[PATCH] vfio: platform: Fix using devices in PM Domains

[PATCH] vfio: platform: Fix using devices in PM Domains

[Geert Uytterhoeven]

If a device is part of a PM Domain (e.g. power and/or clock domain), its
power state is managed using Runtime PM.  Without Runtime PM, the device
may not be powered up, causing subtle failures, crashes, or system
lock-ups when the device is accessed by the guest.

Fix this by adding Runtime PM support, powering the device when the VFIO
device is opened by the guest.

Note that while more fine-grained power management could be implemented
on the guest side, if exported, this would be inherently unsafe, as
abusing it may kill the whole system.

Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
---
This depends on "[PATCH v3 2/2] vfio: platform: Add generic DT reset
support" due to a small contextual change (addition of "#include
<linux/reset.h>").

 drivers/vfio/platform/vfio_platform_common.c | 9 +++++++++
 1 file changed, 9 insertions(+)

diff --git a/drivers/vfio/platform/vfio_platform_common.c b/drivers/vfio/platform/vfio_platform_common.c
index ef9b9e3220ebe939..4db0a143992c3353 100644
--- a/drivers/vfio/platform/vfio_platform_common.c
+++ b/drivers/vfio/platform/vfio_platform_common.c
@@ -17,6 +17,7 @@
 #include <linux/iommu.h>
 #include <linux/module.h>
 #include <linux/mutex.h>
+#include <linux/pm_runtime.h>
 #include <linux/reset.h>
 #include <linux/slab.h>
 #include <linux/types.h>
@@ -249,6 +250,8 @@ static void vfio_platform_release(void *device_data)
 		const char *extra_dbg = NULL;
 		int ret;
 
+		pm_runtime_put(vdev->device);
+
 		ret = vfio_platform_call_reset(vdev, &extra_dbg);
 		if (ret && vdev->reset_required) {
 			dev_warn(vdev->device, "reset driver is required and reset call failed in release (%d) %s\n",
@@ -291,6 +294,10 @@ static int vfio_platform_open(void *device_data)
 				 ret, extra_dbg ? extra_dbg : "");
 			goto err_rst;
 		}
+
+		ret = pm_runtime_get_sync(vdev->device);
+		if (ret < 0)
+			goto err_rst;
 	}
 
 	vdev->refcnt++;
@@ -706,6 +713,7 @@ int vfio_platform_probe_common(struct vfio_platform_device *vdev,
 
 	mutex_init(&vdev->igate);
 
+	pm_runtime_enable(vdev->device);
 	return 0;
 
 put_iommu:
@@ -723,6 +731,7 @@ struct vfio_platform_device *vfio_platform_remove_common(struct device *dev)
 	vdev = vfio_del_group_dev(dev);
 
 	if (vdev) {
+		pm_runtime_disable(vdev->device);
 		vfio_platform_put_reset(vdev);
 		vfio_iommu_group_put(dev->iommu_group, dev);
 	}
-- 
2.7.4

Re: [PATCH] vfio: platform: Fix using devices in PM Domains

[Simon Horman]

On Wed, Apr 11, 2018 at 11:24:08AM +0200, Geert Uytterhoeven wrote:
> If a device is part of a PM Domain (e.g. power and/or clock domain), its
> power state is managed using Runtime PM.  Without Runtime PM, the device
> may not be powered up, causing subtle failures, crashes, or system
> lock-ups when the device is accessed by the guest.
> 
> Fix this by adding Runtime PM support, powering the device when the VFIO
> device is opened by the guest.
> 
> Note that while more fine-grained power management could be implemented
> on the guest side, if exported, this would be inherently unsafe, as
> abusing it may kill the whole system.
> 
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>

Reviewed-by: Simon Horman <horms+renesas@verge.net.au>

Re: [PATCH] vfio: platform: Fix using devices in PM Domains

[Auger Eric]

Hi Geert,

On 11/04/18 11:24, Geert Uytterhoeven wrote:
> If a device is part of a PM Domain (e.g. power and/or clock domain), its
> power state is managed using Runtime PM.  Without Runtime PM, the device
> may not be powered up, causing subtle failures, crashes, or system
> lock-ups when the device is accessed by the guest.
the device may not be powered up/clcoked or power/clock may be switched
off while the guest uses it.
> 
> Fix this by adding Runtime PM support, powering the device when the VFIO
> device is opened by the guest.
> 
> Note that while more fine-grained power management could be implemented
> on the guest side, if exported, this would be inherently unsafe, as
> abusing it may kill the whole system.
Please can you elaborate on this remark please?

