Re: [PATCH v3 2/2] remoteproc: qcom: Introduce Non-PAS ADSP PIL driver

[Sibi Sankar <sibis@codeaurora.org>]

On 2018-09-24 12:19, Rohit Kumar wrote:
> Thanks Sibi for reviewing.
> 
> 
> On 9/22/2018 1:11 AM, Sibi Sankar wrote:
>> Hi Rohit,
>> 
>> On 2018-09-03 17:22, Rohit kumar wrote:
>>> This adds Non PAS ADSP PIL driver for Qualcomm
>>> Technologies Inc SoCs.
>>> Added initial support for SDM845 with ADSP bootup and
>>> shutdown operation handled from Application Processor
>>> SubSystem(APSS).
>>> 
>>> Signed-off-by: Rohit kumar <rohitkr@codeaurora.org>
>>> ---
>> ....
>>> +    select QCOM_MDT_LOADER
>>> +    select QCOM_Q6V5_COMMON
>>> +    select QCOM_RPROC_COMMON
>>> +    help
>>> +      Say y here to support the Peripherial Image Loader
>> 
>> replace Peripherial/Peripheral.
> 
> sure
>> ....
>>> +#include <linux/clk.h>
>>> +#include <linux/firmware.h>
>>> +#include <linux/interrupt.h>
>>> +#include <linux/kernel.h>
>>> +#include <linux/module.h>
>>> +#include <linux/delay.h>
>>> +#include <linux/io.h>
>>> +#include <linux/reset.h>
>>> +#include <linux/iopoll.h>
>>> +#include <linux/pm_domain.h>
>>> +#include <linux/pm_runtime.h>
>>> +#include <linux/regmap.h>
>>> +#include <linux/mfd/syscon.h>
>>> +#include <linux/of_address.h>
>>> +#include <linux/of_device.h>
>>> +#include <linux/platform_device.h>
>>> +#include <linux/remoteproc.h>
>>> +#include <linux/soc/qcom/mdt_loader.h>
>>> +#include <linux/soc/qcom/smem.h>
>>> +#include <linux/soc/qcom/smem_state.h>
>>> +
>> 
>> The headers should be in alphabetical order.
>> 
> 
> ok, will do
>> ....
>>> +    struct reset_control *pdc_sync_reset;
>>> +    struct reset_control *cc_lpass_restart;
>>> +
>>> +    struct regmap *halt_map;
>>> +    unsigned int  halt_lpass;
>> 
>> remove the double spaces above.
> 
> ok
>> ....
>>> +    if (ret || val)
>>> +        goto reset;
>>> +
>>> +    regmap_write(adsp->halt_map,
>>> +            adsp->halt_lpass + LPASS_HALTREQ_REG, 1);
>>> +
>>> +    /*  Wait for halt ACK from QDSP6 */
>> 
>> Minor nit, please remove the double spaces in the comment above.
> 
> ok
>> 
>>> +    timeout = jiffies + msecs_to_jiffies(ACK_TIMEOUT);
>>> +    for (;;) {
>>> +        ret = regmap_read(adsp->halt_map,
>>> +            adsp->halt_lpass + LPASS_HALTACK_REG, &val);
>>> +        if (ret || val || time_after(jiffies, timeout))
>>> +            break;
>>> +
>>> +        udelay(1000);
>> 
>> According to Documentation/timers/timers-howto.txt please use 
>> usleep_range()
>> when the delay range is between(10us - 20ms).
> 
> okay, will update in next spin.
>> 
>>> +    }
>>> +
>>> +    ret = regmap_read(adsp->halt_map,
>>> +            adsp->halt_lpass + LPASS_MASTER_IDLE_REG, &val);
>> ....
>>> +    /* wait after asserting subsystem restart from AOSS */
>>> +    udelay(200);
>> 
>> ditto
> 
> ok
>> 
>>> +
>>> +    /* Clear the halt request for the AXIM and AHBM for Q6 */
>>> +    regmap_write(adsp->halt_map, adsp->halt_lpass + 
>>> LPASS_HALTREQ_REG, 0);
>>> +
>>> +    /* De-assert the LPASS PDC Reset */
>>> +    reset_control_deassert(adsp->pdc_sync_reset);
>>> +    /* Remove the LPASS reset */
>>> +    reset_control_deassert(adsp->cc_lpass_restart);
>>> +    /* wait after de-asserting subsystem restart from AOSS */
>>> +    udelay(200);
>> 
>> ditto
>> 
>>> +
>>> +    return 0;
>>> +}
>> ....
>>> +static int adsp_start(struct rproc *rproc)
>>> +{
>>> +    struct qcom_adsp *adsp = (struct qcom_adsp *)rproc->priv;
>>> +    int ret;
>>> +    unsigned int val;
>>> +
>>> +    qcom_q6v5_prepare(&adsp->q6v5);
>>> +
>>> +    ret = clk_prepare_enable(adsp->xo);
>>> +    if (ret)
>>> +        return ret;
>> 
>> please call qcom_q6v5_unprepare on clk_prepare_enable failure to 
>> disable_irqs
>> 
> 
> sure, will do that
>>> +
>>> +    dev_pm_genpd_set_performance_state(adsp->dev, INT_MAX);
>>> +    ret = pm_runtime_get_sync(adsp->dev);
>>> +    if (ret)
>>> +        goto disable_xo_clk;
>>> +
>> ....
>>> +static int adsp_init_reset(struct qcom_adsp *adsp)
>>> +{
>>> +    adsp->pdc_sync_reset = 
>>> devm_reset_control_get_exclusive(adsp->dev,
>>> +            "pdc_sync");
>>> +    if (IS_ERR(adsp->pdc_sync_reset)) {
>>> +        dev_err(adsp->dev, "failed to acquire pdc_sync reset\n");
>>> +        return PTR_ERR(adsp->pdc_sync_reset);
>>> +    }
>> 
>> Bjorn, should we return EPROBE_DEFER here since PDC global reset 
>> controller can be a module?
>> 
> 
> devm_reset_control_get_exclusive itself returns EPROBE_DEFER until PDC
> reset driver is probed.
> return PTR_ERR(adsp->pdc_sync_reset) handles this case.
> 

Thanks for pointing this out, missed this.

