Return-path: Received: from smtp.nokia.com ([147.243.128.26]:65130 "EHLO mgw-da02.nokia.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751707Ab0KWIrf (ORCPT ); Tue, 23 Nov 2010 03:47:35 -0500 Received: from nokia.com (localhost [127.0.0.1]) by mgw-da02.nokia.com (Switch-3.4.3/Switch-3.4.3) with ESMTP id oAN8l4ZE026614 for ; Tue, 23 Nov 2010 10:47:34 +0200 Subject: Re: [PATCH] wl12xx: Fix kernel crash related to hw recovery and interface shutdown From: Luciano Coelho To: Juuso Oikarinen Cc: linux-wireless@vger.kernel.org In-Reply-To: <1290491470.4284.9.camel@wimaxnb.nmp.nokia.com> References: <1290086342-12015-1-git-send-email-juuso.oikarinen@nokia.com> <1290430346.3322.4.camel@chilepepper> <1290491470.4284.9.camel@wimaxnb.nmp.nokia.com> Content-Type: text/plain; charset="UTF-8" Date: Tue, 23 Nov 2010 10:47:37 +0200 Message-ID: <1290502057.10621.32.camel@chilepepper> Mime-Version: 1.0 Sender: linux-wireless-owner@vger.kernel.org List-ID: On Tue, 2010-11-23 at 07:51 +0200, Juuso Oikarinen wrote: > On Mon, 2010-11-22 at 14:52 +0200, Luciano Coelho wrote: > > On Thu, 2010-11-18 at 15:19 +0200, juuso.oikarinen@nokia.com wrote: > > > From: Juuso Oikarinen > > > > > > It is possible that the op_remove_interface function is invoked exactly at > > > the same time has hw recovery is started. In this case it is possible for the > > > interface to be already removed in the op_remove_interface call, which > > > currently leads to a kernel warning and a subsequent kernel crash. > > > > > > Fix this by ignoring the op_remove_interface call if the interface is already > > > down at that point. > > > > > > Signed-off-by: Juuso Oikarinen > > > --- > > > > [...] > > > > > diff --git a/drivers/net/wireless/wl12xx/main.c b/drivers/net/wireless/wl12xx/main.c > > > index 31f0e2f..11b0477 100644 > > > --- a/drivers/net/wireless/wl12xx/main.c > > > +++ b/drivers/net/wireless/wl12xx/main.c > > > @@ -1157,10 +1157,16 @@ static void wl1271_op_remove_interface(struct ieee80211_hw *hw, > > > struct wl1271 *wl = hw->priv; > > > > > > mutex_lock(&wl->mutex); > > > - WARN_ON(wl->vif != vif); > > > - __wl1271_op_remove_interface(wl); > > > - mutex_unlock(&wl->mutex); > > > + /* > > > + * wl->vif can be null here if someone shuts down the interface > > > + * just when hardware recovery has been started. > > > + */ > > > + if (wl->vif) { > > > + WARN_ON(wl->vif != vif); > > > + __wl1271_op_remove_interface(wl); > > > + } > > > > Should you still remove the interface if the vif you received is wrong? > > Surely, something is totally wrong if you get a different vif, but maybe > > removing the interface here will just confuse things even more? > > Dunno if it would be better to remove or leave unremoved - probably does > not matter. If the vif is wrong, there is some serious bug somewhere, > and probably the result is serious instability anyway. > > The WARN_ON is there just to validate the assumption the driver makes > about the vif, and this function call. If there is a bug somewhere, or > the assumption somehow changes, the warning will be a clear indication > of it. Yep, it's okay like this. And it was like that earlier anyway. I was just wondering if it would be possible for a similar thing as wl->vif being NULL (which your patch addresses) happen, such as wl->vif changing before we had the chance to fully remove the interface. It will probably not happen and, as you say, the WARN_ON will be a good indication that something went wrong. Applied to wl12xx/master. Thanks! -- Cheers, Luca.