oops sorry, this was supposed to go to Linville.
>
> And probably "[RFT]" instead of "[RTF]" too. :-)

Heh yes, sorry been WTF'ing a bit too much lately.

> Does anyone even have this hardware?

Re: [RTF] iwmc3200wifi: fix device registration ordering for cfg80211

[Luis R. Rodriguez]

> On Fri, Mar 18, 2011 at 6:50 AM, John W. Linville
>> Does anyone even have this hardware?

So similarly here we can likely hit a case where we hit the double
wiphy_unregister() twice and I think we'd oops. My preference is to
make wiphy_unregister() re-entrant and let us detect when the device
already was unregistered. Johannes?

  Luis

Re: [RTF] iwmc3200wifi: fix device registration ordering for cfg80211

[Johannes Berg]

 On Fri, 18 Mar 2011 10:41:53 -0700, Luis R. Rodriguez wrote:
>> On Fri, Mar 18, 2011 at 6:50 AM, John W. Linville
>>> Does anyone even have this hardware?

 I don't think it even exists any more.

> So similarly here we can likely hit a case where we hit the double
> wiphy_unregister() twice and I think we'd oops. My preference is to
> make wiphy_unregister() re-entrant and let us detect when the device
> already was unregistered. Johannes?

 I don't like that, it encourages sloppy driver writing. Seriously,
 why is it so hard to properly write the code?

 You can fail during init, or you can properly deregister, and that's
 about it, no?

 johannes

Re: [RTF] iwmc3200wifi: fix device registration ordering for cfg80211

[Luis R. Rodriguez]

On Fri, Mar 18, 2011 at 11:03 AM, Johannes Berg
<johannes@sipsolutions.net> wrote:
> On Fri, 18 Mar 2011 10:41:53 -0700, Luis R. Rodriguez wrote:
>>>
>>> On Fri, Mar 18, 2011 at 6:50 AM, John W. Linville
>>>>
>>>> Does anyone even have this hardware?
>
> I don't think it even exists any more.

Can the driver be removed?

>> So similarly here we can likely hit a case where we hit the double
>> wiphy_unregister() twice and I think we'd oops. My preference is to
>> make wiphy_unregister() re-entrant and let us detect when the device
>> already was unregistered. Johannes?
>
> I don't like that, it encourages sloppy driver writing. Seriously,
> why is it so hard to properly write the code?
>
> You can fail during init, or you can properly deregister, and that's
> about it, no?

That's fine too.

  Luis