linux-kernel - Re: [PATCH] dmaengine: dma_request_chan_by

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] [day] [month] [year] [list]

Message-ID: <4dd1caa7-4b95-4e06-a5ac-e2d33ce88d04@arm.com>
Date: Fri, 29 Nov 2024 16:44:36 +0000
From: Robin Murphy <robin.murphy@....com>
To: Enric Balletbo i Serra <eballetb@...hat.com>,
 Vinod Koul <vkoul@...nel.org>
Cc: dmaengine@...r.kernel.org, linux-kernel@...r.kernel.org,
 linux-arm-kernel@...ts.infradead.org, u-kumar1@...com, vigneshr@...com,
 nm@...com
Subject: Re: [PATCH] dmaengine: dma_request_chan_by_mask() defer probing
 unconditionally

On 2024-11-27 8:23 am, Enric Balletbo i Serra wrote:
> Having no DMA devices registered is not a guarantee that the device
> doesn't exist, it could be that is not registered yet, so return
> EPROBE_DEFER unconditionally so the caller can wait for the required
> DMA device registered.
> 
> Signed-off-by: Enric Balletbo i Serra <eballetb@...hat.com>
> ---
> This patch fixes the following error on TI AM69-SK
> 
> [    2.854501] cadence-qspi 47040000.spi: error -ENODEV: No Rx DMA available
> 
> The DMA device is probed after cadence-qspi driver, so deferring it
> solves the problem.

Conversely, though, it does carry some risk that if there really is no 
DMA device/driver, other callers (e.g. spi-ti-qspi) may now get stuck 
deferring forever where the -ENODEV would have let them proceed with a 
fallback to non-DMA operation. driver_deferred_probe_check_state() is 
typically a good tool for these situations, but I guess it's a bit 
tricky in a context where we don't actually have the dependent device to 
hand :/

Thanks,
Robin.

> ---
>   drivers/dma/dmaengine.c | 8 ++++----
>   1 file changed, 4 insertions(+), 4 deletions(-)
> 
> diff --git a/drivers/dma/dmaengine.c b/drivers/dma/dmaengine.c
> index c1357d7f3dc6ca7899c4d68a039567e73b0f089d..57f07b477a5d9ad8f2656584b8c0d6dffb2ab469 100644
> --- a/drivers/dma/dmaengine.c
> +++ b/drivers/dma/dmaengine.c
> @@ -889,10 +889,10 @@ struct dma_chan *dma_request_chan_by_mask(const dma_cap_mask_t *mask)
>   	chan = __dma_request_channel(mask, NULL, NULL, NULL);
>   	if (!chan) {
>   		mutex_lock(&dma_list_mutex);
> -		if (list_empty(&dma_device_list))
> -			chan = ERR_PTR(-EPROBE_DEFER);
> -		else
> -			chan = ERR_PTR(-ENODEV);
> +		/* If the required DMA device is not registered yet,
> +		 * return EPROBE_DEFER
> +		 */
> +		chan = ERR_PTR(-EPROBE_DEFER);
>   		mutex_unlock(&dma_list_mutex);
>   	}
>   
> 
> ---
> base-commit: 43fb83c17ba2d63dfb798f0be7453ed55ca3f9c2
> change-id: 20241127-defer-dma-request-chan-4f26c62c8691
> 
> Best regards,