Re: [PATCH v2 04/12] PCI: brcmstb: Use swinit reset if available

[Stanimir Varbanov <svarbanov@suse.de>]

Hi Philipp,

On 7/8/24 12:37, Philipp Zabel wrote:
> On Fr, 2024-07-05 at 13:46 -0400, Jim Quinlan wrote:
>> On Thu, Jul 4, 2024 at 8:56 AM Stanimir Varbanov <svarbanov@suse.de> wrote:
>>>
>>> Hi Jim,
>>>
>>> On 7/3/24 21:02, Jim Quinlan wrote:
>>>> The 7712 SOC adds a software init reset device for the PCIe HW.
>>>> If found in the DT node, use it.
>>>>
>>>> Signed-off-by: Jim Quinlan <james.quinlan@broadcom.com>
>>>> ---
>>>>  drivers/pci/controller/pcie-brcmstb.c | 19 +++++++++++++++++++
>>>>  1 file changed, 19 insertions(+)
>>>>
>>>> diff --git a/drivers/pci/controller/pcie-brcmstb.c b/drivers/pci/controller/pcie-brcmstb.c
>>>> index 4104c3668fdb..69926ee5c961 100644
>>>> --- a/drivers/pci/controller/pcie-brcmstb.c
>>>> +++ b/drivers/pci/controller/pcie-brcmstb.c
>>>> @@ -266,6 +266,7 @@ struct brcm_pcie {
>>>>       struct reset_control    *rescal;
>>>>       struct reset_control    *perst_reset;
>>>>       struct reset_control    *bridge;
>>>> +     struct reset_control    *swinit;
>>>>       int                     num_memc;
>>>>       u64                     memc_size[PCIE_BRCM_MAX_MEMC];
>>>>       u32                     hw_rev;
>>>> @@ -1626,6 +1627,13 @@ static int brcm_pcie_probe(struct platform_device *pdev)
>>>>               dev_err(&pdev->dev, "could not enable clock\n");
>>>>               return ret;
>>>>       }
>>>> +
>>>> +     pcie->swinit = devm_reset_control_get_optional_exclusive(&pdev->dev, "swinit");
>>>> +     if (IS_ERR(pcie->swinit)) {
>>>> +             ret = dev_err_probe(&pdev->dev, PTR_ERR(pcie->swinit),
>>>> +                                 "failed to get 'swinit' reset\n");
>>>> +             goto clk_out;
>>>> +     }
>>>>       pcie->rescal = devm_reset_control_get_optional_shared(&pdev->dev, "rescal");
>>>>       if (IS_ERR(pcie->rescal)) {
>>>>               ret = PTR_ERR(pcie->rescal);
>>>> @@ -1637,6 +1645,17 @@ static int brcm_pcie_probe(struct platform_device *pdev)
>>>>               goto clk_out;
>>>>       }
>>>>
>>>> +     ret = reset_control_assert(pcie->swinit);
>>>> +     if (ret) {
>>>> +             dev_err_probe(&pdev->dev, ret, "could not assert reset 'swinit'\n");
>>>> +             goto clk_out;
>>>> +     }
>>>> +     ret = reset_control_deassert(pcie->swinit);
>>>> +     if (ret) {
>>>> +             dev_err(&pdev->dev, "could not de-assert reset 'swinit' after asserting\n");
>>>> +             goto clk_out;
>>>> +     }
>>>
>>> why not call reset_control_reset(pcie->swinit) directly?
>> Hi Stan,
>>
>> There is no reset_control_reset() method defined for reset-brcmstb.c.
>> The only reason I can
>> think of for this is that it allows the callers of assert/deassert to
>> insert a delay if desired.
> 
> The main reason for the existence of reset_control_reset() is that
> there are reset controllers that can only be triggered (e.g. by writing
> a bit to a self-clearing register) to produce a complete reset pulse,
> with assertion, delay, and deassertion all handled by the reset
> controller.


Got it. Thank you for explanation.

But, IMO that means that the consumer driver should have knowledge of
low-level reset implementation, which is not generic API?

Otherwise, I don't see a problem to implement asset/deassert sequence in
.reset op in this particular reset-brcmstb.c low-level driver.


~Stan