[PATCH] PCI/AER: Block runtime suspend when handling errors

[PATCH] PCI/AER: Block runtime suspend when handling errors

[Stanislaw Gruszka]

PM runtime can be done simultaneously with AER error handling.
Avoid that by using pm_runtime_get_sync() before and pm_runtime_put()
after reset in pcie_do_recovery() for all recovering devices.

pm_runtime_get_sync() will increase dev->power.usage_count counter
to prevent any possible future request to runtime suspend a device,
as well as resume device is was in D3hot state.

I tested with igc device by doing simultaneous aer_inject and
rpm suspend/resume via /sys/bus/pci/devices/PCI_ID/power/control
and can reproduce:

igc 0000:02:00.0: not ready 65535ms after bus reset; giving up
pcieport 0000:00:1c.2: AER: Root Port link has been reset (-25)
pcieport 0000:00:1c.2: AER: subordinate device reset failed
pcieport 0000:00:1c.2: AER: device recovery failed
igc 0000:02:00.0: Unable to change power state from D3hot to D0, device inaccessible

The problem disappears when applied this patch.

Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
---
RFC -> v1:
 add runtime callbacks to pcie_do_recovery(), this covers DPC case
 as well as case of recovering multiple devices under same port.
 
 drivers/pci/pcie/err.c | 20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/drivers/pci/pcie/err.c b/drivers/pci/pcie/err.c
index 59c90d04a609..705893b5f7b0 100644
--- a/drivers/pci/pcie/err.c
+++ b/drivers/pci/pcie/err.c
@@ -13,6 +13,7 @@
 #define dev_fmt(fmt) "AER: " fmt
 
 #include <linux/pci.h>
+#include <linux/pm_runtime.h>
 #include <linux/module.h>
 #include <linux/kernel.h>
 #include <linux/errno.h>
@@ -85,6 +86,18 @@ static int report_error_detected(struct pci_dev *dev,
 	return 0;
 }
 
+static int pci_pm_runtime_get_sync(struct pci_dev *pdev, void *data)
+{
+	pm_runtime_get_sync(&pdev->dev);
+	return 0;
+}
+
+static int pci_pm_runtime_put(struct pci_dev *pdev, void *data)
+{
+	pm_runtime_put(&pdev->dev);
+	return 0;
+}
+
 static int report_frozen_detected(struct pci_dev *dev, void *data)
 {
 	return report_error_detected(dev, pci_channel_io_frozen, data);
@@ -207,6 +220,8 @@ pci_ers_result_t pcie_do_recovery(struct pci_dev *dev,
 	else
 		bridge = pci_upstream_bridge(dev);
 
+	pci_walk_bridge(bridge, pci_pm_runtime_get_sync, NULL);
+
 	pci_dbg(bridge, "broadcast error_detected message\n");
 	if (state == pci_channel_io_frozen) {
 		pci_walk_bridge(bridge, report_frozen_detected, &status);
@@ -251,10 +266,15 @@ pci_ers_result_t pcie_do_recovery(struct pci_dev *dev,
 		pcie_clear_device_status(dev);
 		pci_aer_clear_nonfatal_status(dev);
 	}
+
+	pci_walk_bridge(bridge, pci_pm_runtime_put, NULL);
+
 	pci_info(bridge, "device recovery successful\n");
 	return status;
 
 failed:
+	pci_walk_bridge(bridge, pci_pm_runtime_put, NULL);
+
 	pci_uevent_ers(bridge, PCI_ERS_RESULT_DISCONNECT);
 
 	/* TODO: Should kernel panic here? */
-- 
2.34.1

Re: [PATCH] PCI/AER: Block runtime suspend when handling errors

[Rafael J. Wysocki]

On Fri, Feb 9, 2024 at 3:09 PM Stanislaw Gruszka
<stanislaw.gruszka@linux.intel.com> wrote:
>
> PM runtime can be done simultaneously with AER error handling.
> Avoid that by using pm_runtime_get_sync() before and pm_runtime_put()
> after reset in pcie_do_recovery() for all recovering devices.
>
> pm_runtime_get_sync() will increase dev->power.usage_count counter
> to prevent any possible future request to runtime suspend a device,
> as well as resume device is was in D3hot state.
>
> I tested with igc device by doing simultaneous aer_inject and
> rpm suspend/resume via /sys/bus/pci/devices/PCI_ID/power/control
> and can reproduce:
>
> igc 0000:02:00.0: not ready 65535ms after bus reset; giving up
> pcieport 0000:00:1c.2: AER: Root Port link has been reset (-25)
> pcieport 0000:00:1c.2: AER: subordinate device reset failed
> pcieport 0000:00:1c.2: AER: device recovery failed
> igc 0000:02:00.0: Unable to change power state from D3hot to D0, device inaccessible
>
> The problem disappears when applied this patch.
>
> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>