>>> +
>>> +    adsp->cc_lpass_restart = 
>>> devm_reset_control_get_exclusive(adsp->dev,
>>> +            "cc_lpass");
>>> +    if (IS_ERR(adsp->cc_lpass_restart)) {
>>> +        dev_err(adsp->dev, "failed to acquire cc_lpass restart\n");
>>> +        return PTR_ERR(adsp->cc_lpass_restart);
>>> +    }
>>> +
>>> +    return 0;
>> ....
>>> +static int adsp_remove(struct platform_device *pdev)
>>> +{
>>> +    struct qcom_adsp *adsp = platform_get_drvdata(pdev);
>>> +
>>> +    rproc_del(adsp->rproc);
>>> +
>>> +    qcom_remove_glink_subdev(adsp->rproc, &adsp->glink_subdev);
>>> +    qcom_remove_sysmon_subdev(adsp->sysmon);
>>> +    qcom_remove_ssr_subdev(adsp->rproc, &adsp->ssr_subdev);
>>> +    rproc_free(adsp->rproc);
>>> +    pm_runtime_disable(adsp->dev);
>>> +
>> 
>> rmmod of the driver resulted in the following kernel panic:
>> having a pm_runtime_disable after rproc_free seems to be the cause of 
>> the kernel panic.
>> Please call pm_runtime_disable before rproc_free.
>> 
> 
> Thanks for pointing out, will update.
>> do_raw_spin_lock+0x28/0x118
>> __raw_spin_lock_irq+0x30/0x3c
>> __pm_runtime_disable+0x28/0xf4
>> adsp_remove+0x4c/0x5c [qcom_adsp_pil]
>> platform_drv_remove+0x28/0x50
>> device_release_driver_internal+0x124/0x1c8
>> driver_detach+0x44/0x80
>> bus_remove_driver+0x78/0x9c
>> driver_unregister+0x34/0x54
>> platform_driver_unregister+0x1c/0x28
>> cleanup_module+0x14/0x6bc [qcom_adsp_pil]
>> SyS_delete_module+0x1c4/0x214
>> 
>>> +    return 0;
>>> +}
>>> +
>>> +static const struct adsp_pil_data adsp_resource_init = {
>>> +    .crash_reason_smem = 423,
>>> +    .firmware_name = "adsp.mdt",
>>> +    .ssr_name = "lpass",
>>> +    .sysmon_name = "adsp",
>>> +    .ssctl_id = 0x14,
>>> +};
>> ....
>>> +module_platform_driver(adsp_pil_driver);
>>> +MODULE_DESCRIPTION("QTi SDM845 ADSP Peripherial Image Loader");
>> 
>> replace QTi/QTI and Peripherial/Peripheral.
>> 
> ok
>> ....
>> Also I see the following warns on stopping the adsp remoteproc, 
>> couldn't root cause it though:
> 
> It should be issue in Q6 drivers. I will check and update q6 drivers.
> Thanks for reporting.
> 
>>  device_del+0x84/0x29c
>>  platform_device_del+0x2c/0x88
>>  platform_device_unregister+0x1c/0x30
>>  of_platform_device_destroy+0x98/0xa8
>>  device_for_each_child+0x54/0xa4
>>  of_platform_depopulate+0x38/0x54
>>  q6asm_remove+0x1c/0x2c
>>  apr_device_remove+0x38/0x70
>>  device_release_driver_internal+0x124/0x1c8
>>  device_release_driver+0x24/0x30
>>  bus_remove_device+0xcc/0xe4
>>  device_del+0x1f8/0x29c
>>  device_unregister+0x1c/0x30
>>  apr_remove_device+0x1c/0x2c
>>  device_for_each_child+0x54/0xa4
>>  apr_remove+0x28/0x34
>>  rpmsg_dev_remove+0x48/0x70
>>  device_release_driver_internal+0x124/0x1c8
>>  device_release_driver+0x24/0x30
>>  bus_remove_device+0xcc/0xe4
>>  device_del+0x1f8/0x29c
>>  device_unregister+0x1c/0x30
>>  qcom_glink_remove_device+0x1c/0x2c
>>  device_for_each_child+0x54/0xa4
>>  qcom_glink_native_remove+0x54/0x15c
>>  qcom_glink_smem_unregister+0x1c/0x30
>>  glink_subdev_stop+0x1c/0x2c [qcom_common]
>>  rproc_stop+0x40/0xc0
>>  rproc_shutdown+0x6c/0xc0
>>  rproc_del+0x28/0xa0
>>  adsp_remove+0x20/0x5c [qcom_adsp_pil]
>>  platform_drv_remove+0x28/0x50
>>  device_release_driver_internal+0x124/0x1c8
>>  driver_detach+0x44/0x80
>>  bus_remove_driver+0x78/0x9c
>>  driver_unregister+0x34/0x54
>>  platform_driver_unregister+0x1c/0x28
>>  cleanup_module+0x14/0x6bc [qcom_adsp_pil]
>>  SyS_delete_module+0x1c4/0x214
>> 
> 
> Thanks,
> Rohit

-- 
-- Sibi Sankar --
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project.