Received-SPF: pass (google.com: domain of linux-kernel+bounces-96829-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45d1:ec00::1 as permitted sender) client-ip=2604:1380:45d1:ec00::1;
Message-ID: <0b5aa4be-c466-4e1b-91c4-45a26f2f67bf@arm.com>
Date: Fri, 8 Mar 2024 10:29:08 +0000
Precedence: bulk
MIME-Version: 1.0
User-Agent: Mozilla Thunderbird
Subject: Re: [PATCH V5 07/11] coresight: catu: Move ACPI support from AMBA
 driver to platform driver
To: Anshuman Khandual <anshuman.khandual@arm.com>,
 linux-arm-kernel@lists.infradead.org
Cc: Lorenzo Pieralisi <lpieralisi@kernel.org>,
 Sudeep Holla <sudeep.holla@arm.com>, Mike Leach <mike.leach@linaro.org>,
 James Clark <james.clark@arm.com>,
 Maxime Coquelin <mcoquelin.stm32@gmail.com>,
 Alexandre Torgue <alexandre.torgue@foss.st.com>, linux-acpi@vger.kernel.org,
 linux-kernel@vger.kernel.org, coresight@lists.linaro.org,
 linux-stm32@st-md-mailman.stormreply.com
References: <20240222082142.3663983-1-anshuman.khandual@arm.com>
 <20240222082142.3663983-8-anshuman.khandual@arm.com>
 <c43fcd3a-9813-4e1f-adb3-25cc32c54438@arm.com>
 <c52865b2-8608-4a47-967a-6cf3e11b197a@arm.com>
 <2bd6ed98-fd58-42ef-8b86-fddac28df5c7@arm.com>
 <a553951a-9432-4336-82a6-32fc338d0c91@arm.com>
Content-Language: en-GB
From: Suzuki K Poulose <suzuki.poulose@arm.com>
In-Reply-To: <a553951a-9432-4336-82a6-32fc338d0c91@arm.com>
Content-Type: text/plain; charset=UTF-8; format=flowed
Content-Transfer-Encoding: 8bit

On 08/03/2024 04:25, Anshuman Khandual wrote:
> 
> 
> On 3/6/24 22:51, Suzuki K Poulose wrote:
>> On 06/03/2024 06:14, Anshuman Khandual wrote:
>>>
>>>
>>> On 3/5/24 23:02, Suzuki K Poulose wrote:
>>>> On 22/02/2024 08:21, Anshuman Khandual wrote:
>>>>> Add support for the catu devices in a new platform driver, which can then
>>>>> be used on ACPI based platforms. This change would now allow runtime power
>>>>> management for ACPI based systems. The driver would try to enable the APB
>>>>> clock if available. But first this renames and then refactors catu_probe()
>>>>> and catu_remove(), making sure it can be used both for platform and AMBA
>>>>> drivers. This also moves pm_runtime_put() from catu_probe() to the callers.
>>>>>
>>>>> Cc: Lorenzo Pieralisi <lpieralisi@kernel.org>
>>>>> Cc: Sudeep Holla <sudeep.holla@arm.com>
>>>>> Cc: Suzuki K Poulose <suzuki.poulose@arm.com>
>>>>> Cc: Mike Leach <mike.leach@linaro.org>
>>>>> Cc: James Clark <james.clark@arm.com>
>>>>> Cc: linux-acpi@vger.kernel.org
>>>>> Cc: linux-arm-kernel@lists.infradead.org
>>>>> Cc: linux-kernel@vger.kernel.org
>>>>> Cc: coresight@lists.linaro.org
>>>>> Acked-by: Sudeep Holla <sudeep.holla@arm.com> # For ACPI related changes
>>>>> Reviewed-by: James Clark <james.clark@arm.com>
>>>>> Signed-off-by: Anshuman Khandual <anshuman.khandual@arm.com>
>>>>> ---
>>>>> Changes in V5:
>>>>>
>>>>> - Updated commit message regarding catu_probe/remove() refactoring and renaming
>>>>>
>>>>>     drivers/acpi/arm64/amba.c                    |   1 -
>>>>>     drivers/hwtracing/coresight/coresight-catu.c | 142 ++++++++++++++++---
>>>>>     drivers/hwtracing/coresight/coresight-catu.h |   1 +
>>>>>     3 files changed, 124 insertions(+), 20 deletions(-)
>>>>>
