linux-kernel - Re: [PATCH v2 3/4] spi: s3c64xx: add sleep during transfer

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: <874bb3d7-9b71-2ab0-9f0c-7f4066db50a4@samsung.com>
Date:   Fri, 21 Apr 2023 11:53:15 +0900
From:   Jaewon Kim <jaewon02.kim@...sung.com>
To:     Andi Shyti <andi.shyti@...nel.org>
Cc:     Krzysztof Kozlowski <krzysztof.kozlowski@...aro.org>,
        Mark Brown <broonie@...nel.org>, Andi Shyti <andi@...zian.org>,
        Alim Akhtar <alim.akhtar@...sung.com>,
        linux-spi@...r.kernel.org, linux-samsung-soc@...r.kernel.org,
        linux-arm-kernel@...ts.infradead.org, linux-kernel@...r.kernel.org,
        Chanho Park <chanho61.park@...sung.com>
Subject: Re: [PATCH v2 3/4] spi: s3c64xx: add sleep during transfer

Hi Andi,


On 23. 4. 20. 00:56, Andi Shyti wrote:
> Hi Jaewon,
>
>>>> In polling mode, the status register is constantly read to check transfer
>>>> completion. It cause excessive CPU usage.
>>>> So, it calculates the SPI transfer time and made it sleep.
>>>>
>>>> Signed-off-by: Jaewon Kim <jaewon02.kim@...sung.com>
>>>> ---
>>>>    drivers/spi/spi-s3c64xx.c | 8 ++++++++
>>>>    1 file changed, 8 insertions(+)
>>>>
>>>> diff --git a/drivers/spi/spi-s3c64xx.c b/drivers/spi/spi-s3c64xx.c
>>>> index 886722fb40ea..cf3060b2639b 100644
>>>> --- a/drivers/spi/spi-s3c64xx.c
>>>> +++ b/drivers/spi/spi-s3c64xx.c
>>>> @@ -561,6 +561,14 @@ static int s3c64xx_wait_for_pio(struct s3c64xx_spi_driver_data *sdd,
>>>>    	u32 cpy_len;
>>>>    	u8 *buf;
>>>>    	int ms;
>>>> +	u32 tx_time;
>>>> +
>>>> +	/* sleep during signal transfer time */
>>>> +	status = readl(regs + S3C64XX_SPI_STATUS);
>>>> +	if (RX_FIFO_LVL(status, sdd) < xfer->len) {
>>>> +		tx_time = (xfer->len * 8 * 1000 * 1000) / sdd->cur_speed;
>>>> +		usleep_range(tx_time / 2, tx_time);
>>>> +	}
>>> Did you actually check the delays introduced by it? Is it worth?
>> Yes, I already test it.
>>
>> Throughput was the same, CPU utilization decreased to 30~40% from 100%.
>>
>> Tested board is ExynosAutov9 SADK.
>>
>>
>>>>    
>>>>    	/* millisecs to xfer 'len' bytes @ 'cur_speed' */
>>>>    	ms = xfer->len * 8 * 1000 / sdd->cur_speed;
>>> You have now some code duplication so this could be combined.
> you could put the 'if' under the 'ms = ...' and just use ms
> without declaring any tx_time.
>
> Andi


The unit of 'tx_time' is 'us'.


tx_time = (xfer->len * 8 * 1000 * 1000) / sdd->cur_speed;

ms = xfer->len * 8 * 1000 / sdd->cur_speed;


I add tx_time to minimize existing code modifications.

If we are not using tx_time, we need to change ms to us and change the 
related code.


Thanks

Jaewon Kim