Subject: Re: [PATCH 2/2] mtd: spi-nor: add driver for STM32 quad spi flash
 controller
To: Ludovic BARRE <ludovic.barre@st.com>,
        Cyrille Pitchen <cyrille.pitchen@atmel.com>
References: <1490619296-8168-1-git-send-email-ludovic.Barre@st.com>
 <1490619296-8168-3-git-send-email-ludovic.Barre@st.com>
 <1be39452-83b0-5c32-39fe-d6dd5134d1ef@gmail.com>
 <2c364b99-512c-8eb6-7044-7989ba21d53b@st.com>
Cc: David Woodhouse <dwmw2@infradead.org>,
        Brian Norris <computersforpeace@gmail.com>,
        Boris Brezillon <boris.brezillon@free-electrons.com>,
        Richard Weinberger <richard@nod.at>,
        Alexandre Torgue <alexandre.torgue@st.com>,
        Rob Herring <robh+dt@kernel.org>, linux-mtd@lists.infradead.org,
        linux-kernel@vger.kernel.org, devicetree@vger.kernel.org
From: Marek Vasut <marek.vasut@gmail.com>
Message-ID: <28454969-0f56-7752-b087-5e02a1a20c23@gmail.com>
Date: Wed, 29 Mar 2017 15:09:30 +0200
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:45.0) Gecko/20100101
 Icedove/45.4.0
MIME-Version: 1.0
In-Reply-To: <2c364b99-512c-8eb6-7044-7989ba21d53b@st.com>
Content-Type: text/plain; charset=utf-8
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 10222
Lines: 333

On 03/29/2017 02:24 PM, Ludovic BARRE wrote:
> hi Marek

Hi!

> thanks for review
> comment below
> 
> On 03/29/2017 12:54 PM, Marek Vasut wrote:
>> On 03/27/2017 02:54 PM, Ludovic Barre wrote:
>>> From: Ludovic Barre <ludovic.barre@st.com>
>>>
>>> The quadspi is a specialized communication interface targeting single,
>>> dual or quad SPI Flash memories.
>>>
>>> It can operate in any of the following modes:
>>> -indirect mode: all the operations are performed using the quadspi
>>>   registers
>>> -read memory-mapped mode: the external Flash memory is mapped to the
>>>   microcontroller address space and is seen by the system as if it was
>>>   an internal memory
>>>
>>> Signed-off-by: Ludovic Barre <ludovic.barre@st.com>
>> Hi! I presume this is not compatible with the Cadence QSPI or any other
>> QSPI controller, huh ?
> it's not a cadence QSPI, it's specific for stm32 platform

OK, got it. Didn't ST use Cadence IP somewhere too though ?
That's probably what confused me, oh well ...

>> Mostly minor nits below ...
>>
>>> ---
>>>   drivers/mtd/spi-nor/Kconfig         |   7 +
>>>   drivers/mtd/spi-nor/Makefile        |   1 +
>>>   drivers/mtd/spi-nor/stm32-quadspi.c | 679
>>> ++++++++++++++++++++++++++++++++++++
>>>   3 files changed, 687 insertions(+)
>>>   create mode 100644 drivers/mtd/spi-nor/stm32-quadspi.c
>> [...]
>>
>>> +struct stm32_qspi_cmd {
>>> +    struct {
>>> +        u32 addr_width:8;
>>> +        u32 dummy:8;
>>> +        u32 data:1;
>>> +    } conf;
>> This could all be u8 ? Why this awful construct ?
> yes I could replace each u32 by u8

Yeah, please do .

>>> +    u8 opcode;
>>> +    u32 framemode;
>>> +    u32 qspimode;
>>> +    u32 addr;
>>> +    size_t len;
>>> +    void *buf;
>>> +};
>>> +
>>> +static int stm32_qspi_wait_cmd(struct stm32_qspi *qspi)
>>> +{
>>> +    u32 cr;
>>> +    int err = 0;
>> Can readl_poll_timeout() or somesuch replace this function ?
> in fact stm32_qspi_wait_cmd test if the transfer has been complete (TCF
> flag)
> else initialize an interrupt completion.
> like the time may be long I prefer keep the interrupt wait, if you agree.