Looks reasonable to me:

Acked-by: Rafael J. Wysocki <rafael@kernel.org>

> ---
> RFC -> v1:
>  add runtime callbacks to pcie_do_recovery(), this covers DPC case
>  as well as case of recovering multiple devices under same port.
>
>  drivers/pci/pcie/err.c | 20 ++++++++++++++++++++
>  1 file changed, 20 insertions(+)
>
> diff --git a/drivers/pci/pcie/err.c b/drivers/pci/pcie/err.c
> index 59c90d04a609..705893b5f7b0 100644
> --- a/drivers/pci/pcie/err.c
> +++ b/drivers/pci/pcie/err.c
> @@ -13,6 +13,7 @@
>  #define dev_fmt(fmt) "AER: " fmt
>
>  #include <linux/pci.h>
> +#include <linux/pm_runtime.h>
>  #include <linux/module.h>
>  #include <linux/kernel.h>
>  #include <linux/errno.h>
> @@ -85,6 +86,18 @@ static int report_error_detected(struct pci_dev *dev,
>         return 0;
>  }
>
> +static int pci_pm_runtime_get_sync(struct pci_dev *pdev, void *data)
> +{
> +       pm_runtime_get_sync(&pdev->dev);
> +       return 0;
> +}
> +
> +static int pci_pm_runtime_put(struct pci_dev *pdev, void *data)
> +{
> +       pm_runtime_put(&pdev->dev);
> +       return 0;
> +}
> +
>  static int report_frozen_detected(struct pci_dev *dev, void *data)
>  {
>         return report_error_detected(dev, pci_channel_io_frozen, data);
> @@ -207,6 +220,8 @@ pci_ers_result_t pcie_do_recovery(struct pci_dev *dev,
>         else
>                 bridge = pci_upstream_bridge(dev);
>
> +       pci_walk_bridge(bridge, pci_pm_runtime_get_sync, NULL);
> +
>         pci_dbg(bridge, "broadcast error_detected message\n");
>         if (state == pci_channel_io_frozen) {
>                 pci_walk_bridge(bridge, report_frozen_detected, &status);
> @@ -251,10 +266,15 @@ pci_ers_result_t pcie_do_recovery(struct pci_dev *dev,
>                 pcie_clear_device_status(dev);
>                 pci_aer_clear_nonfatal_status(dev);
>         }
> +
> +       pci_walk_bridge(bridge, pci_pm_runtime_put, NULL);
> +
>         pci_info(bridge, "device recovery successful\n");
>         return status;
>
>  failed:
> +       pci_walk_bridge(bridge, pci_pm_runtime_put, NULL);
> +
>         pci_uevent_ers(bridge, PCI_ERS_RESULT_DISCONNECT);
>
>         /* TODO: Should kernel panic here? */
> --
> 2.34.1
>

Re: [PATCH] PCI/AER: Block runtime suspend when handling errors

[Kuppuswamy Sathyanarayanan]

On 2/9/24 6:08 AM, Stanislaw Gruszka wrote:
> PM runtime can be done simultaneously with AER error handling.
> Avoid that by using pm_runtime_get_sync() before and pm_runtime_put()
> after reset in pcie_do_recovery() for all recovering devices.
>
> pm_runtime_get_sync() will increase dev->power.usage_count counter
> to prevent any possible future request to runtime suspend a device,
> as well as resume device is was in D3hot state.
runtime suspend a device or resume a device that was in D3hot state.
>
> I tested with igc device by doing simultaneous aer_inject and
> rpm suspend/resume via /sys/bus/pci/devices/PCI_ID/power/control
> and can reproduce:
>
> igc 0000:02:00.0: not ready 65535ms after bus reset; giving up
> pcieport 0000:00:1c.2: AER: Root Port link has been reset (-25)
> pcieport 0000:00:1c.2: AER: subordinate device reset failed
> pcieport 0000:00:1c.2: AER: device recovery failed
> igc 0000:02:00.0: Unable to change power state from D3hot to D0, device inaccessible
>
> The problem disappears when applied this patch.
>
> Signed-off-by: Stanislaw Gruszka <stanislaw.gruszka@linux.intel.com>
> ---

