Message-ID: <878rr86vml.fsf@kernel.org>
Date:   Wed, 11 May 2022 16:34:26 +0300
From:   Kalle Valo <kvalo@...nel.org>
To:     Abhishek Kumar <kuabhs@...omium.org>
Cc:     netdev@...r.kernel.org, dianders@...omium.org,
        linux-wireless@...r.kernel.org, linux-kernel@...r.kernel.org,
        ath10k@...ts.infradead.org,
        "David S. Miller" <davem@...emloft.net>,
        Eric Dumazet <edumazet@...gle.com>,
        Jakub Kicinski <kuba@...nel.org>,
        Paolo Abeni <pabeni@...hat.com>
Subject: Re: [PATCH v3] ath10k: improve BDF search fallback strategy

Abhishek Kumar <kuabhs@...omium.org> writes:

> Board data files wrapped inside board-2.bin files are
> identified based on a combination of bus architecture,
> chip-id, board-id or variants. Here is one such example
> of a BDF entry in board-2.bin file:
> bus=snoc,qmi-board-id=67,qmi-chip-id=320,variant=GO_XXXX
> It is possible for few platforms none of the combinations
> of bus,qmi-board,chip-id or variants match, e.g. if
> board-id is not programmed and thus reads board-id=0xff,
> there won't be any matching BDF to be found. In such
> situations, the wlan will fail to enumerate.
>
> Currently, to search for BDF, there are two fallback
> boardnames creates to search for BDFs in case the full BDF
> is not found. It is still possible that even the fallback
> boardnames do not match.
>
> As an improvement, search for BDF with full BDF combination
> and perform the fallback searches by stripping down the last
> elements until a BDF entry is found or none is found for all
> possible BDF combinations.e.g.
> Search for initial BDF first then followed by reduced BDF
> names as follows:
> bus=snoc,qmi-board-id=67,qmi-chip-id=320,variant=GO_XXXX
> bus=snoc,qmi-board-id=67,qmi-chip-id=320
> bus=snoc,qmi-board-id=67
> bus=snoc
> <No BDF found>
>
> Tested-on: WCN3990/hw1.0 WLAN.HL.3.2.2.c10-00754-QCAHLSWMTPL-1
> Signed-off-by: Abhishek Kumar <kuabhs@...omium.org>
> ---
>
> Changes in v3:
> - As discussed, instead of adding support for default BDF in DT, added
> a method to drop the last elements from full BDF until a BDF is found.
> - Previous patch was "ath10k: search for default BDF name provided in DT"
>
>  drivers/net/wireless/ath/ath10k/core.c | 65 +++++++++++++-------------
>  1 file changed, 32 insertions(+), 33 deletions(-)
>
> diff --git a/drivers/net/wireless/ath/ath10k/core.c b/drivers/net/wireless/ath/ath10k/core.c
> index 688177453b07..ebb0d2a02c28 100644
> --- a/drivers/net/wireless/ath/ath10k/core.c
> +++ b/drivers/net/wireless/ath/ath10k/core.c
> @@ -1426,15 +1426,31 @@ static int ath10k_core_search_bd(struct ath10k *ar,
>  	return ret;
>  }
>  
> +static bool ath10k_create_reduced_boardname(struct ath10k *ar, char *boardname)


"static int" as you use error values. Strange that the compiler doesn't
warn about that.

> +{
> +	/* Find last BDF element */
> +	char *last_field = strrchr(boardname, ',');
> +
> +	if (last_field) {
> +		/* Drop the last BDF element */
> +		last_field[0] = '\0';
> +		ath10k_dbg(ar, ATH10K_DBG_BOOT,
> +			   "boardname =%s\n", boardname);
> +		return 0;
> +	}
> +	return -ENODATA;

I would invert the check:

	if (!last_field)
        	return -ENODATA;

	/* Drop the last BDF element */
	last_field[0] = '\0';
	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boot reduced boardname %s\n",
		   boardname);

	return 0;

Also I changed the style used in the debug message.

https://wireless.wiki.kernel.org/en/users/drivers/ath10k/codingstyle#debug_messages

>  	data += magic_len;
>  	len -= magic_len;
>  
> -	/* attempt to find boardname in the IE list */
> -	ret = ath10k_core_search_bd(ar, boardname, data, len);
> +	memcpy(temp_boardname, boardname, board_len);
> +	ath10k_dbg(ar, ATH10K_DBG_BOOT, "boardname :%s\n", boardname);

"boot boardname %s\n"

>  
> -	/* if we didn't find it and have a fallback name, try that */
> -	if (ret == -ENOENT && fallback_boardname1)
> -		ret = ath10k_core_search_bd(ar, fallback_boardname1, data, len);
> +retry_search:
> +	/* attempt to find boardname in the IE list */
> +	ret = ath10k_core_search_bd(ar, temp_boardname, data, len);
>  
> -	if (ret == -ENOENT && fallback_boardname2)
> -		ret = ath10k_core_search_bd(ar, fallback_boardname2, data, len);
> +	/* If the full BDF entry was not found then drop the last element and
> +	 * recheck until a BDF is found or until all options are exhausted.
> +	 */
> +	if (ret == -ENOENT)
> +		if (!ath10k_create_reduced_boardname(ar, temp_boardname))
> +			goto retry_search;
>  
>  	if (ret == -ENOENT) {
>  		ath10k_err(ar,
>  			   "failed to fetch board data for %s from %s/%s\n",
> -			   boardname, ar->hw_params.fw.dir, filename);
> +			   temp_boardname, ar->hw_params.fw.dir, filename);
>  		ret = -ENODATA;
>  	}

This is hard to read, what about the loop below? Though I'm sure I
missed some corner case but I hope you get the idea. It's always good to
have a limit for the loops to avoid looping forever.

	for (i = 0; i < 20; i++) {
		/* attempt to find boardname in the IE list */
		ret = ath10k_core_search_bd(ar, temp_boardname, data, len);
		if (ret != -ENOENT)
			/* found it or something was wrong */
			break;

		/* If the full BDF entry was not found then drop the last element and
		 * recheck until a BDF is found or until all options are exhausted.
		 */
		ret = ath10k_create_reduced_boardname(ar, temp_boardname);
		if (ret) {
			ath10k_err(ar,
				   "failed to fetch board data for %s from %s/%s\n",
				   temp_boardname, ar->hw_params.fw.dir, filename);
			ret = -ENODATA;
			break;
		}
	};

	if (ret)
		goto err;

        return 0;

-- 
https://patchwork.kernel.org/project/linux-wireless/list/

https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches

Powered by blists - more mailing lists