linux-kernel - Re: [PATCH v2 1/4] dmaengine: sf-pdma: Support of_dma_controller

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <e5d2e96b-9b16-4aaf-9291-76d1d2222c44@sifive.com>
Date:   Tue, 17 Oct 2023 13:33:27 -0500
From:   Samuel Holland <samuel.holland@...ive.com>
To:     shravan chippa <shravan.chippa@...rochip.com>,
        green.wan@...ive.com, vkoul@...nel.org, robh+dt@...nel.org,
        krzysztof.kozlowski+dt@...aro.org, palmer@...belt.com,
        paul.walmsley@...ive.com, conor+dt@...nel.org
Cc:     dmaengine@...r.kernel.org, devicetree@...r.kernel.org,
        linux-riscv@...ts.infradead.org, linux-kernel@...r.kernel.org,
        nagasuresh.relli@...rochip.com, praveen.kumar@...rochip.com
Subject: Re: [PATCH v2 1/4] dmaengine: sf-pdma: Support
 of_dma_controller_register()

Hi,

On 2023-10-02 11:22 PM, shravan chippa wrote:
> From: Shravan Chippa <shravan.chippa@...rochip.com>
> 
> Update sf-pdma driver to adopt generic DMA device tree bindings.
> It calls of_dma_controller_register() with sf-pdma specific
> of_dma_xlate to get the generic DMA device tree helper support
> and the DMA clients can look up the sf-pdma controller using
> standard APIs.
> 
> Signed-off-by: Shravan Chippa <shravan.chippa@...rochip.com>
> ---
>  drivers/dma/sf-pdma/sf-pdma.c | 44 +++++++++++++++++++++++++++++++++++
>  1 file changed, 44 insertions(+)
> 
> diff --git a/drivers/dma/sf-pdma/sf-pdma.c b/drivers/dma/sf-pdma/sf-pdma.c
> index d1c6956af452..06a0912a12a1 100644
> --- a/drivers/dma/sf-pdma/sf-pdma.c
> +++ b/drivers/dma/sf-pdma/sf-pdma.c
> @@ -20,6 +20,7 @@
>  #include <linux/mod_devicetable.h>
>  #include <linux/dma-mapping.h>
>  #include <linux/of.h>
> +#include <linux/of_dma.h>
>  #include <linux/slab.h>
>  
>  #include "sf-pdma.h"
> @@ -490,6 +491,33 @@ static void sf_pdma_setup_chans(struct sf_pdma *pdma)
>  	}
>  }
>  
> +static struct dma_chan *sf_pdma_of_xlate(struct of_phandle_args *dma_spec,
> +					 struct of_dma *ofdma)
> +{
> +	struct sf_pdma *pdma = ofdma->of_dma_data;
> +	struct device *dev = pdma->dma_dev.dev;
> +	struct sf_pdma_chan  *chan;
> +	struct dma_chan *c;
> +	u32 channel_id;
> +
> +	if (dma_spec->args_count != 1) {
> +		dev_err(dev, "Bad number of cells\n");
> +		return NULL;
> +	}
> +
> +	channel_id = dma_spec->args[0];
> +
> +	chan = &pdma->chans[channel_id];
> +
> +	c = dma_get_slave_channel(&chan->vchan.chan);

This does not look right to me. All of the channels in the controller are
identical and support arbitrary addresses, so there is no need to use a specific
physical channel. And unless Microchip has added something on top, the only way
to trigger a transfer is through the MMIO interface, so there is no request ID
to differentiate virtual channels either.

So it seems to me that #dma-cells should really be 0, and this function should
just call dma_get_any_slave_channel().

Regards,
Samuel

> +	if (!c) {
> +		dev_err(dev, "No more channels available\n");
> +		return NULL;
> +	}
> +
> +	return c;
> +}
> +
>  static int sf_pdma_probe(struct platform_device *pdev)
>  {
>  	struct sf_pdma *pdma;
> @@ -563,7 +591,20 @@ static int sf_pdma_probe(struct platform_device *pdev)
>  		return ret;
>  	}
>  
> +	ret = of_dma_controller_register(pdev->dev.of_node,
> +					 sf_pdma_of_xlate, pdma);
> +	if (ret < 0) {
> +		dev_err(&pdev->dev,
> +			"Can't register SiFive Platform OF_DMA. (%d)\n", ret);
> +		goto err_unregister;
> +	}
> +
>  	return 0;
> +
> +err_unregister:
> +	dma_async_device_unregister(&pdma->dma_dev);
> +
> +	return ret;
>  }
>  
>  static int sf_pdma_remove(struct platform_device *pdev)
> @@ -583,6 +624,9 @@ static int sf_pdma_remove(struct platform_device *pdev)
>  		tasklet_kill(&ch->err_tasklet);
>  	}
>  
> +	if (pdev->dev.of_node)
> +		of_dma_controller_free(pdev->dev.of_node);
> +
>  	dma_async_device_unregister(&pdma->dma_dev);
>  
>  	return 0;