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: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <1e71c153-e482-409c-b229-9b9c0662b67e@arm.com>
Date: Mon, 5 Feb 2024 17:55:03 +0000
From: Robin Murphy <robin.murphy@....com>
To: Andrea della Porta <andrea.porta@...e.com>, Vinod Koul
 <vkoul@...nel.org>, Florian Fainelli <florian.fainelli@...adcom.com>,
 Ray Jui <rjui@...adcom.com>, Scott Branden <sbranden@...adcom.com>,
 Broadcom internal kernel review list
 <bcm-kernel-feedback-list@...adcom.com>, dmaengine@...r.kernel.org,
 linux-rpi-kernel@...ts.infradead.org, linux-arm-kernel@...ts.infradead.org,
 linux-kernel@...r.kernel.org
Cc: Maxime Ripard <maxime@...no.tech>, Dom Cobley <popcornmix@...il.com>,
 Phil Elwell <phil@...pberrypi.com>
Subject: Re: [PATCH 04/12] bcm2835-dma: Advertise the full DMA range

On 2024-02-04 6:59 am, Andrea della Porta wrote:
> From: Phil Elwell <phil@...pberrypi.com>
> 
> Unless the DMA mask is set wider than 32 bits, DMA mapping will use a
> bounce buffer.
> 
> Signed-off-by: Phil Elwell <phil@...pberrypi.com>
> ---
>   drivers/dma/bcm2835-dma.c | 18 +++++++++++++++---
>   1 file changed, 15 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/dma/bcm2835-dma.c b/drivers/dma/bcm2835-dma.c
> index 36bad198b655..237dcdb8d726 100644
> --- a/drivers/dma/bcm2835-dma.c
> +++ b/drivers/dma/bcm2835-dma.c
> @@ -39,6 +39,7 @@
>   #define BCM2711_DMA_MEMCPY_CHAN 14
>   
>   struct bcm2835_dma_cfg_data {
> +	u64	dma_mask;
>   	u32	chan_40bit_mask;
>   };
>   
> @@ -308,10 +309,12 @@ DEFINE_SPINLOCK(memcpy_lock);
>   
>   static const struct bcm2835_dma_cfg_data bcm2835_dma_cfg = {
>   	.chan_40bit_mask = 0,
> +	.dma_mask = DMA_BIT_MASK(32),
>   };
>   
>   static const struct bcm2835_dma_cfg_data bcm2711_dma_cfg = {
>   	.chan_40bit_mask = BIT(11) | BIT(12) | BIT(13) | BIT(14),
> +	.dma_mask = DMA_BIT_MASK(36),
>   };
>   
>   static inline size_t bcm2835_dma_max_frame_length(struct bcm2835_chan *c)
> @@ -1263,6 +1266,8 @@ static struct dma_chan *bcm2835_dma_xlate(struct of_phandle_args *spec,
>   
>   static int bcm2835_dma_probe(struct platform_device *pdev)
>   {
> +	const struct bcm2835_dma_cfg_data *cfg_data;
> +	const struct of_device_id *of_id;
>   	struct bcm2835_dmadev *od;
>   	struct resource *res;
>   	void __iomem *base;
> @@ -1272,13 +1277,20 @@ static int bcm2835_dma_probe(struct platform_device *pdev)
>   	int irq_flags;
>   	uint32_t chans_available;
>   	char chan_name[BCM2835_DMA_CHAN_NAME_SIZE];
> -	const struct of_device_id *of_id;
>   	int chan_count, chan_start, chan_end;
>   
> +	of_id = of_match_node(bcm2835_dma_of_match, pdev->dev.of_node);
> +	if (!of_id) {
> +		dev_err(&pdev->dev, "Failed to match compatible string\n");
> +		return -EINVAL;
> +	}
> +
> +	cfg_data = of_id->data;

We've had of_device_get_match_data() for nearly 9 years now, and even a 
generic device_get_match_data() for 6 ;)

> +
>   	if (!pdev->dev.dma_mask)
>   		pdev->dev.dma_mask = &pdev->dev.coherent_dma_mask;

[ Passing nit: that also really shouldn't be there, especially since 
cdfee5623290 ]

>   
> -	rc = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(32));
> +	rc = dma_set_mask_and_coherent(&pdev->dev, cfg_data->dma_mask);

Wait, does chan_40bit_mask mean that you still have some channels which 
*can't* address this full mask? If so this can't work properly. You may 
well need to redesign a bit further to have a separate DMA device for 
each channel such they can each have different masks.

Thanks,
Robin.

>   	if (rc) {
>   		dev_err(&pdev->dev, "Unable to set DMA mask\n");
>   		return rc;
> @@ -1342,7 +1354,7 @@ static int bcm2835_dma_probe(struct platform_device *pdev)
>   		return -EINVAL;
>   	}
>   
> -	od->cfg_data = of_id->data;
> +	od->cfg_data = cfg_data;
>   
>   	/* Request DMA channel mask from device tree */
>   	if (of_property_read_u32(pdev->dev.of_node,

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