>>>>> diff --git a/drivers/acpi/arm64/amba.c b/drivers/acpi/arm64/amba.c
>>>>> index afb6afb66967..587061b0fd2f 100644
>>>>> --- a/drivers/acpi/arm64/amba.c
>>>>> +++ b/drivers/acpi/arm64/amba.c
>>>>> @@ -27,7 +27,6 @@ static const struct acpi_device_id amba_id_list[] = {
>>>>>         {"ARMHC503", 0}, /* ARM CoreSight Debug */
>>>>>         {"ARMHC979", 0}, /* ARM CoreSight TPIU */
>>>>>         {"ARMHC97C", 0}, /* ARM CoreSight SoC-400 TMC, SoC-600 ETF/ETB */
>>>>> -    {"ARMHC9CA", 0}, /* ARM CoreSight CATU */
>>>>>         {"", 0},
>>>>>     };
>>>>>     diff --git a/drivers/hwtracing/coresight/coresight-catu.c b/drivers/hwtracing/coresight/coresight-catu.c
>>>>> index 3949ded0d4fa..a3ea46b53898 100644
>>>>> --- a/drivers/hwtracing/coresight/coresight-catu.c
>>>>> +++ b/drivers/hwtracing/coresight/coresight-catu.c
>>>>> @@ -7,6 +7,8 @@
>>>>>      * Author: Suzuki K Poulose <suzuki.poulose@arm.com>
>>>>>      */
>>>>>     +#include <linux/platform_device.h>
>>>>> +#include <linux/acpi.h>
>>>>>     #include <linux/amba/bus.h>
>>>>>     #include <linux/device.h>
>>>>>     #include <linux/dma-mapping.h>
>>>>> @@ -502,28 +504,20 @@ static const struct coresight_ops catu_ops = {
>>>>>         .helper_ops = &catu_helper_ops,
>>>>>     };
>>>>>     -static int catu_probe(struct amba_device *adev, const struct amba_id *id)
>>>>> +static int __catu_probe(struct device *dev, struct resource *res)
>>>>>     {
>>>>>         int ret = 0;
>>>>>         u32 dma_mask;
>>>>> -    struct catu_drvdata *drvdata;
>>>>> +    struct catu_drvdata *drvdata = dev_get_drvdata(dev);
>>>>>         struct coresight_desc catu_desc;
>>>>>         struct coresight_platform_data *pdata = NULL;
>>>>> -    struct device *dev = &adev->dev;
>>>>>         void __iomem *base;
>>>>>           catu_desc.name = coresight_alloc_device_name(&catu_devs, dev);
>>>>>         if (!catu_desc.name)
>>>>>             return -ENOMEM;
>>>>>     -    drvdata = devm_kzalloc(dev, sizeof(*drvdata), GFP_KERNEL);
>>>>> -    if (!drvdata) {
>>>>> -        ret = -ENOMEM;
>>>>> -        goto out;
>>>>> -    }
>>>>> -
>>>>> -    dev_set_drvdata(dev, drvdata);
>>>>> -    base = devm_ioremap_resource(dev, &adev->res);
>>>>> +    base = devm_ioremap_resource(dev, res);
>>>>>         if (IS_ERR(base)) {
>>>>>             ret = PTR_ERR(base);
>>>>>             goto out;
>>>>> @@ -567,19 +561,39 @@ static int catu_probe(struct amba_device *adev, const struct amba_id *id)
>>>>>         drvdata->csdev = coresight_register(&catu_desc);
>>>>>         if (IS_ERR(drvdata->csdev))
>>>>>             ret = PTR_ERR(drvdata->csdev);
>>>>> -    else
>>>>> -        pm_runtime_put(&adev->dev);
>>>>>     out:
>>>>>         return ret;
>>>>>     }
>>>>>     -static void catu_remove(struct amba_device *adev)
>>>>> +static int catu_probe(struct amba_device *adev, const struct amba_id *id)
>>>>>     {
>>>>> -    struct catu_drvdata *drvdata = dev_get_drvdata(&adev->dev);
>>>>> +    struct catu_drvdata *drvdata;
>>>>> +    int ret;
>>>>> +
>>>>> +    drvdata = devm_kzalloc(&adev->dev, sizeof(*drvdata), GFP_KERNEL);
>>>>> +    if (!drvdata)
>>>>> +        return -ENOMEM;
>>>>> +
>>>>> +    amba_set_drvdata(adev, drvdata);
