Re: [PATCH] PCI / hotplug: Propagate the "ignore hotplug" setting to parent

[Bjorn Helgaas]

On Wed, Apr 15, 2015 at 12:01 PM, Rafael J. Wysocki <rjw@xxxxxxxxxxxxx> wrote:
> On Wednesday, April 15, 2015 05:36:42 PM Rafael J. Wysocki wrote:
>> On Apr 15, 2015 3:16 PM, "Rafael J. Wysocki" <rafael@xxxxxxxxxx> wrote:
>> >
>> > On Wed, Apr 15, 2015 at 2:55 PM, Bjorn Helgaas <bhelgaas@xxxxxxxxxx>
>> wrote:
>> > > On Tue, Apr 14, 2015 at 8:03 PM, Rafael J. Wysocki <rjw@xxxxxxxxxxxxx>
>> wrote:
>> > >> On Tuesday, April 14, 2015 12:28:12 PM Henrique de Moraes Holschuh
>> wrote:
>> > >>> On Mon, Apr 13, 2015, at 11:23, Rafael J. Wysocki wrote:
>> > >>> > From: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
>> > >>> >
>> > >>> > Refine the mechanism introduced by commit f244d8b623da (ACPIPHP /
>> radeon
>> > >>> > / nouveau: Fix VGA switcheroo problem related to hotplug) to
>> propagate
>> > >>> > the ignore_hotplug setting of the device to its parent bridge in
>> case
>> > >>> > hotplug notifications related to the graphics adapter switching are
>> > >>> > given for the bridge rather than for the device itself (the need to
>> > >>> > be ignored in both cases).
>> > >>>
>> > >>> I do apologise if this is a stupid question, but is there any chance
>> the
>> > >>> bridge will be connected to other devices that do require hotplug
>> handling,
>> > >>> and not just to the GPU?
>> > >>
>> > >> The bridge is actually a downstream PCIe port holding the GPU, so no.
>> :-)
>> > >
>> > > When radeon/nouveau call pci_ignore_hotplug(), that's the case, but in
>> > > general all we know is that pci_ignore_hotplug() receives a PCI
>> > > device.  We don't know whether it's PCI or PCIe.  In the hotplug
>> > > topologies I'm familiar with, a bridge only leads to one hot-pluggable
>> > > slot, but I don't remember anything that would guarantee that.  For
>> > > PCIe, I think there can only be one slot, but for PCI I would think it
>> > > possible to have one bridge leading to several hotpluggable slots,
>> > > with the hotplug controller(s) being separate from the bridge.
>> >
>> > Realistically, the switcheroo people are the only users of
>> > pci_ignore_hotplug() today and if somebody wants the hotplug events to
>> > be ignored for him and perhaps not for someone else on the same
>> > bridge, then something is seriously broken about that system anyway.
>>
>> There may be a problem if somebody *already* calls this function for the
>> parent bridge (which some callers do ISTR) in which case the setting will
>> be propagated unnecessarily.
>>
>> That said, since all of the existing callers of pci_ignore_hotplug() appear
>> to need to set the flag for the parent too, IMO it's better to put that
>> logic into the interface instead of duplicating it in the callers. So, why
>> don't we add a second argument indicating whether or not to propagate the
>> setting?
>
> My previous message didn't seem to reach the mailing lists, so above is a
> quote from it and below is a new version of the patch with the additional
> argument to pci_ignore_hotplug().

I can't remember why you added the "propagate" argument.  Both of the
existing callers supply "true".  How would a new caller decide whether
to supply "true" or "false"?  The question of whether hotplug
notifications go to the device or the bridge seems like a property of
the platform/ACPI design that the driver wouldn't know about.

For pciehp, the notification always goes to the bridge, and we check
all the devices on the secondary bus for "ignore_hotplug".  If it's
feasible, it would be nice if acpiphp could use a similar design that
doesn't require the propagation.  That would reduce code differences
and also remove a little ambiguity about what "bridge->ignore_hotplug"
means (does it mean "ignore hotplug of devices below the bridge" or
"ignore hotplug of this bridge"?)

