Return-path: Received: from nebensachen.de ([195.34.83.29]:41680 "EHLO mail.nebensachen.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753635AbYJGUwe (ORCPT ); Tue, 7 Oct 2008 16:52:34 -0400 From: Elias Oltmanns To: Bob Copeland Cc: jirislaby@gmail.com, toralf.foerster@gmx.de, ath5k-devel@lists.ath5k.org, linux-wireless@vger.kernel.org, mickflemm@gmail.com Subject: Re: [ath5k-devel] Oops with current kernel and ath5k References: <20081002144600.M31352@bobcopeland.com> <87bpy32c1w.fsf@denkblock.local> <20081003134648.M45120@bobcopeland.com> <87fxnd2112.fsf@denkblock.local> <20081003193739.M19356@bobcopeland.com> <877i8nky65.fsf@denkblock.local> <20081007013529.GA9691@hash.localnet> <873aj8vg45.fsf@denkblock.local> <20081007130613.GB10693@hash.localnet> Date: Tue, 07 Oct 2008 22:52:22 +0200 In-Reply-To: <20081007130613.GB10693@hash.localnet> (Bob Copeland's message of "Tue, 7 Oct 2008 09:06:13 -0400") Message-ID: <87ljx0t9fd.fsf@denkblock.local> (sfid-20081007_225237_886805_A0B3A629) MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Sender: linux-wireless-owner@vger.kernel.org List-ID: Bob Copeland wrote: > On Tue, Oct > Not sure 07, 2008 at 12:44:58PM +0200, Elias Oltmanns wrote: >> Looking through the code, I'm wondering about the atomicity requirements > >> of sc->status. In my opinion, __set_bit() is not permissible in various >> places (including your use case). But since this is a problem that has >> been around before, I will send a separate patch once yours has been >> merged. > > Ok, I don't see why it's a problem in this case, but I'll look for > your patch. Note we should be doing sc->status updates under a mutex > already; if not that's a bug. Alright, in that case __set_bit() is quite sufficient, of course. In order to enforce this policy, I'd suggest the following patch. Regards, Elias --- From: Elias Oltmanns Subject: [PATCH] ath5k: Ensure atomicity of bitops on sc->status Bitops on sc->status have to be protected by the sc->lock as soon as ieee80211_register_hw() has been called. Signed-off-by: Elias Oltmanns --- drivers/net/wireless/ath5k/base.c | 8 ++++++-- 1 files changed, 6 insertions(+), 2 deletions(-) diff --git a/drivers/net/wireless/ath5k/base.c b/drivers/net/wireless/ath5k/base.c index 0676c6d..2c737da 100644 --- a/drivers/net/wireless/ath5k/base.c +++ b/drivers/net/wireless/ath5k/base.c @@ -490,6 +490,9 @@ ath5k_pci_probe(struct pci_dev *pdev, if (ret) goto err_ah; + mutex_lock(&sc->lock); + ath5k_init_leds(sc); + ATH5K_INFO(sc, "Atheros AR%s chip found (MAC: 0x%x, PHY: 0x%x)\n", ath5k_chip_name(AR5K_VERSION_VER,sc->ah->ah_mac_srev), sc->ah->ah_mac_srev, @@ -541,6 +544,7 @@ ath5k_pci_probe(struct pci_dev *pdev, /* ready to process interrupts */ __clear_bit(ATH_STAT_INVALID, sc->status); + mutex_unlock(&sc->lock); return 0; err_ah: @@ -749,8 +753,6 @@ ath5k_attach(struct pci_dev *pdev, struct ieee80211_hw *hw) goto err_queues; } - ath5k_init_leds(sc); - return 0; err_queues: ath5k_txq_release(sc); @@ -2881,12 +2883,14 @@ static void ath5k_configure_filter(struct ieee80211_hw *hw, AR5K_RX_FILTER_MCAST); if (changed_flags & (FIF_PROMISC_IN_BSS | FIF_OTHER_BSS)) { + mutex_lock(&sc->lock); if (*new_flags & FIF_PROMISC_IN_BSS) { rfilt |= AR5K_RX_FILTER_PROM; __set_bit(ATH_STAT_PROMISC, sc->status); } else __clear_bit(ATH_STAT_PROMISC, sc->status); + mutex_unlock(&sc->lock); } /* Note, AR5K_RX_FILTER_MCAST is already enabled */