Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <499f0940-3d7f-0ed5-e469-5c0979e6f761@baylibre.com>
Date:   Wed, 14 Sep 2022 14:35:36 +0200
MIME-Version: 1.0
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:91.0) Gecko/20100101
 Thunderbird/91.11.0
Subject: Re: [PATCH] spi: meson-spicc: add support for DMA
Content-Language: en-US
To:     Arnd Bergmann <arnd@arndb.de>, Mark Brown <broonie@kernel.org>
Cc:     linux-spi@vger.kernel.org, linux-amlogic@lists.infradead.org,
        linux-arm-kernel@lists.infradead.org, linux-kernel@vger.kernel.org
References: <20220913140303.437994-1-narmstrong@baylibre.com>
 <667da463-1edc-4a99-9ac1-85303d9187c6@www.fastmail.com>
From:   Neil Armstrong <narmstrong@baylibre.com>
Organization: Baylibre
In-Reply-To: <667da463-1edc-4a99-9ac1-85303d9187c6@www.fastmail.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 7bit
Precedence: bulk

On 14/09/2022 13:19, Arnd Bergmann wrote:
> On Tue, Sep 13, 2022, at 4:03 PM, Neil Armstrong wrote:
> 
>> +static void meson_spicc_setup_dma_burst(struct meson_spicc_device *spicc)
>> +{
> ...
>> +	/* Setup burst length */
>> +	writel_relaxed(ld_ctr1, spicc->base + SPICC_LD_CNTL1);
>> +
>> +	writel_relaxed(SPICC_DMA_ENABLE | SPICC_DMA_URGENT |
>> +		       FIELD_PREP(SPICC_TXFIFO_THRESHOLD_MASK, txfifo_thres) |
>> +		       FIELD_PREP(SPICC_READ_BURST_MASK, read_req) |
>> +		       FIELD_PREP(SPICC_RXFIFO_THRESHOLD_MASK, rxfifo_thres) |
>> +		       FIELD_PREP(SPICC_WRITE_BURST_MASK, write_req),
>> +		       spicc->base + SPICC_DMAREG);
>> +}
> 
> It looks like this last writel_relaxed() starts the DMA, but I don't
> see any barrier that serializes it against the memory access, which
> could still be in a store buffer.

You're right this one restarts a DMA, so yeah it should not have the relaxed accessor.

But the one starting the first DMA is in another place and aswell should not use relaxed.

> 
>> + /* Sometimes, TC gets triggered while the RX fifo isn't fully flushed *
>> + if (spicc->using_dma) {
>> +          unsigned int rxfifo_count = FIELD_GET(SPICC_RXCNT_MASK,
>> +                       readl_relaxed(spicc->base + SPICC_TESTREG));
> 
> Same here in the interrupt controller, I don't see anything enforcing
> the DMA to actually complete before the readl_relaxed().

I don't see the relathionship between a register relaxed read and the DMA not finishing
writing the data in uncached memory, for me it's 2 unrelated things.

> 
> At the very minimum, I think these two have to use non-relaxed
> accessors when adding DMA support, but an easier approach would
> be to use those consistently throughout the driver.

I'll do a check of those accessors for v2.

> 
>       Arnd
Thanks,
Neil