since it is a bug fix, may be Cc: stable@vger.kernel.org

Reviewed-by: Kuppuswamy Sathyanarayanan <sathyanarayanan.kuppuswamy@linux.intel.com>
> RFC -> v1:
>  add runtime callbacks to pcie_do_recovery(), this covers DPC case
>  as well as case of recovering multiple devices under same port.
>  
>  drivers/pci/pcie/err.c | 20 ++++++++++++++++++++
>  1 file changed, 20 insertions(+)
>
> diff --git a/drivers/pci/pcie/err.c b/drivers/pci/pcie/err.c
> index 59c90d04a609..705893b5f7b0 100644
> --- a/drivers/pci/pcie/err.c
> +++ b/drivers/pci/pcie/err.c
> @@ -13,6 +13,7 @@
>  #define dev_fmt(fmt) "AER: " fmt
>  
>  #include <linux/pci.h>
> +#include <linux/pm_runtime.h>
>  #include <linux/module.h>
>  #include <linux/kernel.h>
>  #include <linux/errno.h>
> @@ -85,6 +86,18 @@ static int report_error_detected(struct pci_dev *dev,
>  	return 0;
>  }
>  
> +static int pci_pm_runtime_get_sync(struct pci_dev *pdev, void *data)
> +{
> +	pm_runtime_get_sync(&pdev->dev);
> +	return 0;
> +}
> +
> +static int pci_pm_runtime_put(struct pci_dev *pdev, void *data)
> +{
> +	pm_runtime_put(&pdev->dev);
> +	return 0;
> +}
> +
>  static int report_frozen_detected(struct pci_dev *dev, void *data)
>  {
>  	return report_error_detected(dev, pci_channel_io_frozen, data);
> @@ -207,6 +220,8 @@ pci_ers_result_t pcie_do_recovery(struct pci_dev *dev,
>  	else
>  		bridge = pci_upstream_bridge(dev);
>  
> +	pci_walk_bridge(bridge, pci_pm_runtime_get_sync, NULL);
> +
>  	pci_dbg(bridge, "broadcast error_detected message\n");
>  	if (state == pci_channel_io_frozen) {
>  		pci_walk_bridge(bridge, report_frozen_detected, &status);
> @@ -251,10 +266,15 @@ pci_ers_result_t pcie_do_recovery(struct pci_dev *dev,
>  		pcie_clear_device_status(dev);
>  		pci_aer_clear_nonfatal_status(dev);
>  	}
> +
> +	pci_walk_bridge(bridge, pci_pm_runtime_put, NULL);
> +
>  	pci_info(bridge, "device recovery successful\n");
>  	return status;
>  
>  failed:
> +	pci_walk_bridge(bridge, pci_pm_runtime_put, NULL);
> +
>  	pci_uevent_ers(bridge, PCI_ERS_RESULT_DISCONNECT);
>  
>  	/* TODO: Should kernel panic here? */

-- 
Sathyanarayanan Kuppuswamy
Linux Kernel Developer

Re: [PATCH] PCI/AER: Block runtime suspend when handling errors

[Stanislaw Gruszka]

On Fri, Feb 09, 2024 at 07:45:05AM -0800, Kuppuswamy Sathyanarayanan wrote:
> 
> On 2/9/24 6:08 AM, Stanislaw Gruszka wrote:
> > PM runtime can be done simultaneously with AER error handling.
> > Avoid that by using pm_runtime_get_sync() before and pm_runtime_put()
> > after reset in pcie_do_recovery() for all recovering devices.
> >
> > pm_runtime_get_sync() will increase dev->power.usage_count counter
> > to prevent any possible future request to runtime suspend a device,
> > as well as resume device is was in D3hot state.
> runtime suspend a device or resume a device that was in D3hot state.

I think "or" is not proper here, since both: resume and prevention
of suspend are done. I'll reword this way:

pm_runtime_get_sync() will increase dev->power.usage_count counter
to prevent any possible future request to runtime suspend a device
It will also resume a device, if it was previously in D3hot state. 

Hope that's clearer.

Thanks
Stanislaw