Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753256AbZJHFQB (ORCPT ); Thu, 8 Oct 2009 01:16:01 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1752894AbZJHFQB (ORCPT ); Thu, 8 Oct 2009 01:16:01 -0400 Received: from 74-93-104-97-Washington.hfc.comcastbusiness.net ([74.93.104.97]:32978 "EHLO sunset.davemloft.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752850AbZJHFQA convert rfc822-to-8bit (ORCPT ); Thu, 8 Oct 2009 01:16:00 -0400 Date: Wed, 07 Oct 2009 22:15:55 -0700 (PDT) Message-Id: <20091007.221555.71697986.davem@davemloft.net> To: vapier.adi@gmail.com Cc: michael.hennerich@analog.com, netdev@vger.kernel.org, linux-kernel@vger.kernel.org, strakh@ispras.ru, uclinux-dist-devel@blackfin.uclinux.org Subject: Re: [BUG] znet.c sleeping function called from invalid context From: David Miller In-Reply-To: <8bd0f97a0910071144k1d0bf30bv60656181edae8af7@mail.gmail.com> References: <200910071847.38163.strakh@ispras.ru> <8bd0f97a0910071144k1d0bf30bv60656181edae8af7@mail.gmail.com> X-Mailer: Mew version 6.2.51 on Emacs 22.1 / Mule 5.0 (SAKAKI) Mime-Version: 1.0 Content-Type: Text/Plain; charset=iso-8859-1 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2778 Lines: 87 From: Mike Frysinger Date: Wed, 7 Oct 2009 14:44:45 -0400 > On Wed, Oct 7, 2009 at 14:47, Alexander Strakh wrote: >> ? ? ? ?KERNEL_VERSION: 2.6.31 >> ? ? ? ?DESCRIBE: >> Driver drivers/net/znet.c might sleep in atomic context, because it calls >> free_dma under claim_dma_lock: >> >> .drivers/net/znet.c: >> ?168 static int znet_request_resources (struct net_device *dev) >> ... >> ?189 ? ? ? ?flags = claim_dma_lock(); >> ?190 ? ? ? ?free_dma (znet->tx_dma); >> ?191 ? ? ? ?release_dma_lock (flags); >> ... >> >> Path to might_sleep macro from znet_request_resources: >> 1. znet_request_resources calls free_dma at >> arch/blackfin/kernel/bfin_dma_5xx.c:181 >> 2. free_dma calls arch/blackfin/kernel/bfin_dma_5xx.c:195 > > i dont think we need the dmalock mutex. it's only used to protect > read/writes to .chan_status, and that should be atomic already. > -mike I'm checking in the obvious fix to net-2.6, thanks for the report: znet: Don't claim DMA lock around free_dma() calls. It's not necessary and it's illegal too. Reported-by: Alexander Strakh Signed-off-by: David S. Miller --- drivers/net/znet.c | 8 -------- 1 files changed, 0 insertions(+), 8 deletions(-) diff --git a/drivers/net/znet.c b/drivers/net/znet.c index a0384b6..b423473 100644 --- a/drivers/net/znet.c +++ b/drivers/net/znet.c @@ -169,7 +169,6 @@ static void znet_tx_timeout (struct net_device *dev); static int znet_request_resources (struct net_device *dev) { struct znet_private *znet = netdev_priv(dev); - unsigned long flags; if (request_irq (dev->irq, &znet_interrupt, 0, "ZNet", dev)) goto failed; @@ -187,13 +186,9 @@ static int znet_request_resources (struct net_device *dev) free_sia: release_region (znet->sia_base, znet->sia_size); free_tx_dma: - flags = claim_dma_lock(); free_dma (znet->tx_dma); - release_dma_lock (flags); free_rx_dma: - flags = claim_dma_lock(); free_dma (znet->rx_dma); - release_dma_lock (flags); free_irq: free_irq (dev->irq, dev); failed: @@ -203,14 +198,11 @@ static int znet_request_resources (struct net_device *dev) static void znet_release_resources (struct net_device *dev) { struct znet_private *znet = netdev_priv(dev); - unsigned long flags; release_region (znet->sia_base, znet->sia_size); release_region (dev->base_addr, znet->io_size); - flags = claim_dma_lock(); free_dma (znet->tx_dma); free_dma (znet->rx_dma); - release_dma_lock (flags); free_irq (dev->irq, dev); } -- 1.6.4.4 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/