I don't really mind if it fits the hardware better and I agree with you
it does here.

>>> +    if (readl_relaxed(qspi->io_base + QUADSPI_SR) & SR_TCF)
>>> +        return 0;
>>> +
>>> +    reinit_completion(&qspi->cmd_completion);
>>> +    cr = readl_relaxed(qspi->io_base + QUADSPI_CR);
>>> +    writel_relaxed(cr | CR_TCIE, qspi->io_base + QUADSPI_CR);
>>> +
>>> +    if
>>> (!wait_for_completion_interruptible_timeout(&qspi->cmd_completion,
>>> +                               msecs_to_jiffies(1000)))
>>> +        err = -ETIMEDOUT;
>>> +
>>> +    writel_relaxed(cr, qspi->io_base + QUADSPI_CR);
>>> +    return err;
>>> +}
>>> +
>>> +static int stm32_qspi_wait_nobusy(struct stm32_qspi *qspi)
>>> +{
>>> +    u32 sr;
>>> +
>>> +    return readl_relaxed_poll_timeout(qspi->io_base + QUADSPI_SR, sr,
>>> +                      !(sr & SR_BUSY), 10, 100000);
>>> +}
>>> +
>>> +static void stm32_qspi_set_framemode(struct spi_nor *nor,
>>> +                     struct stm32_qspi_cmd *cmd, bool read)
>>> +{
>>> +    u32 dmode = CCR_DMODE_1;
>>> +
>>> +    cmd->framemode = CCR_IMODE_1;
>>> +
>>> +    if (read) {
>>> +        switch (nor->flash_read) {
>>> +        case SPI_NOR_NORMAL:
>>> +        case SPI_NOR_FAST:
>>> +            dmode = CCR_DMODE_1;
>>> +            break;
>>> +        case SPI_NOR_DUAL:
>>> +            dmode = CCR_DMODE_2;
>>> +            break;
>>> +        case SPI_NOR_QUAD:
>>> +            dmode = CCR_DMODE_4;
>>> +            break;
>>> +        }
>>> +    }
>>> +
>>> +    cmd->framemode |= cmd->conf.data ? dmode : 0;
>>> +    cmd->framemode |= cmd->conf.addr_width ? CCR_ADMODE_1 : 0;
>>> +}
>>> +
>>> +static void stm32_qspi_read_fifo(u8 *val, void __iomem *addr)
>>> +{
>>> +    *val = readb_relaxed(addr);
>>> +}
>>> +
>>> +static void stm32_qspi_write_fifo(u8 *val, void __iomem *addr)
>>> +{
>>> +    writeb_relaxed(*val, addr);
>>> +}
>>> +
>>> +static int stm32_qspi_tx_poll(struct stm32_qspi *qspi,
>>> +                  const struct stm32_qspi_cmd *cmd)
>>> +{
>>> +    void (*tx_fifo)(u8 *, void __iomem *);
>>> +    u32 len = cmd->len, sr;
>>> +    u8 *buf = cmd->buf;
>>> +    int ret;
>>> +
>>> +    if (cmd->qspimode == CCR_FMODE_INDW)
>>> +        tx_fifo = stm32_qspi_write_fifo;
>>> +    else
>>> +        tx_fifo = stm32_qspi_read_fifo;
>>> +
>>> +    while (len--) {
>>> +        ret = readl_relaxed_poll_timeout(qspi->io_base + QUADSPI_SR,
>>> +                         sr, (sr & SR_FTF),
>>> +                         10, 30000);
>> Add macros for those ad-hoc timeouts.
> I will add 2 defines (for stm32_qspi_wait_nobusy, stm32_qspi_tx_poll)
> #define STM32_QSPI_FIFO_TIMEOUT_US 30000
> #define STM32_QSPI_BUSY_TIMEOUT_US 100000

