From mboxrd@z Thu Jan  1 00:00:00 1970
From: Stephen Warren <swarren@wwwdotorg.org>
Subject: Re: [PATCH 2/2] dma: tegra20-apbdma: channel freeing correction
Date: Mon, 29 Oct 2012 09:27:09 -0600
Message-ID: <508EA04D.5050805@wwwdotorg.org>
References: <1351433873-14082-1-git-send-email-digetx@gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Return-path: <linux-kernel-owner@vger.kernel.org>
In-Reply-To: <1351433873-14082-1-git-send-email-digetx@gmail.com>
Sender: linux-kernel-owner@vger.kernel.org
To: Dmitry Osipenko <digetx@gmail.com>, Laxman Dewangan <ldewangan@nvidia.com>
Cc: vinod.koul@intel.com, linux-tegra@vger.kernel.org, linux-kernel@vger.kernel.org
List-Id: linux-tegra@vger.kernel.org

On 10/28/2012 08:17 AM, Dmitry Osipenko wrote:
> Fixed channel "lock" after free.
> 
> Example: Channel 1 was allocated and prepared as slave_sg, used and freed. Now preparation of cyclic dma on channel 1 will fail with err "DMA
> configuration conflict" because tdc->isr_handler still selected to handle_once_dma_done.
> 
> This happens because tegra_dma_abort_all() won't be called on channel freeing if pending list is empty.

That commit description isn't correctly wrapped.

> diff --git a/drivers/dma/tegra20-apb-dma.c b/drivers/dma/tegra20-apb-dma.c

> @@ -1147,6 +1147,7 @@ static void tegra_dma_free_chan_resources(struct dma_chan *dc)
>  
>  	if (tdc->busy)
>  		tegra_dma_terminate_all(dc);
> +	tdc->isr_handler = NULL;

Should we remove that assignment from tegra_dma_abort_all(); perhaps it
is redundant now?

Actually, I wonder if the correct fix isn't to:

a) Always call tegra_dma_terminate_all() from
tegra_dma_free_chan_resources() irrespective of busy state.

b) Make tegra_dma_terminate_all() always call tegra_dma_abort_all()
irrespective of whether list_empty(&tdc->pending_sg_req).

But then I wonder: should tdc->isr_handler get left set to non-NULL in
the scenario mentioned in your commit description at all; should it be
cleared as soon as the channel is idle in all cases, so that it doesn't
need to be cleared when freeing the channel?

I CC'd Laxman, the driver author, to comment here.