>>>>> +    ret = __catu_probe(&adev->dev, &adev->res);
>>>>> +    if (!ret)
>>>>> +        pm_runtime_put(&adev->dev);
>>>>> +
>>>>> +    return ret;
>>>>> +}
>>>>> +
>>>>> +static void __catu_remove(struct device *dev)
>>>>> +{
>>>>> +    struct catu_drvdata *drvdata = dev_get_drvdata(dev);
>>>>>           coresight_unregister(drvdata->csdev);
>>>>>     }
>>>>>     +static void catu_remove(struct amba_device *adev)
>>>>> +{
>>>>> +    __catu_remove(&adev->dev);
>>>>> +}
>>>>> +
>>>>>     static struct amba_id catu_ids[] = {
>>>>>         CS_AMBA_ID(0x000bb9ee),
>>>>>         {},
>>>>> @@ -598,13 +612,103 @@ static struct amba_driver catu_driver = {
>>>>>         .id_table            = catu_ids,
>>>>>     };
>>>>>     +static int catu_platform_probe(struct platform_device *pdev)
>>>>> +{
>>>>> +    struct resource *res = platform_get_resource(pdev, IORESOURCE_MEM, 0);
>>>>> +    struct catu_drvdata *drvdata;
>>>>> +    int ret = 0;
>>>>> +
>>>>> +    drvdata = devm_kzalloc(&pdev->dev, sizeof(*drvdata), GFP_KERNEL);
>>>>> +    if (!drvdata)
>>>>> +        return -ENOMEM;
>>>>> +
>>>>> +    drvdata->pclk = coresight_get_enable_apb_pclk(&pdev->dev);
>>>>> +    if (IS_ERR(drvdata->pclk))
>>>>> +        return -ENODEV;
>>>>
>>>>
>>>> ---8>---
>>>>
>>>>> +
>>>>> +    if (res) {
>>>>> +        drvdata->base = devm_ioremap_resource(&pdev->dev, res);
>>>>> +        if (IS_ERR(drvdata->base)) {
>>>>> +            clk_put(drvdata->pclk);
>>>>> +            return PTR_ERR(drvdata->base);
>>>>> +        }
>>>>> +    }
>>>>
>>>> ---<8---
>>>>
>>>> The above section seems unncessary as we already try to map the base in __catu_probe ?
>>>
>>> Agreed, though it seems unnecessary, there is a small difference in there. In the platform
>>> driver case i.e catu_platform_probe(), clk_put() is called on platform clock drvdata->pclk
>>> (just enabled earlier) for cases when devm_ioremap_resource() fails.
>>>
>>> To remove this redundancy, let's move devm_ioremap_resource() into it's AMBA caller i.e
>>> catu_probe() thus dropping struct resource argument from __catu_probe(). Similar situation
>>> is present in coresight-cpu-debug driver as well, will fix that.
>>>
>>> But there are some other drivers in the series where coresight_get_enable_apb_pclk() called
>>> on 'drvdata->pclk' and devm_ioremap_resource() is attempted inside the factored __xxx_probe()
>>> function which is common for both AMBA and platform drivers.
>>>
>>> Such drivers are ...
>>>
>>> - tpiu
>>> - tmc
>>> - stm
>>> - replicator
>>>
>>> IMHO it would be better to follow same scheme for all drivers in the series. Please do let
>>> me know which method will be preferred.
>>
>> Lets pass the "res" to the common probe and deal it there.
> 
> Sure, but then will have to deal with drvdata->pclk in the caller, where it was originally
> acquired, when the __xxxx_probe() fails. For example - in case with the debug cpu driver,
> following changes will be required.

Of course, why not. Because that is specific to the platform driver (for 
a reason). The whole point is, push everything that is common across 
AMBA or Platform to a common __xxx_probe()/remove() routine, rather than
duplicating things around.


