linux-kernel - Re: [PATCH] wifi: ath10k: Drop of_get

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: <fdaab417-62b7-4151-be35-34fb7a354eae@oss.qualcomm.com>
Date: Thu, 13 Mar 2025 10:29:55 -0700
From: Jeff Johnson <jeff.johnson@....qualcomm.com>
To: "Rob Herring (Arm)" <robh@...nel.org>, Jeff Johnson <jjohnson@...nel.org>
Cc: linux-wireless@...r.kernel.org, ath10k@...ts.infradead.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] wifi: ath10k: Drop of_get_property() call

On 3/12/2025 4:42 PM, Rob Herring (Arm) wrote:
> There's no need to check the property presence and length before calling
> of_property_read_u8_array() as it will return an error if the property
> is missing or the length is too small. The return errno doesn't matter
> to the caller, so no change in behavior there.
> 
> Change of_property_read_u8_array() to of_property_read_variable_u8_array()
> as the former allows properties to be longer than the requested length.
> Now the property has to be the exact length requested as the removed
> check required.
> 
> This part of a larger effort to remove DT functions like
> of_get_property() and of_find_property() which return raw DT data
> having no reference counting.
> 
> Signed-off-by: Rob Herring (Arm) <robh@...nel.org>
> ---
>  drivers/net/wireless/ath/ath10k/core.c | 22 ++++++----------------
>  1 file changed, 6 insertions(+), 16 deletions(-)
> 
> diff --git a/drivers/net/wireless/ath/ath10k/core.c b/drivers/net/wireless/ath/ath10k/core.c
> index b3294287bce1..db7444af251d 100644
> --- a/drivers/net/wireless/ath/ath10k/core.c
> +++ b/drivers/net/wireless/ath/ath10k/core.c
> @@ -1889,7 +1889,7 @@ static int ath10k_download_cal_file(struct ath10k *ar,
>  static int ath10k_download_cal_dt(struct ath10k *ar, const char *dt_name)
>  {
>  	struct device_node *node;
> -	int data_len;
> +	int data_len = ar->hw_params.cal_data_len;
>  	void *data;
>  	int ret;
>  
> @@ -1900,28 +1900,18 @@ static int ath10k_download_cal_dt(struct ath10k *ar, const char *dt_name)
>  		 */
>  		return -ENOENT;
>  
> -	if (!of_get_property(node, dt_name, &data_len)) {
> -		/* The calibration data node is optional */
> -		return -ENOENT;
> -	}

note that there is one instance where .cal_data_len = 0

so i suggest that there still needs to be an early return here to avoid
kmalloc(0):

	if (!data_len)
		return -ENOENT;

> -
> -	if (data_len != ar->hw_params.cal_data_len) {
> -		ath10k_warn(ar, "invalid calibration data length in DT: %d\n",
> -			    data_len);
> -		ret = -EMSGSIZE;
> -		goto out;
> -	}
> -
>  	data = kmalloc(data_len, GFP_KERNEL);
>  	if (!data) {
>  		ret = -ENOMEM;
>  		goto out;
>  	}
>  
> -	ret = of_property_read_u8_array(node, dt_name, data, data_len);
> +	ret = of_property_read_variable_u8_array(node, dt_name, data, data_len, data_len);
>  	if (ret) {
> -		ath10k_warn(ar, "failed to read calibration data from DT: %d\n",
> -			    ret);
> +		/* Don't warn if optional property not found  */
> +		if (ret != -EINVAL)
> +			ath10k_warn(ar, "failed to read calibration data from DT: %d\n",
> +				    ret);
>  		goto out_free;
>  	}
>  

this could made even cleaner using cleanup.h functionality (all of the gotos
could disappear). perhaps I'll do that as a follow-up patch on top of your patch.

/jeff