> ---
> From: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> Subject: PCI / hotplug: Propagate the "ignore hotplug" setting to parent
>
> Refine the mechanism introduced by commit f244d8b623da (ACPIPHP / radeon
> / nouveau: Fix VGA switcheroo problem related to hotplug) to optionally
> propagate the ignore_hotplug setting of the device to its parent bridge
> in case hotplug notifications related to the graphics adapter switching
> are given for the bridge rather than for the device itself (the need to
> be ignored in both cases).
>
> Link: https://bugzilla.kernel.org/show_bug.cgi?id=61891
> Link: https://bugs.freedesktop.org/show_bug.cgi?id=88927
> Reported-by: tiagdtd-lava <tiagdtd-lava@xxxxxxxx>
> Signed-off-by: Rafael J. Wysocki <rafael.j.wysocki@xxxxxxxxx>
> ---
>  drivers/gpu/drm/nouveau/nouveau_drm.c |    2 +-
>  drivers/gpu/drm/radeon/radeon_drv.c   |    2 +-
>  drivers/pci/pci.c                     |   16 ++++++++++++++++
>  include/linux/pci.h                   |    6 +-----
>  4 files changed, 19 insertions(+), 7 deletions(-)
>
> Index: linux-pm/drivers/pci/pci.c
> ===================================================================
> --- linux-pm.orig/drivers/pci/pci.c
> +++ linux-pm/drivers/pci/pci.c
> @@ -4319,6 +4319,22 @@ bool pci_device_is_present(struct pci_de
>  }
>  EXPORT_SYMBOL_GPL(pci_device_is_present);
>
> +/**
> + * pci_ignore_hotplug - Ignore subsequent hotplug notifies for this device.
> + * @dev: Target device.
> + * @propagate: Whether or not to ignore hotplug notifies for the parent bridge.
> + */
> +void pci_ignore_hotplug(struct pci_dev *dev, bool propagate)
> +{
> +       struct pci_dev *bridge = dev->bus->self;
> +
> +       dev->ignore_hotplug = 1;
> +       /* Propagate the "ignore hotplug" setting to the parent bridge. */
> +       if (bridge && propagate)
> +               bridge->ignore_hotplug = 1;
> +}
> +EXPORT_SYMBOL_GPL(pci_ignore_hotplug);
> +
>  #define RESOURCE_ALIGNMENT_PARAM_SIZE COMMAND_LINE_SIZE
>  static char resource_alignment_param[RESOURCE_ALIGNMENT_PARAM_SIZE] = {0};
>  static DEFINE_SPINLOCK(resource_alignment_lock);
> Index: linux-pm/include/linux/pci.h
> ===================================================================
> --- linux-pm.orig/include/linux/pci.h
> +++ linux-pm/include/linux/pci.h
> @@ -1002,6 +1002,7 @@ int __must_check pci_assign_resource(str
>  int __must_check pci_reassign_resource(struct pci_dev *dev, int i, resource_size_t add_size, resource_size_t align);
>  int pci_select_bars(struct pci_dev *dev, unsigned long flags);
>  bool pci_device_is_present(struct pci_dev *pdev);
> +void pci_ignore_hotplug(struct pci_dev *dev, bool propagate);
>
>  /* ROM control related routines */
>  int pci_enable_rom(struct pci_dev *pdev);
> @@ -1039,11 +1040,6 @@ bool pci_dev_run_wake(struct pci_dev *de
>  bool pci_check_pme_status(struct pci_dev *dev);
>  void pci_pme_wakeup_bus(struct pci_bus *bus);
>
> -static inline void pci_ignore_hotplug(struct pci_dev *dev)
> -{
> -       dev->ignore_hotplug = 1;
> -}
> -
>  static inline int pci_enable_wake(struct pci_dev *dev, pci_power_t state,
>                                   bool enable)
>  {
> Index: linux-pm/drivers/gpu/drm/nouveau/nouveau_drm.c
> ===================================================================
> --- linux-pm.orig/drivers/gpu/drm/nouveau/nouveau_drm.c
> +++ linux-pm/drivers/gpu/drm/nouveau/nouveau_drm.c
> @@ -731,7 +731,7 @@ nouveau_pmops_runtime_suspend(struct dev
>         ret = nouveau_do_suspend(drm_dev, true);
>         pci_save_state(pdev);
>         pci_disable_device(pdev);
> -       pci_ignore_hotplug(pdev);
> +       pci_ignore_hotplug(pdev, true);
>         pci_set_power_state(pdev, PCI_D3cold);
>         drm_dev->switch_power_state = DRM_SWITCH_POWER_DYNAMIC_OFF;
>         return ret;
> Index: linux-pm/drivers/gpu/drm/radeon/radeon_drv.c
> ===================================================================
> --- linux-pm.orig/drivers/gpu/drm/radeon/radeon_drv.c
> +++ linux-pm/drivers/gpu/drm/radeon/radeon_drv.c
> @@ -453,7 +453,7 @@ static int radeon_pmops_runtime_suspend(
>         ret = radeon_suspend_kms(drm_dev, false, false);
>         pci_save_state(pdev);
>         pci_disable_device(pdev);
> -       pci_ignore_hotplug(pdev);
> +       pci_ignore_hotplug(pdev, true);
>         pci_set_power_state(pdev, PCI_D3cold);
>         drm_dev->switch_power_state = DRM_SWITCH_POWER_DYNAMIC_OFF;
>
>
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/