> 
> diff --git a/drivers/hwtracing/coresight/coresight-cpu-debug.c b/drivers/hwtracing/coresight/coresight-cpu-debug.c
> index 68d077174a6a..2d77f9a17692 100644
> --- a/drivers/hwtracing/coresight/coresight-cpu-debug.c
> +++ b/drivers/hwtracing/coresight/coresight-cpu-debug.c
> @@ -577,11 +577,11 @@ static int __debug_probe(struct device *dev, struct resource *res)
>          }
>   
>          drvdata->dev = dev;
> -
> -       /* Validity for the resource is already checked by the AMBA core */
> -       base = devm_ioremap_resource(dev, res);
> -       if (IS_ERR(base))
> -               return PTR_ERR(base);
> +       if (res) {
> +               base = devm_ioremap_resource(dev, res);
> +               if (IS_ERR(base))
> +                       return PTR_ERR(base);
> +       }
>   
>          drvdata->base = base;
>   
> @@ -703,14 +703,6 @@ static int debug_platform_probe(struct platform_device *pdev)
>          if (IS_ERR(drvdata->pclk))
>                  return -ENODEV;
>   
> -       if (res) {
> -               drvdata->base = devm_ioremap_resource(&pdev->dev, res);
> -               if (IS_ERR(drvdata->base)) {
> -                       clk_put(drvdata->pclk);
> -                       return PTR_ERR(drvdata->base);
> -               }
> -       }
> -
>          dev_set_drvdata(&pdev->dev, drvdata);
>          pm_runtime_get_noresume(&pdev->dev);
>          pm_runtime_set_active(&pdev->dev);
> @@ -720,6 +712,8 @@ static int debug_platform_probe(struct platform_device *pdev)
>          if (ret) {
>                  pm_runtime_put_noidle(&pdev->dev);
>                  pm_runtime_disable(&pdev->dev);
> +               if (drvdata && !IS_ERR_OR_NULL(drvdata->pclk))
> +                       clk_put(drvdata->pclk);
>          }
>          return ret;
> 
>>>
>>>>
>>>>> +
>>>
>>>>> +    pm_runtime_get_noresume(&pdev->dev);
>>>>> +    pm_runtime_set_active(&pdev->dev);
>>>>> +    pm_runtime_enable(&pdev->dev);
>>>>> +
>>>>> +    dev_set_drvdata(&pdev->dev, drvdata);
>>>>> +    ret = __catu_probe(&pdev->dev, res);
>>>>> +    pm_runtime_put(&pdev->dev);
>>>>> +    if (ret)
>>>>> +        pm_runtime_disable(&pdev->dev);
>>>>> +
>>>>> +    return ret;
>>>>> +}
>>>>> +
>>>>> +static int catu_platform_remove(struct platform_device *pdev)
>>>>> +{
>>>>> +    struct catu_drvdata *drvdata = dev_get_drvdata(&pdev->dev);
>>>>> +
>>>>> +    if (drvdata)
>>>>> +        __catu_remove(&pdev->dev);
>>>>
>>>> I don't understand the need for if () check here (and on all the other drivers). Even if we have a drvdata != NULL, what guarantees that
>>>> the drvdata->csdev is valid (which is used in xx_remove) ?
>>>
>>> Agreed, although drvdata is derived in __xxx_remove() functions, a pre-check here is not
>>> required - similar to the AMBA remove path. Sure, will drop them across drivers.
>>
>> No, my point is : Is there a case where :
>>
>> 1) drvdata == NULL, but we have a device to do some cleanup ?
> 
> NO, I don't believe there are any such cases.
> 
>> 2) If drvdata != NULL, but drvdata->csdev is not valid, because we failed to register the coresight device ?
> 
> I don't believe so. If drvdata->csdev is not valid because coresight_register() has failed,
> that would have also failed driver' probe() as well. Hence driver remove() could not have
> been called in such situations.

So, all of those seem to be cases we don't expect. To be on the safer 
side, add a if (WARN_ON(!drvdata)) and return. Deal the rest with
drvdata and drvdata->csdev being valid.