Super

>>> +        if (ret) {
>>> +            dev_err(qspi->dev, "fifo timeout (stat:%#x)\n", sr);
>>> +            break;
>>> +        }
>>> +        tx_fifo(buf++, qspi->io_base + QUADSPI_DR);
>>> +    }
>>> +
>>> +    return ret;
>>> +}
>>
>> [...]
>>
>>> +static int stm32_qspi_send(struct stm32_qspi_flash *flash,
>>> +               const struct stm32_qspi_cmd *cmd)
>>> +{
>>> +    struct stm32_qspi *qspi = flash->qspi;
>>> +    u32 ccr, dcr, cr;
>>> +    int err;
>>> +
>>> +    err = stm32_qspi_wait_nobusy(qspi);
>>> +    if (err)
>>> +        goto abort;
>>> +
>>> +    dcr = readl_relaxed(qspi->io_base + QUADSPI_DCR) & ~DCR_FSIZE_MASK;
>>> +    dcr |= DCR_FSIZE(flash->fsize);
>>> +    writel_relaxed(dcr, qspi->io_base + QUADSPI_DCR);
>>> +
>>> +    cr = readl_relaxed(qspi->io_base + QUADSPI_CR);
>>> +    cr &= ~CR_PRESC_MASK & ~CR_FSEL;
>>> +    cr |= CR_PRESC(flash->presc);
>>> +    cr |= flash->cs ? CR_FSEL : 0;
>>> +    writel_relaxed(cr, qspi->io_base + QUADSPI_CR);
>>> +
>>> +    if (cmd->conf.data)
>>> +        writel_relaxed(cmd->len - 1, qspi->io_base + QUADSPI_DLR);
>>> +
>>> +    ccr = cmd->framemode | cmd->qspimode;
>>> +
>>> +    if (cmd->conf.dummy)
>>> +        ccr |= CCR_DCYC(cmd->conf.dummy);
>>> +
>>> +    if (cmd->conf.addr_width)
>>> +        ccr |= CCR_ADSIZE(cmd->conf.addr_width - 1);
>>> +
>>> +    ccr |= CCR_INST(cmd->opcode);
>>> +    writel_relaxed(ccr, qspi->io_base + QUADSPI_CCR);
>>> +
>>> +    if (cmd->conf.addr_width && cmd->qspimode != CCR_FMODE_MM)
>>> +        writel_relaxed(cmd->addr, qspi->io_base + QUADSPI_AR);
>>> +
>>> +    err = stm32_qspi_tx(qspi, cmd);
>>> +    if (err)
>>> +        goto abort;
>>> +
>>> +    if (cmd->qspimode != CCR_FMODE_MM) {
>>> +        err = stm32_qspi_wait_cmd(qspi);
>>> +        if (err)
>>> +            goto abort;
>>> +        writel_relaxed(FCR_CTCF, qspi->io_base + QUADSPI_FCR);
>>> +    }
>>> +
>>> +    return err;
>>> +
>>> +abort:
>>> +    writel_relaxed(readl_relaxed(qspi->io_base + QUADSPI_CR)
>>> +               | CR_ABORT, qspi->io_base + QUADSPI_CR);
>> Use a helper variable here too.
> ok
>>> +    dev_err(qspi->dev, "%s abort err:%d\n", __func__, err);
>>> +    return err;
>>> +}
>> [...]
>>
>>> +static int stm32_qspi_flash_setup(struct stm32_qspi *qspi,
>>> +                  struct device_node *np)
>>> +{
>>> +    u32 width, flash_read, presc, cs_num, max_rate = 0;
>>> +    struct stm32_qspi_flash *flash;
>>> +    struct mtd_info *mtd;
>>> +    int ret;
>>> +
>>> +    of_property_read_u32(np, "reg", &cs_num);
>>> +    if (cs_num >= STM32_MAX_NORCHIP)
>>> +        return -EINVAL;
>>> +
>>> +    of_property_read_u32(np, "spi-max-frequency", &max_rate);
>>> +    if (!max_rate)
>>> +        return -EINVAL;
>>> +
>>> +    presc = DIV_ROUND_UP(qspi->clk_rate, max_rate) - 1;
>>> +
>>> +    if (of_property_read_u32(np, "spi-rx-bus-width", &width))
>>> +        width = 1;
>>> +
>>> +    if (width == 4)
>>> +        flash_read = SPI_NOR_QUAD;
>>> +    else if (width == 2)
>>> +        flash_read = SPI_NOR_DUAL;
>>> +    else if (width == 1)
>>> +        flash_read = SPI_NOR_NORMAL;
>>> +    else
>>> +        return -EINVAL;
>>> +
>>> +    flash = &qspi->flash[cs_num];
>>> +    flash->qspi = qspi;
>>> +    flash->cs = cs_num;
>>> +    flash->presc = presc;
>>> +
>>> +    flash->nor.dev = qspi->dev;
>>> +    spi_nor_set_flash_node(&flash->nor, np);
>>> +    flash->nor.priv = flash;
>>> +    mtd = &flash->nor.mtd;
>>> +    mtd->priv = &flash->nor;
>>> +
>>> +    flash->nor.read = stm32_qspi_read;
>>> +    flash->nor.write = stm32_qspi_write;
>>> +    flash->nor.erase = stm32_qspi_erase;
>>> +    flash->nor.read_reg = stm32_qspi_read_reg;
>>> +    flash->nor.write_reg = stm32_qspi_write_reg;
>>> +    flash->nor.prepare = stm32_qspi_prep;
>>> +    flash->nor.unprepare = stm32_qspi_unprep;
>>> +
>>> +    writel_relaxed(LPTR_DFT_TIMEOUT, qspi->io_base + QUADSPI_LPTR);
>>> +
>>> +    writel_relaxed(CR_PRESC(presc) | CR_FTHRES(3) | CR_TCEN | CR_SSHIFT
>>> +               | CR_EN, qspi->io_base + QUADSPI_CR);
>>> +
>>> +    /* a minimum fsize must be set to sent the command id */
>>> +    flash->fsize = 25;
>> I don't understand why this is needed and the comment doesn't make
>> sense. Please fix.
> fsize field defines the size of external memory.

