Re: [PATCH v5 4/4] platform/x86: amd-pmc: Drop CPU QoS workaround

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi,

On 3/17/22 15:14, Mario Limonciello wrote:
> A workaround was previously introduced as part of commit 646f429ec2de
> ("platform/x86: amd-pmc: Set QOS during suspend on CZN w/ timer wakeup")
> to prevent CPUs from going into the deepest state during the execution
> of the `noirq` stage of `amd_pmc`.  As `amd_pmc` has been pushed to the
> last step for suspend on AMD platforms, this workaround is no longer
> necessary.
> 
> Signed-off-by: Mario Limonciello <mario.limonciello@xxxxxxx>
> ---
> changes from v4->v5:
>  * rebase on earlier patches
> changes from v3->v4:
>  * rebase on earlier patches
> changes from v2->v3:
>  * No changes
> changes from v1->v2:
>  * No changes
> 
>  drivers/platform/x86/amd-pmc.c | 32 +++-----------------------------
>  1 file changed, 3 insertions(+), 29 deletions(-)
> 
> diff --git a/drivers/platform/x86/amd-pmc.c b/drivers/platform/x86/amd-pmc.c
> index f36cf125b284..7317993cd91b 100644
> --- a/drivers/platform/x86/amd-pmc.c
> +++ b/drivers/platform/x86/amd-pmc.c
> @@ -21,7 +21,6 @@
>  #include <linux/module.h>
>  #include <linux/pci.h>
>  #include <linux/platform_device.h>
> -#include <linux/pm_qos.h>
>  #include <linux/rtc.h>
>  #include <linux/suspend.h>
>  #include <linux/seq_file.h>
> @@ -96,9 +95,6 @@
>  #define PMC_MSG_DELAY_MIN_US		50
>  #define RESPONSE_REGISTER_LOOP_MAX	20000
>  
> -/* QoS request for letting CPUs in idle states, but not the deepest */
> -#define AMD_PMC_MAX_IDLE_STATE_LATENCY	3
> -
>  #define SOC_SUBSYSTEM_IP_MAX	12
>  #define DELAY_MIN_US		2000
>  #define DELAY_MAX_US		3000
> @@ -153,7 +149,6 @@ struct amd_pmc_dev {
>  	struct device *dev;
>  	struct pci_dev *rdev;
>  	struct mutex lock; /* generic mutex lock */
> -	struct pm_qos_request amd_pmc_pm_qos_req;
>  #if IS_ENABLED(CONFIG_DEBUG_FS)
>  	struct dentry *dbgfs_dir;
>  #endif /* CONFIG_DEBUG_FS */
> @@ -628,14 +623,6 @@ static int amd_pmc_verify_czn_rtc(struct amd_pmc_dev *pdev, u32 *arg)
>  	rc = rtc_alarm_irq_enable(rtc_device, 0);
>  	dev_dbg(pdev->dev, "wakeup timer programmed for %lld seconds\n", duration);
>  
> -	/*
> -	 * Prevent CPUs from getting into deep idle states while sending OS_HINT
> -	 * which is otherwise generally safe to send when at least one of the CPUs
> -	 * is not in deep idle states.
> -	 */
> -	cpu_latency_qos_update_request(&pdev->amd_pmc_pm_qos_req, AMD_PMC_MAX_IDLE_STATE_LATENCY);
> -	wake_up_all_idle_cpus();
> -
>  	return rc;
>  }
>  
> @@ -655,7 +642,7 @@ static void amd_pmc_s2idle_prepare(void)
>  		rc = amd_pmc_verify_czn_rtc(pdev, &arg);
>  		if (rc) {
>  			dev_err(pdev->dev, "failed to set RTC: %d\n", rc);
> -			goto fail;
> +			return;
>  		}
>  	}
>  
> @@ -665,20 +652,13 @@ static void amd_pmc_s2idle_prepare(void)
>  	rc = amd_pmc_send_cmd(pdev, arg, NULL, msg, 0);
>  	if (rc) {
>  		dev_err(pdev->dev, "suspend failed: %d\n", rc);
> -		goto fail;
> +		return;
>  	}
>  
>  	if (enable_stb)
>  		rc = amd_pmc_write_stb(pdev, AMD_PMC_STB_PREDEF);
> -	if (rc) {
> +	if (rc)
>  		dev_err(pdev->dev, "error writing to STB: %d\n", rc);
> -		goto fail;
> -	}

Ok, this entire series has been merged here now:
https://git.kernel.org/pub/scm/linux/kernel/git/pdx86/platform-drivers-x86.git/log/?h=for-next

now lets hope none of the various build tests become unhappy because of this.

Mario, one request, the:

	if (rc)
		dev_err(pdev->dev, "error writing to STB: %d\n", rc);

above looks weird, can you do a follow-up patch moving that to
inside/under the "if (enable_stb)" check ?

Currently it also rechecks the amd_pmc_verify_czn_rtc() when
enable_stb is false, which is weird.

Regards,

Hans





> -	return;
> -fail:
> -	if (pdev->cpu_id == AMD_CPU_ID_CZN)
> -		cpu_latency_qos_update_request(&pdev->amd_pmc_pm_qos_req,
> -						PM_QOS_DEFAULT_VALUE);
>  }
>  
>  static void amd_pmc_s2idle_restore(void)



> @@ -704,11 +684,6 @@ static void amd_pmc_s2idle_restore(void)
>  	if (rc)
>  		dev_err(pdev->dev, "error writing to STB: %d\n", rc);
>  
> -	/* Restore the QoS request back to defaults if it was set */
> -	if (pdev->cpu_id == AMD_CPU_ID_CZN)
> -		cpu_latency_qos_update_request(&pdev->amd_pmc_pm_qos_req,
> -						PM_QOS_DEFAULT_VALUE);
> -
>  	/* Notify on failed entry */
>  	amd_pmc_validate_deepest(pdev);
>  }
> @@ -887,7 +862,6 @@ static int amd_pmc_probe(struct platform_device *pdev)
>  		dev_warn(dev->dev, "failed to register LPS0 sleep handler, expect increased power consumption\n");
>  
>  	amd_pmc_dbgfs_register(dev);
> -	cpu_latency_qos_add_request(&dev->amd_pmc_pm_qos_req, PM_QOS_DEFAULT_VALUE);
>  	return 0;
>  
>  err_pci_dev_put:




[Index of Archives]     [Linux Kernel Development]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]

  Powered by Linux