[RFC PATCH v6 1/9] mmc: dw_mmc: Add external dma interface support

Shawn Lin shawn.lin at rock-chips.com
Fri Aug 21 00:27:43 PDT 2015


On 2015/8/21 14:35, Jaehoon Chung wrote:
> On 08/21/2015 03:30 PM, Shawn Lin wrote:
>> On 2015/8/21 14:27, Jaehoon Chung wrote:
>>> Hi, Shawn.
>>>
>>> Is this based on Ulf's repository?
>>
>>
>> no, it's based on "https://github.com/jh80chung/dw-mmc.git tags/dw-mmc-for-ulf-v4.2" :)
>
> Oh..I will rebase to Ulf's next branch on this weekend.
> Then could you rebase this patch? And i added more comments at below.. :)
>

Okay, I will rebase to Ulf's next.

> Best Regards,
> Jaehoon Chung
>
>>

[...]

>>>> index ec6dbcd..7e1d13b 100644
>>>> --- a/drivers/mmc/host/dw_mmc-pltfm.c
>>>> +++ b/drivers/mmc/host/dw_mmc-pltfm.c
>>>> @@ -59,6 +59,8 @@ int dw_mci_pltfm_register(struct platform_device *pdev,
>>>>        host->pdata = pdev->dev.platform_data;
>>>>
>>>>        regs = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>>>> +    /* Get registers' physical base address */
>>>> +    host->phy_regs = (void *)(regs->start);
>>>>        host->regs = devm_ioremap_resource(&pdev->dev, regs);
>>>>        if (IS_ERR(host->regs))
>>>>            return PTR_ERR(host->regs);
>>>
>>> Is this board specific code? If so, separate the patch.

It's might not board specific code.
dmaengine need dw_mmc's *physical* fifo address for data transfer, so I 
get controller physical address  here in order to calculate physical 
fifo address.

regs is from dt-bindings, for instance:
          dwmmc0 at 12200000 {
                  compatible = "snps,dw-mshc";
                   clocks = <&clock 351>, <&clock 132>;
                  clock-names = "biu", "ciu";
                  reg = <0x12200000 0x1000>;
                  interrupts = <0 75 0>;
                  #address-cells = <1>;
                  #size-cells = <0>;
          };

so, host->phy_regs will be 0x12200000 .

[...]

>>>> +static void dw_mci_dmac_complete_dma(void *arg)
>>>>    {
>>>> +    struct dw_mci *host = arg;
>>>>        struct mmc_data *data = host->data;
>>>>
>>>>        dev_vdbg(host->dev, "DMA complete\n");
>>>>
>>>> +    if (host->use_dma == TRANS_MODE_EDMAC)
>>>> +        if (data && (data->flags & MMC_DATA_READ))
>>>
>>> Combine one condition.

okay.

[...]

>>>> +    u32 fifo_offset = host->fifo_reg - host->regs;
>>>> +    int ret = 0;
>>>> +
>>>> +    /* Set external dma config: burst size, burst width */
>>>> +    cfg.dst_addr = (dma_addr_t)(host->phy_regs + fifo_offset);
>>>
>>> host->phy_regs is not assigned?

we got it at dw_mci_pltfm_register. See comments above. :)

[...]

>>>>            mmc->max_blk_count = mmc->max_req_size / 512;
>>>> +    } else if (host->use_dma == TRANS_MODE_EDMAC) {
>>>> +            mmc->max_segs = 64;
>>>> +            mmc->max_blk_size = 65536;
>>>> +            mmc->max_blk_count = 65535;
>>>> +            mmc->max_req_size =
>>>> +                    mmc->max_blk_size * mmc->max_blk_count;
>>>> +            mmc->max_seg_size = mmc->max_req_size;
>>>
>>> Fix the indention

Hmm..I check it attentively but can't find the indention . Might it's 
because you apply it against Ulf's repo?

>>>

[...]

>>>>
>>>> -    /* Alloc memory for sg translation */
>>>> -    host->sg_cpu = dmam_alloc_coherent(host->dev, PAGE_SIZE,
>>>> -                      &host->sg_dma, GFP_KERNEL);
>>>> -    if (!host->sg_cpu) {
>>>> -        dev_err(host->dev, "%s: could not alloc DMA memory\n",
>>>> -            __func__);
>>>> +    /* Check tansfer mode */
>>>> +    trans_mode = SDMMC_GET_TRANS_MODE(mci_readl(host, HCON));
>>>> +    if (trans_mode == 0) {
>>>> +        trans_mode = TRANS_MODE_IDMAC;
>>>> +    } else if (trans_mode == 1 || trans_mode == 2) {
>>>> +        trans_mode = TRANS_MODE_EDMAC;
>>>> +    } else {
>>>> +        trans_mode = TRANS_MODE_PIO;
>>>
>>> what are trans_mode "0, 1, 2"?
>>> (00 - none) is NO-DMA interface, isn't? is it IDMAC mode?
>>>

No, I guess the databook's ambiguous description confuse everybody.

I got double comfirm from my ASCI team as well as Synoposys
2b'00: NO-DMA interface -->  It support IDMAC actually
2b'01 & 2b'02: DW_DMA & GENERIC_DMA --> Support 2 types of external dma.
2b'02: NON-DW-DMA --> only support PIO

refer to the description below:
Parameter Name: DMA_INTERFACE
Legal Values: 0-3
Default Value: 0
Description:
  0- No DMA Interface
  1- DesignWare DMA Interface
  2- Generic DMA Interface
  3- Non DW DMA Interface
In DesignWare DMA mode, request/acknowledge protocol meets DW_ahb_dmac
controller protocol. In this mode, host data bus is also used for DMA 
transfers.Generic DMA-type interface has simpler request/acknowledge 
handshake and has dedicated read/write data bus for DMA transfers. Non 
DW DMAC interface uses dw_dma_single interface in addition to the 
existing interface and uses host data bus for DMA transfers. This is 
configurable only if INTERNAL_DMAC=0.

>>>>            goto no_dma;
>>>>        }
>
>>>> +        dev_info(host->dev, "Using external DMA controller.\n");
>>>> +    }
>>>>
>>>>        if (!host->dma_ops)
>>>>            goto no_dma;
>>>> @@ -2562,12 +2720,12 @@ static void dw_mci_init_dma(struct dw_mci *host)
>>>>            goto no_dma;
>>>>        }
>>>>
>>>> -    host->use_dma = 1;
>>>> +    host->use_dma = trans_mode;
>>>
>>> Also confuse, if trans_mode is assigned host->use_dma, can mode value be directly assigned to host->use_dma?
>>>

Good idea, I will do it.  :)

>>> trans_mode = TRAMS_MODE_PIO;
>>> host->use_dma = trans_mode;
>>>

[...]

>>>>
>>>> @@ -2890,7 +3047,7 @@ int dw_mci_probe(struct dw_mci *host)
>>>>         * Get the host data width - this assumes that HCON has been set with
>>>>         * the correct values.
>>>>         */
>>>> -    i = (mci_readl(host, HCON) >> 7) & 0x7;
>>>> +    i = SDMMC_GET_HDATA_WIDTH(mci_readl(host, HCON));
>>>
>>> This is not related with supporting external dma interface.
>>> Separate this.
>>>

Okay, It will be split into another one.

>>>>        if (!i) {
>>>>            host->push_data = dw_mci_push_data16;
>>>>            host->pull_data = dw_mci_pull_data16;



-- 
Best Regards
Shawn Lin




More information about the Linux-rockchip mailing list