What external memory ? Unclear

> Normaly, this field is used only for memory map mode,
> but in fact is check in indirect mode.
> So while flash scan "spi_nor_scan":
> -I can't let 0.
> -I not know yet the size of flash.
> So I fix a temporary value
> 
> I will update my comment

Please do, also please consider that I'm reading the comment and I
barely have any clue about this hardware , so make sure I can understand it.

>>> +    ret = spi_nor_scan(&flash->nor, NULL, flash_read);
>>> +    if (ret) {
>>> +        dev_err(qspi->dev, "device scan failed\n");
>>> +        return ret;
>>> +    }
>>> +
>>> +    flash->fsize = __fls(mtd->size) - 1;
>>> +
>>> +    writel_relaxed(DCR_CSHT(1), qspi->io_base + QUADSPI_DCR);
>>> +
>>> +    ret = mtd_device_register(mtd, NULL, 0);
>>> +    if (ret) {
>>> +        dev_err(qspi->dev, "mtd device parse failed\n");
>>> +        return ret;
>>> +    }
>>> +
>>> +    dev_dbg(qspi->dev, "read mm:%s cs:%d bus:%d\n",
>>> +        qspi->read_mode == CCR_FMODE_MM ? "yes" : "no", cs_num, width);
>>> +
>>> +    return 0;
>>> +}
>> [...]
>>
> 


-- 
Best regards,
Marek Vasut