Thanks

Eric
> 
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>
> ---
> This depends on "[PATCH v3 2/2] vfio: platform: Add generic DT reset
> support" due to a small contextual change (addition of "#include
> <linux/reset.h>").
> 
>  drivers/vfio/platform/vfio_platform_common.c | 9 +++++++++
>  1 file changed, 9 insertions(+)
> 
> diff --git a/drivers/vfio/platform/vfio_platform_common.c b/drivers/vfio/platform/vfio_platform_common.c
> index ef9b9e3220ebe939..4db0a143992c3353 100644
> --- a/drivers/vfio/platform/vfio_platform_common.c
> +++ b/drivers/vfio/platform/vfio_platform_common.c
> @@ -17,6 +17,7 @@
>  #include <linux/iommu.h>
>  #include <linux/module.h>
>  #include <linux/mutex.h>
> +#include <linux/pm_runtime.h>
>  #include <linux/reset.h>
>  #include <linux/slab.h>
>  #include <linux/types.h>
> @@ -249,6 +250,8 @@ static void vfio_platform_release(void *device_data)
>  		const char *extra_dbg = NULL;
>  		int ret;
>  
> +		pm_runtime_put(vdev->device);
> +
>  		ret = vfio_platform_call_reset(vdev, &extra_dbg);
>  		if (ret && vdev->reset_required) {
>  			dev_warn(vdev->device, "reset driver is required and reset call failed in release (%d) %s\n",
> @@ -291,6 +294,10 @@ static int vfio_platform_open(void *device_data)
>  				 ret, extra_dbg ? extra_dbg : "");
>  			goto err_rst;
>  		}
> +
> +		ret = pm_runtime_get_sync(vdev->device);
> +		if (ret < 0)
> +			goto err_rst;
>  	}
>  
>  	vdev->refcnt++;
> @@ -706,6 +713,7 @@ int vfio_platform_probe_common(struct vfio_platform_device *vdev,
>  
>  	mutex_init(&vdev->igate);
>  
> +	pm_runtime_enable(vdev->device);
>  	return 0;
>  
>  put_iommu:
> @@ -723,6 +731,7 @@ struct vfio_platform_device *vfio_platform_remove_common(struct device *dev)
>  	vdev = vfio_del_group_dev(dev);
>  
>  	if (vdev) {
> +		pm_runtime_disable(vdev->device);
>  		vfio_platform_put_reset(vdev);
>  		vfio_iommu_group_put(dev->iommu_group, dev);
>  	}
> 

Re: [PATCH] vfio: platform: Fix using devices in PM Domains

[Geert Uytterhoeven]

Hi Eric,

On Fri, Apr 13, 2018 at 11:14 AM, Auger Eric <eric.auger@redhat.com> wrote:
> On 11/04/18 11:24, Geert Uytterhoeven wrote:
>> If a device is part of a PM Domain (e.g. power and/or clock domain), its
>> power state is managed using Runtime PM.  Without Runtime PM, the device
>> may not be powered up, causing subtle failures, crashes, or system
>> lock-ups when the device is accessed by the guest.
> the device may not be powered up/clcoked or power/clock may be switched
> off while the guest uses it.
>>
>> Fix this by adding Runtime PM support, powering the device when the VFIO
>> device is opened by the guest.
>>
>> Note that while more fine-grained power management could be implemented
>> on the guest side, if exported, this would be inherently unsafe, as
>> abusing it may kill the whole system.
>
> Please can you elaborate on this remark please?

If power-management of the device would be delegated to the guest, and the
guest forgets to enable device power before accessing the device's registers,
this could lock up the system, and thus disturb both the host and other guests.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH] vfio: platform: Fix using devices in PM Domains

[Auger Eric]

Hi Geert,

On 13/04/18 11:19, Geert Uytterhoeven wrote:
> Hi Eric,
> 
> On Fri, Apr 13, 2018 at 11:14 AM, Auger Eric <eric.auger@redhat.com> wrote:
>> On 11/04/18 11:24, Geert Uytterhoeven wrote:
>>> If a device is part of a PM Domain (e.g. power and/or clock domain), its
>>> power state is managed using Runtime PM.  Without Runtime PM, the device
>>> may not be powered up, causing subtle failures, crashes, or system
>>> lock-ups when the device is accessed by the guest.
>> the device may not be powered up/clcoked or power/clock may be switched
>> off while the guest uses it.
>>>
>>> Fix this by adding Runtime PM support, powering the device when the VFIO
>>> device is opened by the guest.
>>>
>>> Note that while more fine-grained power management could be implemented
>>> on the guest side, if exported, this would be inherently unsafe, as
>>> abusing it may kill the whole system.
>>
>> Please can you elaborate on this remark please?
> 
> If power-management of the device would be delegated to the guest, and the
> guest forgets to enable device power before accessing the device's registers,
> this could lock up the system, and thus disturb both the host and other guests.
Wouldn't you need to assign another device or use para-virt to allow the
guest to perform this power management control? I think you can remove
this paragraph from the commit message.

Thanks

Eric
> 
> Gr{oetje,eeting}s,
> 
>                         Geert
> 

Re: [PATCH] vfio: platform: Fix using devices in PM Domains

[Geert Uytterhoeven]

Hi Eric,

On Fri, Apr 13, 2018 at 11:44 AM, Auger Eric <eric.auger@redhat.com> wrote:
> On 13/04/18 11:19, Geert Uytterhoeven wrote:
>> On Fri, Apr 13, 2018 at 11:14 AM, Auger Eric <eric.auger@redhat.com> wrote:
>>> On 11/04/18 11:24, Geert Uytterhoeven wrote:
>>>> If a device is part of a PM Domain (e.g. power and/or clock domain), its
>>>> power state is managed using Runtime PM.  Without Runtime PM, the device
>>>> may not be powered up, causing subtle failures, crashes, or system
>>>> lock-ups when the device is accessed by the guest.
>>> the device may not be powered up/clcoked or power/clock may be switched
>>> off while the guest uses it.
>>>>
>>>> Fix this by adding Runtime PM support, powering the device when the VFIO
>>>> device is opened by the guest.
>>>>
>>>> Note that while more fine-grained power management could be implemented
>>>> on the guest side, if exported, this would be inherently unsafe, as
>>>> abusing it may kill the whole system.
>>>
>>> Please can you elaborate on this remark please?
>>
>> If power-management of the device would be delegated to the guest, and the
>> guest forgets to enable device power before accessing the device's registers,
>> this could lock up the system, and thus disturb both the host and other guests.
>
> Wouldn't you need to assign another device or use para-virt to allow the
> guest to perform this power management control? I think you can remove
> this paragraph from the commit message.

Yes, that needs additional exporting of power-management (cfr. the "if
exported").
I just wanted to mention the caveat, as eventually, someone is going to
suggest doing this ;-)

But if you think the comment should be removed, I can of course do so.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds