Re: [PATCH][STABLE backport 4.4] dmaengine: qcom: bam_dma: Fix resource leak

[Ben Hutchings <ben@decadent.org.uk>]

[-- Attachment #1: Type: text/plain, Size: 2721 bytes --]

On Tue, 2019-11-05 at 12:21 -0800, Jeffrey Hugo wrote:
> Commit 7667819385457b4aeb5fac94f67f52ab52cc10d5 upstream.
> 
> bam_dma_terminate_all() will leak resources if any of the transactions are
> committed to the hardware (present in the desc fifo), and not complete.
> Since bam_dma_terminate_all() does not cause the hardware to be updated,
> the hardware will still operate on any previously committed transactions.
> This can cause memory corruption if the memory for the transaction has been
> reassigned, and will cause a sync issue between the BAM and its client(s).
> 
> Fix this by properly updating the hardware in bam_dma_terminate_all().
> 
> Fixes: e7c0fe2a5c84 ("dmaengine: add Qualcomm BAM dma driver")
> Signed-off-by: Jeffrey Hugo <jeffrey.l.hugo@gmail.com>
> Cc: stable@vger.kernel.org
> Link: https://lore.kernel.org/r/20191017152606.34120-1-jeffrey.l.hugo@gmail.com
> Signed-off-by: Vinod Koul <vkoul@kernel.org>
> ---
> Backported to 4.4 which is lacking 6b4faeac05bc
> ("dmaengine: qcom-bam: Process multiple pending descriptors")

This seems to be needed for 3.16 as well, so I've queued it up.  Let me
know if this is wrong.

Ben.

> ---
>  drivers/dma/qcom_bam_dma.c | 14 ++++++++++++++
>  1 file changed, 14 insertions(+)
> 
> diff --git a/drivers/dma/qcom_bam_dma.c b/drivers/dma/qcom_bam_dma.c
> index 5a250cdc8376..eca5b106d7d4 100644
> --- a/drivers/dma/qcom_bam_dma.c
> +++ b/drivers/dma/qcom_bam_dma.c
> @@ -671,7 +671,21 @@ static int bam_dma_terminate_all(struct dma_chan *chan)
>  
>  	/* remove all transactions, including active transaction */
>  	spin_lock_irqsave(&bchan->vc.lock, flag);
> +	/*
> +	 * If we have transactions queued, then some might be committed to the
> +	 * hardware in the desc fifo.  The only way to reset the desc fifo is
> +	 * to do a hardware reset (either by pipe or the entire block).
> +	 * bam_chan_init_hw() will trigger a pipe reset, and also reinit the
> +	 * pipe.  If the pipe is left disabled (default state after pipe reset)
> +	 * and is accessed by a connected hardware engine, a fatal error in
> +	 * the BAM will occur.  There is a small window where this could happen
> +	 * with bam_chan_init_hw(), but it is assumed that the caller has
> +	 * stopped activity on any attached hardware engine.  Make sure to do
> +	 * this first so that the BAM hardware doesn't cause memory corruption
> +	 * by accessing freed resources.
> +	 */
>  	if (bchan->curr_txd) {
> +		bam_chan_init_hw(bchan, bchan->curr_txd->dir);
>  		list_add(&bchan->curr_txd->vd.node, &bchan->vc.desc_issued);
>  		bchan->curr_txd = NULL;
>  	}
-- 
Ben Hutchings
Larkinson's Law: All laws are basically false.



[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 833 bytes --]