linux-kernel - Re: [PATCH v2 1/3] amd64-agp: convert to generic power management

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [day] [month] [year] [list]

Message-ID: <20211202174348.GA2899481@bhelgaas>
Date:   Thu, 2 Dec 2021 11:43:48 -0600
From:   Bjorn Helgaas <helgaas@...nel.org>
To:     David Airlie <airlied@...ux.ie>,
        "Rafael J . Wysocki" <rafael@...nel.org>
Cc:     Vaibhav Gupta <vaibhav.varodek@...il.com>,
        Vaibhav Gupta <vaibhavgupta40@...il.com>,
        linux-kernel@...r.kernel.org, linux-pci@...r.kernel.org,
        Bjorn Helgaas <bhelgaas@...gle.com>
Subject: Re: [PATCH v2 1/3] amd64-agp: convert to generic power management

[cc->to: Rafael: help :)]

On Tue, Nov 30, 2021 at 08:54:17PM -0600, Bjorn Helgaas wrote:
> From: Vaibhav Gupta <vaibhavgupta40@...il.com>
> 
> Convert amd64-agp from legacy PCI power management to the generic power
> management framework.
> 
> Previously, amd64-agp used legacy PCI power management.
> agp_amd64_suspend() looked like this:
> 
>   agp_amd64_suspend
>     pci_save_state(pdev)
>     pci_set_power_state(pdev, pci_choose_state(pdev, state))
> 
> With generic power management, these are both done by the PCI core in
> pci_pm_runtime_suspend(), so drop agp_amd64_suspend() completely.

I think the *patch* is correct, but my explanation is wrong.  Would
appreciate any corrections!

Prior to this patch, agp_amd64_suspend() is a pci_driver.suspend()
method and is called in this path:

  pci_pm_suspend
    pci_legacy_suspend
      drv->suspend

After this patch, agp_amd64_suspend() is not implemented at all, and
we do the pci_save_state() and pci_set_power_state() in PCI generic
code.

But I think those actually happen in pci_pm_suspend_noirq(), not in
pci_pm_runtime_suspend(), i.e., in this path:

  suspend_devices_and_enter
    dpm_suspend_start(PMSG_SUSPEND)
      dpm_suspend(PMSG_SUSPEND)
        device_suspend
          __device_suspend
            callback = pm_op(dev->bus->pm, state)
            dpm_run_callback(callback)
              pci_pm_suspend                            # PCI bus method
                dev->driver->pm->suspend
                  agp_amd64_suspend                     # <-- no longer needed
    suspend_enter
      dpm_suspend_noirq(PMSG_SUSPEND)
        dpm_noirq_suspend_devices(PMSG_SUSPEND)
          device_suspend_noirq
            __device_suspend_noirq
              callback = pm_noirq_op(dev->bus->pm, state)
              dpm_run_callback(callback)
                pci_pm_suspend_noirq                    # PCI bus method
                  pci_save_state                        # <-- now done here
                  pci_prepare_to_sleep
                    pci_set_power_state                 # <-- and here
                  
I got confused because I couldn't find the call chain leading to
pci_pm_suspend_noirq().  pm_op() and dpm_run_callback() essentially
break the call chain, which makes it a little hard to follow.

> agp_amd64_resume() looked like this:
> 
>   agp_amd64_resume
>     pci_set_power_state(pdev, PCI_D0)
>     pci_restore_state(pdev)
>     ...
> 
> With generic power management, the PCI parts are done by
> pci_pm_runtime_resume(), so drop those from agp_amd64_resume().
> 
> [bhelgaas: commit log]
> Signed-off-by: Vaibhav Gupta <vaibhavgupta40@...il.com>
> Signed-off-by: Bjorn Helgaas <bhelgaas@...gle.com>
> ---
>  drivers/char/agp/amd64-agp.c | 24 ++++++------------------
>  1 file changed, 6 insertions(+), 18 deletions(-)
> 
> diff --git a/drivers/char/agp/amd64-agp.c b/drivers/char/agp/amd64-agp.c
> index b40edae32817..dc78a4fb879e 100644
> --- a/drivers/char/agp/amd64-agp.c
> +++ b/drivers/char/agp/amd64-agp.c
> @@ -588,20 +588,11 @@ static void agp_amd64_remove(struct pci_dev *pdev)
>  	agp_bridges_found--;
>  }
>  
> -#ifdef CONFIG_PM
> +#define agp_amd64_suspend NULL
>  
> -static int agp_amd64_suspend(struct pci_dev *pdev, pm_message_t state)
> +static int __maybe_unused agp_amd64_resume(struct device *dev)
>  {
> -	pci_save_state(pdev);
> -	pci_set_power_state(pdev, pci_choose_state(pdev, state));
> -
> -	return 0;
> -}
> -
> -static int agp_amd64_resume(struct pci_dev *pdev)
> -{
> -	pci_set_power_state(pdev, PCI_D0);
> -	pci_restore_state(pdev);
> +	struct pci_dev *pdev = to_pci_dev(dev);
>  
>  	if (pdev->vendor == PCI_VENDOR_ID_NVIDIA)
>  		nforce3_agp_init(pdev);
> @@ -609,8 +600,6 @@ static int agp_amd64_resume(struct pci_dev *pdev)
>  	return amd_8151_configure();
>  }
>  
> -#endif /* CONFIG_PM */
> -
>  static const struct pci_device_id agp_amd64_pci_table[] = {
>  	{
>  	.class		= (PCI_CLASS_BRIDGE_HOST << 8),
> @@ -738,15 +727,14 @@ static const struct pci_device_id agp_amd64_pci_promisc_table[] = {
>  	{ }
>  };
>  
> +static SIMPLE_DEV_PM_OPS(agp_amd64_pm_ops, agp_amd64_suspend, agp_amd64_resume);
> +
>  static struct pci_driver agp_amd64_pci_driver = {
>  	.name		= "agpgart-amd64",
>  	.id_table	= agp_amd64_pci_table,
>  	.probe		= agp_amd64_probe,
>  	.remove		= agp_amd64_remove,
> -#ifdef CONFIG_PM
> -	.suspend	= agp_amd64_suspend,
> -	.resume		= agp_amd64_resume,
> -#endif
> +	.driver.pm  = &agp_amd64_pm_ops,
>  };
>  
>  
> -- 
> 2.25.1
>