Suzuki


> 
>>
>>
>> Suzuki
>>
>>
>>>
>>>>
>>>> Suzuki
>>>>
>>>>> +
>>>>> +    pm_runtime_disable(&pdev->dev);
>>>>> +    if (drvdata && !IS_ERR_OR_NULL(drvdata->pclk))
>>>>> +        clk_put(drvdata->pclk);
>>>>> +    return 0;
>>>>> +}
>>>>> +
>>>>> +#ifdef CONFIG_PM
>>>>> +static int catu_runtime_suspend(struct device *dev)
>>>>> +{
>>>>> +    struct catu_drvdata *drvdata = dev_get_drvdata(dev);
>>>>> +
>>>>> +    if (drvdata && !IS_ERR_OR_NULL(drvdata->pclk))
>>>>> +        clk_disable_unprepare(drvdata->pclk);
>>>>> +    return 0;
>>>>> +}
>>>>> +
>>>>> +static int catu_runtime_resume(struct device *dev)
>>>>> +{
>>>>> +    struct catu_drvdata *drvdata = dev_get_drvdata(dev);
>>>>> +
>>>>> +    if (drvdata && !IS_ERR_OR_NULL(drvdata->pclk))
>>>>> +        clk_prepare_enable(drvdata->pclk);
>>>>> +    return 0;
>>>>> +}
>>>>> +#endif
>>>>> +
>>>>> +static const struct dev_pm_ops catu_dev_pm_ops = {
>>>>> +    SET_RUNTIME_PM_OPS(catu_runtime_suspend, catu_runtime_resume, NULL)
>>>>> +};
>>>>> +
>>>>> +#ifdef CONFIG_ACPI
>>>>> +static const struct acpi_device_id catu_acpi_ids[] = {
>>>>> +    {"ARMHC9CA", 0}, /* ARM CoreSight CATU */
>>>>> +    {},
>>>>> +};
>>>>> +
>>>>> +MODULE_DEVICE_TABLE(acpi, catu_acpi_ids);
>>>>> +#endif
>>>>> +
>>>>> +static struct platform_driver catu_platform_driver = {
>>>>> +    .probe    = catu_platform_probe,
>>>>> +    .remove    = catu_platform_remove,
>>>>> +    .driver    = {
>>>>> +        .name            = "coresight-catu-platform",
>>>>> +        .acpi_match_table    = ACPI_PTR(catu_acpi_ids),
>>>>> +        .suppress_bind_attrs    = true,
>>>>> +        .pm            = &catu_dev_pm_ops,
>>>>> +    },
>>>>> +};
>>>>> +
>>>>>     static int __init catu_init(void)
>>>>>     {
>>>>>         int ret;
>>>>>     -    ret = amba_driver_register(&catu_driver);
>>>>> -    if (ret)
>>>>> -        pr_info("Error registering catu driver\n");
>>>>> +    ret = coresight_init_driver("catu", &catu_driver, &catu_platform_driver);
>>>>>         tmc_etr_set_catu_ops(&etr_catu_buf_ops);
>>>>>         return ret;
>>>>>     }
>>>>> @@ -612,7 +716,7 @@ static int __init catu_init(void)
>>>>>     static void __exit catu_exit(void)
>>>>>     {
>>>>>         tmc_etr_remove_catu_ops();
>>>>> -    amba_driver_unregister(&catu_driver);
>>>>> +    coresight_remove_driver(&catu_driver, &catu_platform_driver);
>>>>>     }
>>>>>       module_init(catu_init);
>>>>> diff --git a/drivers/hwtracing/coresight/coresight-catu.h b/drivers/hwtracing/coresight/coresight-catu.h
>>>>> index 442e034bbfba..141feac1c14b 100644
>>>>> --- a/drivers/hwtracing/coresight/coresight-catu.h
>>>>> +++ b/drivers/hwtracing/coresight/coresight-catu.h
>>>>> @@ -61,6 +61,7 @@
>>>>>     #define CATU_IRQEN_OFF        0x0
>>>>>       struct catu_drvdata {
>>>>> +    struct clk *pclk;
>>>>>         void __iomem *base;
>>>>>         struct coresight_device *csdev;
>>>>>         int irq;
>>>>
>>